Slashdot Mirror
Microsoft Wants $1M of Larry Ellison
Jabberwocky
writes "Well, it didn't happen overnight, but
Microsoft
claimed that on Wednesday it will be able to
demonstrate that it can indeed meet the $1 million
challenge issued by database arch rival Oracle in November
1998. The whole thing hinges on whether or not "anyone
using Microsoft's SQL Server with a 1 terabyte TPC-D
database to run a standard data-warehouse business query
within 100 times of Oracle's best published
performance." Microsoft is aparantly going to give it
a shot using SQL Server 7.0 which it just released. "
This will be amusing if this happens...
Someone dig up that original challenge. Does Microsoft intend to actually show that their product is better than 1% of the speed of Oracle? That's what the challenge sounds like. MS would have to be an idiot to take this challenge, all they are proving is that Oracle is less than 100 times faster than the MS product. I can see the marketing now: "Oracle is only 50 times faster, not 100 times faster than our product!"
I'm wondering.. is this demonstration going to be on.. videotape? We can -certainly- trust them to show us a good tape, after all, they have quite a history..
I have seen this term before ... anyone want to clue me in?
So Micro$oft is gonna come out and say "we're not 100 times slower, we are only 99 times slower".
Way to go!
If M$ doesn't beat Oracle, it's a lose-lose situation either way. M$ will prove their product is inferior if they even win the million dollars but don't beat Oracle.
After the test....
"Hey, did I just see the word 'Oracle' blink on and off the TV screen???"
hehe
You could see this one coming a long ways away.
Micros~1 has been working on this response since about one week after Larry went and opened his big mouth. So its taken them this long to come up with an optimized dataset that can run at about the same speed as the oracle test.
My prediction is that orable will claim that the challenge is over, and micros~1 will claim victory. There will be lots of press announcements on boths sides, some more name calling, and maybe another similar challenge in the future.
As long as this challenge tied up micros~1 resources for 7 months its better in the long run for all the rest of us. Less bugs were written in that time, it all went into the FUD machine which in the long run is an excercise in futility.
Check out the results on the 400 MHz G3 in PC World.
The only problem is that Steve claimed it's TWICE as fast.
http://www.datamation.com/PlugIn/issues/1996/april 15/04bevalgls.html
On-Line Analytical Processing (OLAP) is a category of software technology that enables analysts, managers and executives to gain insight into data through fast, consistent, interactive access to a wide variety of possible views of information that has been transformed from raw data to reflect the real dimensionality of the enterprise as understood by the user.
OLAP functionality is characterized by dynamic multi-dimensional analysis of consolidated enterprise data supporting end user analytical and navigational activities including:
calculations and modeling applied across dimensions, through hierarchies and/or across members
trend analysis over sequential time periods
slicing subsets for on-screen viewing
drill-down to deeper levels of consolidation
reach-through to underlying detail data
rotation to new dimensional comparisons in the viewing area
OLAP is implemented in a multi-user client/server mode and offers consistently rapid response to queries, regardless of database size and complexity. OLAP helps the user synthesize enterprise information through comparative, personalized viewing, as well as through analysis of historical and projected data in various "what-if" data model scenarios. This is achieved through use of an OLAP Server.
Oracle's query ran in 71 seconds.
e tfinity1tb_011199.html
Their challenge is "can you do it in under 2 hours?" -- 100 times longer.
IBM took that challenge. Their query ran in less than 1 second, on Intel machines running Windows NT.
Check it out at http://www.software.ibm.com/data/db2/benchmarks/n
On-Line Analytical Processing (OLAP) is a category of software technology that enables analysts, managers and executives to gain insight into data through fast, consistent, interactive access to a wide variety of possible views of information that has been transformed from raw data to reflect the real dimensionality of the enterprise as understood by the user.
r il15/04bevalgls.html
OLAP functionality is characterized by dynamic multi-dimensional analysis of consolidated enterprise data supporting end user analytical and navigational activities including:
calculations and modeling applied across dimensions, through hierarchies and/or across members
trend analysis over sequential time periods
slicing subsets for on-screen viewing
drill-down to deeper levels of consolidation
reach-through to underlying detail data
rotation to new dimensional comparisons in the viewing area
http://www.datamation.com/PlugIn/issues/1996/ap
OLAP is implemented in a multi-user client/server mode and offers consistently rapid response to queries, regardless of database size and complexity. OLAP helps the user synthesize enterprise information through comparative, personalized viewing, as well as through analysis of historical and projected data in various "what-if" data model scenarios. This is achieved through use of an OLAP Server.
In the last year or so, some "methods" have become commonplace in world of TPC-D benchmarking. The upshot of these "methods" is that they make use of precomputing results, outside of the benchmark time. These "methods" are completely legal within the rules for the TPC-D benchmark (unlike Spec, TPC doesn't have any meta rules like "no cheating by following the rules but not the spirit").
The end result is that any vendor's TPC-D numbers for the last year or so are meaningless. For MS to use TPC-D to try to win the bet is typical MS distortion.
For more discussion of these TPC-D "methods" look through the last few months of discussion in comp.arch via dejanews, there is probably a lot of discussion in the database newsgroups too, but I don't normally hang out there.
that if they won, the $ would be donated to charity. Nice anti-MS thought though...
The whole point is, Oracle said, "look! you can commute in a Ferrari!" (the Sun machine they ran it on costs MILLIONS...) If Microsoft volleys back and can show a single server ($50000) even making a close approach to Oracle's Ferrari, the point is moot, and Microsoft, IMHO, has won. Who TF can even afford the Sun machine they tested on???
the microsoft press site says the event was scheduled for March 16, 1999. That was Tuesday. Why are there no results? The microsoft site just has some 30 minute advertisement for SQL server 7.0 (CHEESY)
1: The challenge SPECIFICALLY refers to SQL Server 7.
2:You've obviously never even USED SQL 7. The thousands of companies involved in the beta have great evidence that leisure suit Larry will lose his million...
NT will have to run on the multi-million dollar HW that Oracle chose to run it's ridiculous challenge on...
While I was at MS, the SQL Server group claimed that it was IIS's fault :)
If you throw a G3 chip to the ground, and merely drop the PII, the G3 will be substantially faster. Hey, this benchmark is no less valid than some of the other ones used.
I'll bet those 32 boxes were a lot cheaper than the machine from Sun
that Oracle used. I cannot imagine IBM's system costing more than 1M.
How much money can MS continue to pour into products that are not profitable for them, like SQL Server 6.5 (a real dog)?
This revenue streams in from the monopoly they have on the desktop and then streams out to loser projects like WinCE and SQL Server.
Now that SQL Server 7.0 seems to be a better product, it might actually not be a losing proposition like 6.5 was. But the only reason that it exists is due to the MS monopoly and marketing.
Have YOU used Oracle?
Okay, so a lot of you will flame me for this, but hey, who cares, I have to respond. To be honest with you, one of the main reasons why SQL Server can't stand up to Oracle in terms of up time is because the kind of people using SQL Server don't know what the hell they are doing. I'm sure if you were to take your average SQL Server Admin and ask him to Admin an Oracle box, you'd find that they have even LESS success in keeping that up.
However, i do know a very intelligent gentleman who is a vvery highly skilled SQL Server admin. He's been running a SQL Server on a rather nice machine for a e-commerce site that gets quite a bit of traffic. That machine has been up every day for the last year except for 1) when it was moved from one building to another and 2) when the raid controller fried and had to be replaced. Our web servers crash all the time because the web coders are typical MS (L)users, but that SQL server stays up all the time.
I'm not saying that SQL server can compare in performance, but I am saying that on a properly configured machine, administered by a real admin, it can have very nice uptimes.
sam wilson
swilson@bugs.cs.wcupa.edu
Well, just like in the antitrust case, the setup is behind closed doors.. so they probably have a subset of sqlserver running and coded the sample in directly and say its sqlserver that everyone can buy, I would like to see then use an out of the box installation.. I don't trust ms...
the other thing is, it will be interesting to see what data they are going to use and whether they will have a pre-run test and just announce results..
Let's not let facts get in the way of a good argument...
Well thought out.
So who's better? The Guy in the corner with the abacus?
Microsoft? Postgresql? A GNU/Linux box with PERL searches?
None of it is open-source, so who gives a crap?! Don't talk to me unless it is OSS and it says Linux, Linux, Linux on it.
ioqweuhf!
the terms of the challenge state that it is a full
tpc-d benchmark, auditided by a certified tpc-d
person. only query 5 is used for the measure though.
We all know that with current database
engines TPC-D means nothing. Oracle 8i
precomputes queries made by the benchmark.
--P
The challenge isn't just about the software, it's also about the hardware that the software can run on. SQL Server 7.0 is limited to whatever NT can run on (what... 16 processor Xeon box?). Even if Oracle was slower, it can still run on the big iron. The contest doesn't say anything about price/performance.
When absolute performance matters, Oracle will always whomp SQL Server. Then again, DB2 whomps them all.
With advents like beowulf, big iron is having less of an impact these days, maybe MS will get SQL Server to cluster some day...
I imagine that M$ will use SQL 7.0 on some v.big alpha system running a beta of NT 5.0 -- and then just tweak it for speed.
Don't leave out Steve Ballmer! He is to Bill Gates what Darth Vader is to the Emperor!
Actually they say they have over 300 million rows of data... still no big deal. We host approx. 10 million rows of music data and many GB of images and sound clips on our server, with far less hardware than terraserver. Even so, MS SQL 6.5 really struggled to handle it.
We stepped back to using MS Access for the music data. The site is far easier to manage and-- guess what-- pages load faster.
I can't wait until we are on Linux with Oracle8...
the terms of the challenge are that a full, audited 1TB tpc-d test is run, with a certified tpc-d person as auditor. The challenge only uses query-5 from the test, but that it gets compared to oracles 'BEST' published result for this. Oracle look to have been preparing for this too.
oracle own result is 100 times better than when the challenge was issued , but the database load time has nearly doubled, so some cunning indexes are now in place.
when the challenge was issued:
database load ~30hours , query5 = 71.4 sec
january
database load 38hours, query5 = 4.3 sec
february
database load 47hours, query5 = 0.7 sec
the target moves, and microsoft now only have to be as good as Oracle were when the challenge was issued.
if MS can cache 1GB, of a six table join, with arbitrary date and region selections then they deserve some praise.
EMC is at the dead end because network-like
storage is the answer to the performance needs
of upcoming servers.
But their current controllers are quite nice.
I hope EMC can capitalize on current revenue
and bring up a good aside-from-datapath,
switch-based controller.
--P
Shouldn't it say "GNU/Linux, GNU/Linux, GNU/Linux" ?
I think you've missed the boat on this one.
The "Terra" in TerraServer doesn't have anything
to do with the data size. It has to do with the
Earth. Terra == Earth. Besides who the hell
doesn't like to look up their hometown on MS'
terraserver?
Funny. According to Microsoft, they already have met the challenge. In fact, if you go to http://www.microsoft.com, you'll be able to read their propaganda about how well they matched Oracle's mark. If this is true then then Oracle really shot themselves on the foot (mouth, etc.). To get even just half the performance of Oracle at less than 1/16th of the original cost is quite an impressive performance. Oh by the way, who would be crazy enough to buy a SQL Server 7 to serve up terrabytes of data? Then again, who would be crazy enough to pay so much money when a comparable performance can be had for cheaper from other DBMS vendors?
Nice... I had no idea that MySQL was so capable...
Makes a great competitor! (To Access)
Doh?! Do you really think Microsoft is that stupid? Why would you even try if you can get a comparable performance at a fraction of the cost? Surely, Microsoft would have seen that ploy from Larry... By the way, Ellison isn't anymore different that Gates. Bill just has more money.
Actually, you're wrong.
I have every hunch that these guys may goof up the contest as it their in-born-thing behavoir. Also budget-tied corporates weigh price/performance where it is likely that M$ will market happily amd make millions, sorry billions. I say Great brute ideas are the hallmark of Bill Gates.
Who said SQL Server 6.5 wasn't profitable? More well known companies use Oracle, but small buz and startups go for what they can afford.
IMO - Oracle feels that they have a monopoly on the SQL database market and that is why they charge so much. Other example is Photoshop. Why the hell does it cost so much? Cuz historically it's what people use.
Good points all around fellas!
But lets cut to the heart of the matter:
Who's penis is bigger, Bill Gates' or Larry E's?
BTW: If they're the same size, than we'll have to put the two A-Holes in a steel cage wrestling match (demonic voice on) to the death (demonic voice off).
I love that. I ask permission to use it...
follow this link: Microsoft and Hewlett Packard Meet Oracle's Database Challenge, Providing Equivalent Speed, Lower Cost, Higher Value for Customers
if you want to read the details instead of just flaming about how much you hate/envy M$.
it says (in part):
=========================
This month, as part of the premiere event of the "getting Results" Web cast series, Microsoft released its response time to the same query issued in the Challenge, announcing that it had achieved an execution time of 1.075 seconds on query five, significantly faster than Oracle's original mark and on par with Oracle's recent result of 0.7 seconds. Microsoft's results were achieved using Microsoft SQL Server 7.0 Enterprise Edition for a total cost of less than $600,000.
"Our solution not only matches Oracle's performance, but it does so at about one-sixteenth the price," says Leland. "It demonstrates that Microsoft offers powerful data warehousing and business intelligence solutions at a cost of ownership that is in line with real-world business realities. That's the core of our approach: to provide business solutions that drive down the cost of ownership and maximize return."
With a big enough Beowulf cluster you might be able to beat the challange on Intel. Does anyone know if Avalon is faster than an Ultra Enterprise? I'm thinking it is but I don't know about I/O. (Opps, just remebered that Avalon is Alpha but you could build one out of Xeons). This would be some great publicicty for Intel if they did it.
Amusing, but shouldn't that be "MICROS~1"? e.g. it not only can't handle names longer than 8 letters, but it's also not case sensitive...
As http://www.techweb.com/wire/story/TWB19990317S0002
mentions near the end of the article, Microsoft's
claim to have beaten the challenge is false since
they did not perform the specified query against the
specified tables. Making queries against a few huge
objects (which is what MS did) is not the same as
making queries against millions of tiny records (which
is what the challenge was). As usual, Microsoft is
trying to lie its way into success.
Neither company gave oout specifics on the machines and their software. They both optimized it to give a 1 seconf result.
Personally, I would like to see them you a dual proccessor machine with a 200 GB databse with base installs. Its going to be slow but you'll see TRUE results.
Never take benchmarks for face value unless you do them yourself.
Can anyone guess what would happen if M$ where to run the test query 10 mins before anybody comes through the door - and then runs it again having done nothing in the meantime.
I know! I know! It crashes so they're still re-installing the OS when people arrive!
Wow! You're right. The same way they always 'win' - a month late, changing the query and posting their 'interpretation' of the results on their own website for slobbering M$ lovers to lap up like the fecal runoff that it is.
Didja buy Bill's new 'visionary' book yet?
Point being, they tried the 'real' query - it failed. They changed the query until they created something that sounds impressive, but in reality is just more PR. Think "Terra Server".
They still cannot compete with the big boys as far as technology goes. Marketing is another matter.
excuse me but where did you get that from?
the actual quote from the article is:
---------------------------------------
Oracle officials said Tuesday that Microsoft and HP did not meet the challenge, which closed Feb. 22.
"They're a month late, and secondly, this is not the challenge," said Mark Jarvis, senior vice president worldwide marketing at Oracle, in Redwood Shores, Calif. "They have a 1-terabyte database, but did not run the standard TPC-D query against it."
---------------------------------------
it doesn't say a thing about what was in the database.
can you give us a source for this info. or are you just assuming that M$ cheated?
The original poster had it straight.
Some people carry around "DEATH TO BILL GATES" placards so big that they can't see around them.
what are you, on crack?!?!?!?!?!
they never actually ran the query! ms skirted around the issue, being the flamers that they are.
i know that this challanges your preconceptions but it looks like M$ (and HP (a big UNIX fan, let us not forget)) did not cheat.
the M$ presentation asserts that they loaded the standard TPC-D 1TB database and run query-5 in it's standard form.
if you have evidence that they did not do load the std. DB and run the std. query then perhaps you can share it with us.
those of you who are familiar with databases and benchmarking their performance know that this query used to take a lot of time to execute -- but then people started using "materialized views" (i.e. precomputing stratigic pieces of information). that's why the load times for TPC-D have gone through the roof; they're grinding on the data in order to speed up these queries.
this stuff isn't magic, it's s/w architecture.
M$ has simply adopted the same techniques that Oracle is using to speed up query-5.
Larry is not sane. that's the fact. But I do agree with you that even Larry won't do such a stupic thing ;)
seems like microsoft/HP didn't cheat
Would he (Ballmer) finish him (Gates) off then?
If sql server can compete with a 10million dollar database with 50,000$ hardware then I am buying it. Those guys at microsoft really lead the industry with bright innovation that its competitors can not even approach. I used both oracle databases and sql server on NT and microsoft's is always the most scalable and fastest around. With the new olap, oracle can finally go down under and we can have one united soluton by one company which can innovate better. No one cares if oracle is 10 faster on high end databases. People take notice when sql server ccomes close on low end hardware. It shows that of sql server can do these impressive results at the the low end, think about what it can do about the high end. Did any of you read about microsofts "scalabilty day". Bill Gates showed a 10,000$ server compete with a IBM mainframe and the NT server did about 1.5 million transaction in just a few hours. THIS WAS ON LOW END HARDWARE! All the consults finbally decided that unix was dead and NT is the feature and I can't wait untill that day comes.
ORACLE $1 MILLION DATABASE PERFORMANCE CHALLENGE
Is Microsoft SQL Server 7.0 100 Times Slower than Oracle8i
REDWOOD SHORES, Calif., Nov. 23, 1998-Oracle Corp. today detailed its
data warehouse performance challenge issued at Comdex last
week by Oracle Chairman and CEO Larry Ellison. Oracle Corp. will pay $1
million to anyone -- including Microsoft -- who can demonstrate
that SQL Server 7.0 is not 100 times slower than the Oracle database
when running a standard business query against a large database.
The challenge pits the Oracle8i Internet database against Microsoft's
new SQL Server 7.0 database. In the recent announcement for SQL
Server 7.0, Microsoft claimed that SQL Server 7.0 now has the ability to
run OLAP (on-line analytical processing) queries against terabyte
data warehouses. However Microsoft did not publish any data warehousing
performance results. That makes it impossible to compare SQL
Server 7.0 query performance to Oracle query performance. This challenge
was conceived to remedy that problem.
Specifically, the challenge is based on a single database query that
answers the common business question: "How much of last year's revenue
was shipped nationally vs. internationally?" To be fair, Oracle has
based the challenge on query number five of the industry-standard TPC-D
1
terabyte database benchmark. If SQL Sever 7.0 comes within 100 times of
Oracle's best published performance for this query, Oracle will
award the one million dollars.
Additional details of the challenge's specifications are available via
Oracle's Web site at http://www.oracle.com/challenge. Currently Oracle
holds every database performance record in the world including data
warehousing TPC-D, transaction processing TPC-C, and ERP
benchmarks from SAP, PeopleSoft and Baan."
About Oracle Corporation Oracle Corporation is the world's leading
supplier of software for information management, and the world's
second largest independent software company. With annual revenues of
more than $7.5 billion, the company offers its database, application
server, tools, and application products, along with related consulting,
education, and support services in more than 140 countries around the
world. For more information about Oracle, please call 650/506-7000.
Oracle's World Wide Web address is (URL) http://www.oracle.com/.
M$ does show the physical machine that they ran the demo on and they provide the basic details about it's setup.
it's way typical for a TPC-D benchmark. a generic HP LXr 8000 with 4, 400 MHz Xeons and whit a ton (2.3 TB) of RAID arrays.
M$ claims to have loaded the vanilla TPC-D 1TB database and ran query #5. they quoted an average response time of 1.1 seconds.
they also ran some other queries against the same database to show off some other features.
They just keep innovating and innovating and don't know where to stop. I cant wait to take the plunge into sql server and you should take it too!
Microsoft and Hewlett Packard Meet Oracle's Database Challenge, Providing Equivalent Speed, Lower Cost, Higher Value for Customers
and if you want to see the machine and watch them click the buttons then go here
please don't assume that i haven't read the challange; in fact i have.
the challange was to run TPC-D query 5 against the TPC-D 1TB database.
all of the data that i've seen indicates that this is exactly what M$ did.
How impressed would you be if I'd said I would memorize an entire 1 Tb database for 600G's, then spend all of the money in a bar.
Does anyone remember the ads a couple of years ago
that stated "a billion transactions daily" when
Microsoft was trying to scam Tandem for their
transactions processing expertise. The vaporware
was dubbed Tandem NT in the ads that I saw in the
Wall Street Journal.
I'll believe it when it is reported but I am
not holding my breath!
other's have sited the URLs.
if you want to know how TPC-D works then you could visit thier site at:
http://www.tpc.org/
you can download the details of the TPC-D spec at:
http://www.tpc.org/dspec.html
you'll find that the query that Oracle spec'd was simply Q5 of the TPC-D benchmark.
Oracle referes to version 1.3.1 of TPC-D in their challange -- which is cool because Q5 hasn't changed between V1 and V2 of TPC-D.
message says it all.
How does clustering fit into the speed picture?
I sorta figure that you'd need a distributed
database to get the most out of that..
Anyway, wouldn't that make for a higher volume /
lower reaction time payoff?
100 times faster is more than 100% faster ;-)
TAG TEAM! Larry Ellison (He's on the Apple Board, isn't he?) and God^H^H^H Steve Jobs vs. Steve Ballmer and Bill Gates. Heheheh.
Oracle's funky test is a query that benefits from precomputed aggregates, a feature that SQL7 lacks. That's why they were so confident that microsoft couldn't meet it.
I bet SQL crashes. this will be amusing
There's no way MS could have run the standard TPC-D query - they used their OLAP product (Decision Support Services) to pre-aggregate the data.
MS DSS doesn't respond too well to standard SQL - you have to feed it MDX queries to get the results out.
Mh have not adopted the same techniques that Oracle is using:
Oracle's techniques are transparent to end users performing the queries, though additional setup is required.
Microsoft's technique is not transparent to end users (and therefore doesn't qualify to win the TPC-D challenge), and requires just as much additional setup as the Oracle approach.
Having said that, we use Microsoft's OLAP Server here, and for ad-hoc aggregate queries MDX is much nicer to work with than plain SQL.
what they didn't do was to write their press release as well as they should have.
their press release is clearly causing a lot of confusion.
i know that it's a pain in the drain to go slog through the M$ presentation of why SQL Server is the greatest product since sliced bread; but that's exactly what i did.
my 'reward' for listening to all this M$ marketing stuff was that i learned that they did in fact run TPC-D, that they used the 1TB TPC-D data set, and that they ran the Q5 query just like Oracle challanged them to do.
it's all there on the M$ web site. there are a bunch of links to it in the posts preceding this one.
Comparable performance. Get within 1%. Cognitive dissonance is happening here.
DB2 UDB enables Windows NT as a Platform for Enterprise Computing DB2 Proof of Performance January 11, 1999
this more than fits the bill.
of course they used pre-aggregation. that's how all of these vendors are getting such stellar marks in these queries.
i read the TPC-D FAQs page and TPC is 100% clear on this point:
------------------------------------------
3.10 Does TPC-D allow the definition of auxiliary objects such as views or indexes that represent pre-aggregated versions of the underlying table data?
There is no explicit discussion of specialized aggregation objects. However, like any other auxiliary data structure that is derived from the base table data, aggregation objects are a form of replication that must satisfy the requirements of Clause 1.5. Namely, they must be automatically managed by the SUT; the aggregation objects must be transparent to the query and may not be explicitly referenced; updates to the base table are reflected in the aggregation objects in the same transaction; and aggregation objects retain full ACID properties.
------------------------------------------
this is why this year's Sun/Oracle load times are so much longer than last years (for the same size data set). Oracle is pre-aggregating.
it's a level playing filed. M$ used pre-aggregation, Sun/Oracle used pre-aggregation.
If anybody cheated Oracle did cause they basically pre-calculated the query during the database load so it could answer that specific query quickly (completely contrary to the purpose of the TPC-D benchmark which was to simulate random queries). None of the MS haters thought MS would even get into the same ballpark with Oracle from the messages I saw above, but it turns out MS actually beat the Oracle time on some queries and averaged out at a neglible difference. Not bad for 1/16th the cost. Oracle withdrew their challenge cause they knew they were going down. Finally MS haters loyalty to anything non-MS is far more hollow than people like me who look at both sides and call it like they see it. If I needed a solution for a database problem like this, my money would go to MS, nuff said.
So, you're saying a tricycle competes well with a 200mph sport bike, and that customers will jump at giving up their sport bikes for tricycles because tricyles are cheap? Is that it? Hmm. Innovate this, innovation that. Does the MS crowd know what the word means? Ok, Unix being dead, you better stop using the internet as for the most part it runs on Unix (so the internet is dead?). SQL more scalable than Oracle? On which planet? This is some great stuff.
It will be sad if the MS crowd gets forced into mental institutions for their own safety- I hear they've innovated a new type of chronic delusional state that destroys the ability to reason without the subject realizing it.
From what I see here TPC allows pre-aggregration and both Oracel an Microsoft are using pre-aggregration in order to speed up their queries.
Based on Microsoft's press release and the articles at TPC it's fairly clear that TPC-D has been 'out-foxed' by systems that use pre-aggregation and that TPC is going to come up with benchmarks that help us to see the effects of these newer algorithms.
Score 1 in favour of Micrsoft.
You people who get all worked up and answer this stuff seriously should be ashamed of yourselves.
As for pre-aggregation; other posts have already shown that pre-aggregation is completly legal. Perhaps you should find the link to the TPC FAQs page that discusses this topic.
As for Oracle's 71 second query result; it was done using pre-aggregation (that's already been discussed in previous threads).
"lead the industry with innovation" ???!
what has microsoft ever created on it's own, without buying up some other company, or cloning someone else's product ideas?
where did you come up with these "speed/stability arguments?? microsoft SQL server isn't scalable at all. neither is NT.
"and fastest around" ?? --even microsoft itself admits that it isn't faster than Oracles' solution. they're only trying to prove that they are at least better than 1/100 the speed of the competition. 1/100th!
Well not really, take a look at the numbers.
9 /03-16sql.htm
Microsoft's execution time: 1.075 sec
Oracle's execution time: 0.7 sec
Microsoft's cost: $600K
Oracle's cost: ~$10 million
I would say this is pretty much egg on big-mouth Larry's face!
http://www.microsoft.com/presspass/features/199
Well not really, take a look at the numbers.
9 /03-16sql.htm
Microsoft's execution time: 1.075 sec
Oracle's execution time: 0.7 sec
Microsoft's cost: $600K
Oracle's cost: ~$10 million
So Oracle was than a factor of 2 faster but was more expensive by a factor of 16!
I would say this is pretty much egg on big-mouth Larry's face!
http://www.microsoft.com/presspass/features/199
Well not really, take a look at the numbers.
9 /03-16sql.htm
Microsoft's execution time: 1.075 sec
Oracle's execution time: 0.7 sec
Microsoft's cost: $600K
Oracle's cost: ~$10 million
So Oracle was less than a factor of 2 faster but was more expensive by a factor of 16!
I would say this is pretty much egg on big-mouth Larry's face!
http://www.microsoft.com/presspass/features/199
Well not really, take a look at the numbers.
9 /03-16sql.htm
Microsoft's execution time: 1.075 sec
Oracle's execution time: 0.7 sec
Microsoft's cost: $600K
Oracle's cost: ~$10 million
So Oracle was less than a factor of 2 faster but was more expensive by a factor of 16!
Oracle's software cost alone was greater than Microsoft's total cost.
I would say this is pretty much egg on big-mouth Larry's face!
http://www.microsoft.com/presspass/features/199
What was that about struggling to beat the 1/100th performance of Oracle? Take a look at the numbers.
9 /03-16sql.htm
Microsoft's execution time: 1.075 sec
Oracle's execution time: 0.7 sec
Microsoft's cost: $600K
Oracle's cost: ~$10 million
So Oracle was less than a factor of 2 faster but was more expensive by a factor of 16!
Oracle's software cost alone was greater than Microsoft's total cost.
I would say this is pretty much egg on big-mouth Larry's face!
http://www.microsoft.com/presspass/features/199
That was the worst kind of argument I have ever head in a loog time. MS makes ok software, not good and defenetly not great software.
They leave that to ORACLE and Sun and the rest of us.
interesting results but - read that press release again.
it says that oracle cheated by using materialized views to increase performance within TPC-D benchmark, so since the tpc-d benchmark is flawed the microsoft/hp team did a more real world reprasentative test wich is similar in spirit to the original aim of the benchmark.
your costs slightly distort the difference in that you use a five year cost of ownership for the oracle solution against a cost price for the MS/HP solution - admitedly this still leaves the oracle cost at $6M for purchase.
oracle produce a figure, and then tune it to be 100 times faster for the same cost ( ~6M )
;-)
MS/HP claim the test is flawed so they produce their own test, in the spirit of the original
this 'new' test gives half the performance of oracles new figure for a tenth of the price.
nothing proved either way, so media and marketing in both camps can say whatever they like
http://www.techweb.com/wire/story/TWB19990317S0002
"They're a month late, and secondly, this is not the challenge," said Mark Jarvis, senior vice president worldwide marketing at Oracle, in Redwood Shores, Calif. "They have a 1-terabyte database, but did not run the standard TPC-D query against it."
OK - this, to me, says
1. They're a month late
and
2. They did not run the standard TPC-D query.
What is your interpretation?
Even if SQL comes within a factor of 2 they (Oracle) can have a really cool ad:
Microsoft proves that Oracle is X times faster.
It's a win/win situation
So long as it does cluster - M$ seem to think that two nodes and no more constitutes a cluster...
Looks like Oracle lost.
When the issued the challange their query time was 71 seconds (way slow given that they were using pre-aggregation).
Microsoft's time was 1.1 seconds.
Oracle went back to the drawing board and improved their query time to 0.7 seconds.
In other words, SQL Server 7 running a dumpy little quad processor 400 MHz Xeon beat Oracles original time by a factor of 70. What's that tell you? We're looking at a marketing stunt.
Oracle had pre-aggregation in Oracle8 and Microsoft didn't have it in SQL 6.x. Larry shot his mouth off hoping that most people wouldn't know what was going on behind the curtain. Based on what I've seen on the board, Larry was right.
ps I guess I'm wrong about Oracle loosing.
One of the other posters has given us
the link to Oracle's withdrawing the
challange on Feb 25, '99.
How many times are you going to post this message? Remember you are being paid by the hour, not by the post.
But of course this is old news. The 71 second result of Q5 used pre-aggregation.
Larry can protest to the contrary all he wants but the facts speak for themselves.
Smooth move Lawrence. Try spending more time in the lab and less time in your Lear.
Keep your eye on the ball.
Oracle8 had pre-aggregation and SQL 6.x didn't. Therefore, Oracle8 stomped SQL 6.x on Q5 of TPC-D.
Now SQL 7 has pre-aggregation and it's query time is in keeping with the h/w differences between the Sun platform and the HP platform.
In one sense we could say that Oracle did SQL Server users a big favor. They pointed out a hole in the SQL product and embarrased Microsoft into plugging the hole. I suspect that this accelerated the timeframe for integrated Plato into SQL Server.
Megahertz for megahertz there's not much difference between the integer performance of Xeon and an UltrSPARC.
The big advantage to the Sun/Oracle solution is lot's-o-processors in a box (thank you Sun) and lot's-o-boxes in a cluster (thank you Oracle).
Microsoft is slowly getting it's act together in terms of lot's-o-processors (thanks Unisys, 32 processors with 12GBps of switched fabric network between them is good progress (I'll wait an see how well the NT kernal scales to 32 processors)).
But we' still waiting for that parallel version of SQL Server.
On the other hand, IBM's got DB2 and it looks like DB2 on NT rocks.
Doesn't matter if it crashes.. they will sell it anyway. I won't install sql server even if it was free, just because it runs on NT. And I don't rely on NT. :) but they'll probably loose, because of FUD and all that. I just hope that, when the day comes, they release Oracle for linux under gpl.. :)
Today the bussiness is in databases... everybody is programing in stupid VBasic against an SQL server. And M$ is attacking Oracle. Oracle is probably better (it surely is
They didn't meet the challenge...they didn't use the TPC-D query specified in the challenge (they used a modified version of it) and they're a month late. Even if you ignore their use of a different query, the price/performance is 6 times more expensive for Microsoft.
And they're proud of this?
as you will see when you follow this link
The Oracle Million Dollar Challange - Challange Specifications
Oracle did not specify the a close date for the challange. this removes your first claim.
as for your second claim; Mr. Jarvis asserts what Microsoft did not run Q5. on the other hand Microsoft asserts that they did and HP asserts that they did. the burden of proof is on the prosocution. you've offered us hearsay and Microsoft and HP are offering us direct first party.
case not proven.
the prosecution looses.
this is a joke, right? you don't actually believe that TPC specifies the TPC-D must be run "on [a] multi-million $ machine"!
do you have any clue at all as to why one of the figures of merit is stated in terms of $/transactions?
do you have even the most basic concept of what TPC-D is?
if you do then why in the world are you jabbering on about optimized datasets? you can't optimize a data set for a TPC benchmark. these are standard data sets.
do you know what OLAP is? do you know what pre-aggergation is? do you know what Q5 of TPC-D is and how database design effects this query time? do you know why Q5 runs fast on Oracle 8 and SQL Server 7 but not on SQL Server 6.x?
in short, do you understand the technical issues that drive this challange or are you just cluttering up the boards with pointless speculation?
Not surprised at all. They blame that stupid BDC thing for our continuing Exchange woes. Would we do better to buy a new Proliant with NT Enterprise so Xchange can run stand alone...
Or even better...quadruplet servers running Wolfpack...no...MS Clustering
Such nice talk...
I've used and maintained them both...and there is no comparison...I can only imagine Micro$uck has a special hack for the challenge
Ummm..the G3 is faster...
Way faster.
If you had actually read the URL you quoted here, you would have noticed that they neither ran the test on the same hardware, nor on the same query nor in the allotted timeframe, nor under supervision of an independant party It looks more like apples and oranges on Bill's face.
Rather than try to match a result in a benchmark test that was already in dispute, Microsoft and Hewlett-Packard chose to craft a solution that would more closely match actual business conditions, including the need to get maximum value from any enterprise solution.
Maybe they need to tighten them up?
0 3-16sql.htm
:)
1. I don't believe M$ - they have a vested interest.
2. I don't believe HP - see #1 above.
3. You don't believe ORACLE (or love M$ or whatever) (also vested interest)
4. http://www.microsoft.com/presspass/features/1999/
"Rather than try to match a result in a benchmark test that was already in dispute,
Microsoft and Hewlett-Packard chose to craft a solution that would more closely match actual business conditions, including the need to get maximum value from any enterprise solution. They also wanted to develop a solution that met the original intention of TPC-D. "
5. Loose means the opposite of tight.
yeah sure, and maybe some day marketing departments will learn to write clear press releases.
based on the portion of the press release that you have quoted, i can see how you might come to the conclusion that Microsoft did not run Q5.
however, based on this portion of the press release:
This month, as part of the premiere event of the "getting Results" Web cast series, Microsoft released its response time to the same query issued in the Challenge, announcing that it had achieved an execution time of 1.075 seconds on query five, significantly faster than Oracle's original mark and on par with Oracle's recent result of 0.7 seconds. Microsoft's results were achieved using Microsoft SQL Server 7.0 Enterprise Edition for a total cost of less than $600,000.
in conjunction with what i heard on the audio version of their presentation, i come to the opposite consulsion.
in any event i can't quite see your point. what exactly do you think that they did in order to 'cheat'. their story is that they loaded the standard TPC-D 1TB data set and they ran Q5.
if you have a specific objection as to what they did (or failed to do) that was 'cheating' then we can discuss it -- otherwise you place me in the position of disproving a negative.
ok, now that i've demonstrated that i'm a complete idot, make that 'proving a negative'.
I can see your point as well -
As for delving even deeper into the specifics - no thanks, I've got real work to do - I'd rather not do Oracle's work for them. I will say this (now that it's been a couple days) - lacking a rebuttal from Oracle, Microsoft did 'meet the challenge'.
Erich P
324678 Georgia Tech Station
Atlanta, GA
30332-1020
Once I recieve the check, I'll get the computer to you ASAP. I'll even throw in Next-Day air shipping for free!
-- Erich
Slashdot reader since 1997
The drive is probably full from all those
memory.dmp's they are trying to debug =)
The challenge specifically refers to SQL Server 7.
--
While for folks like me (a sysadmin for a web shop) this type of big muscle flexing seems pointless, there are folks out there who are getting terabyte-sized databases, and need a real system to manage it.
... and Oracle wants to prove that they can't.
... I don't think that it even comes close to supporting terabyte sized databases, so a comparison is kind of pointless. mySQL is fast with smaller databases (a gig or less), and that is where it is best used and considered.
Microsoft wants us to believe that NT and MS SQL can scale
As for mySQL
Oracle's whole argument is that Microsoft simply cannot scale ... not with any amount of money ... Oracle has conceeded that Microsft may be able to provide a cheaper solution, but they argue that for a high-end solution, you simply can't use Microsoft's products.
Okay, so MS says they can make a server that is 1/100 the speed of Oracle and Spitfire (or solar flare, or whatever Sun calls that beefy box) for 1/16 the price ($600,000 instead of $10mil).
Okay, so let's say that they can do that. Let's also be *real* generous and say that MS only need to increase the cost of their system by a factor of 50 to match the Oracle performance.
This means that MS has to spend $600,000*50=$30,000,000 to match what Oracle could do for 1/3 the price. And that's being generous!
MS: 1/100 the performance for 1/16 the price.
Sorry; just bugged me.
Posted by FascDot Killed My Previous Use:
Reminds ME of Steve Jobs saying ANYTHING.
Posted by stodge:
;)
If they televise the test watch out for disappearing desktop icons and changing window titles!!!
Posted by stodge:
Fraid not old boy. Well not from the way I see it. It looks to me like Oracle stacked the deck in their favour when defining the challenge to give a good result.
Microsoft then decided the test wasn't valid to the real world (which might mean they couldnt match the result, or it might not), and devised their own alternative test that gave good results.
So in my book they both won. Microsoft didn't respond to the exact challenge, but they also showed results from a similar (?) test that gave similar performance.
What would be a challenge would be to use the same hardware define by an independant third party, using *exactly* the same test as defined by said people. To me (not that I'm an expert) the results prove nothing.
Posted by Kastern:
HAHAHAHAHAHAH!
MS is really making a Stupid descicion in going against Oracle!!!! Those DA. Now Gates has gone mad.
we know they would fabricate evidence videos in a trial - we know because they did.
can you trust them running a benchmark on their own software? if nothing else, then the sql server is AT LEAST patched for this special case.
Assorted stuff I do sometimes: Lemuria.org
Oracle is trash.
Remember that what's inside of you doesn't matter because nobody can see it.
Ummm... so 1000 Gigabyte-sized images? Didn't know we'd hit that point already...
--Matthew
Who knows if M$ hacked up SQL Server 7 to take advantage of things and sway the results in their favor? Remember when graphics card manufacturers made their drivers tuned for benchmark programs?
You know what would happen. Steve would convince Larry to kill Bill and Steve would finish the exhausted Larry off! Then Steve would rule the computer industry! Woo hoo! Yikes.
It's far easier to forgive your enemy after you get even with him.
It's far easier to forgive your enemy after you get even with him.
I'll just keep using my $600 MySQL server to handle my 100 meg database for the time being, thank you. The majority of the database market doesn't need -- nor can it afford -- a server that can chomp thru 1 terabyte of data in 1 second.
No, its a reasonable multitable join on a large dataset. It doesn't seem out of line to me. They did throw a big box at it--- which MS can't do. But then, that lack of scalability to huge hardware is the point.
What, that gigabyte images were used? They're actually common in more areas than GIS: professional digital graphics processing for instance, uses files that are in the hundreds of megabytes for increased colour-depth (like above 32 bit)...
-Stu
It's the TPC-D benchmark, which is an industry standard designed to simulate datawarehousing style jobs. Microsoft don't have anything to compare to oracle's new MVIEWs, which is why they fail.
However, this isn't the point: in the midrange space (say, a few dozen gigabytes) microsoft are poised to make big inroads, mostly in OLAP where they have a relatively mature toolset available. A quad xeon by compaq is competitive with Sun's E250 and E450 boxes in this space.
Note also, however, that oracle runs, and runs well, on NT. An oracle engineer told me the other week that they don't really care about the OS at this stage, they just want to sell database products while they complete their preparation for Raw Iron.
price for price, you can buy a dual PII-400 from dell for the same price as a top-end G3, and the NT box will win for any real-world tasks like photoshop, CAD, rendering, serving, whatever.
I bought my mac (7600/132) when i was at college 'cos i needed the high floating-point performance (running mklinux), but apple these days are trading entirely on past goodwill and slick marketing.
unlike open source advocates, i have no problems with my ego or my sense of self worth. so i'll take the $$$ instead, ohyes.
Well, it ought to be more fun than the Lewis / Holyfield contest, anyway...
:-)
I bet is that MS will do it, although possibly not by very much. Seemed like a hell of a boast on Oracle's part...
-----
I'm not certain, but I think the benchmark test (which they did publish) does all sorts of stuff that MySQL can't even do.
Like, transactions, select into, select where in, and so forth.
Much (most?) of the reason MySQL is so fast is that it doesn't need to log everything it does, because it doesn't let you roll anything back.
-----
What, you mean you don't have one at your home?
HELLO. Give any software company a few months and they will be able to make their software faster JUST for this specific time. Sure, tweak source until you get it the fastest possible, tweak source, hardware, hardware, source. As far as I'm concerned, this isn't legit, or Oracle should get 6 months to make their software faster too.
Blah.
From microsoft's press release on this it says they used 560 disk drives to do the test. Holy shit. That's a lot of drives...
Still, it looks like MS SQL Server proved Oracle wrong (as I was pretty sure they would). But I'm afraid it looks like they used a modified SQL Server to do it. I still say that it's because of SQL Server's heritage (Sybase) that it's fast. I'd be willing to be money that Sybase would kick Oracle's butt on this test if they used materialised views (and equivalent hardware) too.
Matt.
Matt. Want XML + Apache + Stylesheets? Get AxKit.
The Ultra Enterprise 10,000 system is also known as the "Starfire 10,000". We have at least 6 of them. (I know there are a few more in our Dallas data center, but I don't know how many.)
- Randy
- Kate
"DNA is life. The rest is just translation."
I didn't have any problems connecting... how many times did you try? They just might be having server load problems.
-- perl -e'print pack"H*","6e656d6f406d38792e6f7267"'
And still using TPC-D, as well as claiming Microsoft has not responded, while Microsoft claims they worked in co-operation with Oracle to devise a fair challenge.
This whole mess is nothing but PR. Why can't they just rely on the systems proving themselves in the field?
-- perl -e'print pack"H*","6e656d6f406d38792e6f7267"'
M$ is happy because it says that Oracle is treating it as competition, but, when it scomes down to it, the system M$ is going to use is going to 1/16 of the the price of the Oracle solution (admittedly a lot lower) but is going to struggle to beat the 1/100th performance that Oracle wants?
Leave it to M$ to distort the truth.
- (c) 2018 Hank Zimmerman
Umm, IBM wasn't using SQL Server 7.0, like the oracle challenge states...
As another note, for that it only took 128 Xeon processors, not too cheap in its own right. Finally the Sun or SGI solution will be a hell of a lot more stable...
wheee...
PostgreSQL is already available under a Free Software license. Better still, it's a BSD/XFree86 style license that appears to be compatible with the GPL.
M$ doesn't have a chance, unless they ran it on a Compaq AlphaServer 8400 (or GS140). Then they might have a chance, of course Oracle could then just run their database on the same machine...
JMHO,
rbf
--
LONG LIVE ALPHA!!!
Seems from the responses that nobody noticed this already occured yesterday.
The result of 71 seconds was from one of Oracle's initial results. They then optimized their database to work better with the specific benchmark and recorded a result of 1 second.
Microsoft is claiming that they worked with HP, didn't optimize their database to specifically work well on that one single benchmark, but instead used real world business ad-hoc type queries...
And achieved an average of a 1 second result.
Thus achieving very similar performance to the solutions from Oracle and IBM for 1/16th the cost.
Now *THAT* is the competition that Microsoft brings to this market.
And Larry Ellison can eat his shoe now.
But that wouldn't be accurate either.
Database configuration is seldom so easy. A search on a table could take anywhere from a second to two minutes depending on what sort of indexes you have setup and other optimization techniques.
yes, they both probably optimized, but in a sense that's how it's going to be done in the real world as well. Nothing wrong with using your tools to optimize the queries.
Now if they optimized the data, or optimized the engine to work with only that subset of data, I agree that'd be cheating.
You don't think this is going to make Oracle think twice about their next version, as well as their pricing structure?
I think you misunderstood Ellison's challenge.
He wasn't saying that Oracle itself was better software. He was saying that Oracle is capable of running on hardware which will more than beat what SQL Server will run on.
Ellison is a *HUGE* proponent of huge massive servers that cost a lot of money, instead of clusters of smaller cheaper servers.
-Jake
--
Jake
if(is_oracle_challenge(parsetree)) do_fast_oracle_query();
:-)
It wouldn't be the first time a commercial product contains benchmark-specific optimizations.
#1 the response period ended already
#2 they are probably using a modified version of nt/sql 7
#3 they are probably going to use a multiplier to argue that the cost benefit ratio is better (i.e. they shrink the dataset to get closer results to hide its scalability problems)
#4 there's still no mention of the tpc-d benchmark
#5 the microsoft web site that's included in this link is slow as shit! i bet it goes down for the event!
go kick ass larry!!
"The lie, Mr. Mulder, is most convincingly hidden between two truths."
--
And Justice for None
I could use $1 Million.
It probably cost them tens of millions of dollars to develop their new and "improved" SQL server. What's $1 million back?
Besides, Oracle probably came up with some funky test that will most definitely slow to a crawl on M$'s server... It's rather comical.
Oh, and anybody acually register and SEE the results of the competition?
Live!
Before your very eyes!
The Man who made the statue of Liberty dissapear, presents the latest Microsoft benchmarks!
Both Gates and Ellison are manipulative megalomaniacs.
Hell, I say throw them AND Steve Jobs in a ring and let's have an old fashiond Texas Death Cage match and kill the survivors!
No, no..you didn't read the last part of my first post. We kill the survivors at the end, thereby eliminating egomaina, megalomania and freeing the minds and hands of computing forever...hahahahaha
MS's position seems to be that Oracle cheated when they ran their second mark off of pre-cached results, so that made it OK to do it too. Great, I'm impressed.
I was much more impressed by Oracle's original 71 second mark. That was based on really jamming through 1 Tb of data. When MS publishes a result doing the query the way they claim it's supposed to be done, then we'll have some news.
---- "If we have to go on with these damned quantum jumps, then I'm sorry that I ever got involved" - Erwin Schrodinger
It's been used on linux-kernel this week. Along with the fact that when Microsoft splits, we will have Micros~1, Micros~2, Micros~3 and Micros~4. (I think those should have the extension .com though.)
... the bait hasn't been taken until now. What's
slashdot coming to?! I expect cats and dogs raining outside any minute...
Stickboy
-- MS lackeys not allowed.
--- "The problem is not that the world is full of fools, it's that lightning isn't being distributed correctly." -- Mar
100 times faster is more than 100% faster
But, it is 1% as fast which is what he said...
Stickboy
--- "The problem is not that the world is full of fools, it's that lightning isn't being distributed correctly." -- Mar
its a self fulfilling prophecy
=)
Food: It's whats for dinner
The challenge, as I recall, was specifically directed at SQL Server v7. The fact that IBM's DB2 system sticks it to Oracle is nice, but inconsequential :)
:)
And as far as the pricetag for the server hardware... it's price/performance, not the actual pricetag, that will probably influence a lot of big corporate buyers. If you REALLY NEED a DBMS that can do a query over a terabyte of data in 1 second or less, you will have a budget to get whatever ballsy hardware you need
Long live big iron.
Nothing worth doing is worth doing today.
Frankly, I wonder about that too.
Anybody know why the FSF has not adopted PostgreSQL as its standard sql database?
Used to be the FSF made decisions like that for technical reasons, because someone there or that they knew though they could do better than an existing free package...
Oracle's latest results on TPC-D are reported as 0.71 seconds (using materialized views).
TPC-D has been sussed & superseded as a reliable benchmark, they're talking about splitting it into TPC-R (just like current TPC-D) and TPC-H (which is truer to the ad hoc nature of the original test).
Just junk food for thought...
Umm... Is this a joke? How else would he implement user preferences?
A host is a host from coast to coast...
Unless it's down, or slow, or fails to POST!
I have to admit - Microsoft has considerable gall. But Oracle did set themself up for this.. if it works.. they'll have major egg on their face. Reminds me of Steve Jobs proclaiming the G3 faster than the PII.
--
The reason that MS won't run SQL Server 7 on comparable hardware is that they know the product isn't as scalable as Oracle7i. The implication is that if MS can run something reasonably close to Oracle's result on shitty hardware then they must be able top kick Oracle's ass on great hardware.
The unfortunate fact is that Oracle would probably still beat MS on the $50,000 machine. The reason they ran a multi-million $ machine in the benchmark is because that is what the benchmark test called for, not because they "needed" it.
Wouldn't that be a dishonest thing to do?
-Steve
I'd guess that Avalon can harness more raw CPU power than an Enterprise, or at least damn close. But the challenge refers to SQL Server 7, and unless SQL Server 7 has been ported to the PVM (which it sure as hell hasn't), you couldn't even run it.
Secondly, Beowulf-class supercomputer's probably wouldn't handle such a task very well anyway, as Database operations are heavily I/O bound. That's an area where the Enterprise would surely kill Avalon.
"UNIX" is never having to say you're sorry.
Of course, the license terms for SQL server state that you may not publish benchmark results without MS permission...wonder why ?
It's all very well in practice, but it will never work in theory.
Let's see... IIRC, MS compared (under oath and for the record) the download times of two 'comparably equipped machines' in the downloading and installation of MSIE vs. Navigator: The IE machine was running WinXX with a 56k modem and the Navigator machine was running Win3.1 with a 28.8 modem.
.01 capability: i.e., Oracle had the advantage with the more expensive hardware.
/. is recording the addresses of those 'MS is superior' posts (for possible later exposure of MS munchkins for what they are).
Also IIRC, they later compared a 'feltonized' machine vs. a 'non-feltonized' one to demonstrate poorer performance by the feltonized one. After it was proven that their demo of the 'feltonization' was a conglomerate of snippets from several different machines, they requested and were allowed to redo the test under the supervision of the DOJ attorneys. They redid the test, alright. But they didn't allow the DOJ into the room until *after* all of the 'feltonizing' had been done on two identical laptop PCs. Then they claimed they couldn't reliably demonstrate the decreased performance on the 'feltonized' PC because of the unreliability of the connection speeds of the two different machines.
It seems to me, this is just another PR stunt by MS to prove that they know how to lie in impressive ways. If they're not going to run SQL7 on the same hardware, they've failed the scalability test from the outset, plus they've given themselves a scapegoat for when they fail to perform within the
I wonder if Rob's new-and-improved
~~~~~~~~
Signature illegible, could be somebody else.
Gates wouldn't stand a chance again Ellisons :)
Ninja Shuriken stars or hand claws.
Then again when Gates lines Compaq and Dell
executives up in from of him, Ellison might
take alittle longer.
Locutus
"Anyone who stands out in the middle of a road looks like roadkill to me." --Linus
But you cant compare MySQL to Oracle or Microsoft SQL Server.
Its apples and oranges. The caliber of programs are completely different.
I believe that Oracle's already closed down the $1 million dollar offer after about a 3-4 month time with no takers at all!
Still, I'd like to see the results. I don't think M$ has a chance, so this should be amusing...
I wonder if MS will allow someone other than themselves to setup and run this demo. As others have stated, they do not have a stellar track record for videotaped evidence.
The fact that MS is doing it this way suggests that they have already done the benchmark, and they have figured out a way they think they can turn this to their advantage. So don't be too surprised if they blow the benchmark away, (just don't ask them how they did it, I am sure this would just be a waste of time.)
In all my dealings with Microsoft its always someone else at Microsoft's fault :)
Thats funny, all of the sudden during the tape, the NT box disappeared and was replaced with a real machine. Hmmmmm.
Must have been an editing problem.
;-)
This sig is false.
"Three men entah, one man leaves!"
"Three men entah, one man leaves!"
"Three men entah, one man leaves!"
Where is MadMAx?
(c) Associated Press, Redwood Shores, CA - Oracle CEO Larry Ellison was startled today by two masked gunman as he attempted to respond in a demonstration to Microsoft's assertion that SQL Server would indeed meet his database perfomance challenge issued last year. Amid startled gasps in the crowd, one of the attackers held a gun clearly labeled 'SQL Server 7.0' to Mr. Ellison's head while the other took a bag containing the challenge prize of $1 million.
The two men were apprehended shortly thereafter, and were identified as William H. Gates, III and Steven Ballmer. Ballmer was heard at the police station as saying, 'seeya, Billy, I TOLD YOU SQL Server 7.0 could get that million dollars', while a dejected Gates could be heard muttering disgustedly 'just shut up, you bald dork'. A clearly aggravated Melinda Gates would only comment that 'this is absolutely the last time I'll ever bail out those two assholes... jeez, the DOJ mess was bad enough'.
Microsoft immediately announced that rumors of a new WB series, 'When Geeks Go Bad - Billy and Steve, the Thelma and Louise of High Tech', were completely unfounded.
Read the article. It says they used IBM's BD2
*
not Miscrosofts database software. This
challenge was about ORACLE vs Microsoft not
ORACLE vs IBM. Also, IBM used a 32 node cluster
of their netfinity and db2 servers to accomplish
this.
Quote: "32-node configuration used in this benchmark" IT used GigaNet's cLAN DB2 and IBM's serial storage Architecture.
Each Netfinity server has 4 Xeon processors each with 2MB onchip cache. In total that's 128 with 64 megs of on chip cache.
Oracle's machine used only 64 processors.
IBM DIDN'T WIN.
*******************************************
Superstition is a word the ignorant use to describe their ignorance. -Sifu
Yeah... and Oracle didn't use MATERIALIZED VIEWs?
(Thats pre-calculated results of querys... Oracle 8.1.5 has it MS SQL 7.0 doesn't)
'It's kind of fun to do the impossible.'
'It's kind of fun to do the impossible.'
- Walt Disney
...that finally met the challenge.
Wasn't one of the points that Larry Ellison trying to make that Microsoft was effectively prohibiting end-users from performing this type of benchmark that would ultimately show the MS product in a bad light. (Did the EULA claim that such a benchmark constitute some sort of reverse-engineering or something?) I seem to recall some discussion that whoever might claim the $1M might find themselves in legal trouble with MS.
CUR ALLOC 20195.....5804M
ERROR
The requested URL could not be retrieved
While trying to retrieve the URL: http://www.microsoft.com/presspass/style.css
The following error was encountered:
Access Denied.
Access control configuration prevents your request from being allowed at this time. Please contact your service provider if you feel this is
incorrect.
Hmmm... Wat are they trying to hide?
Has MicroSoft said how much slower they think they'll be?
-- Don't Tase me, bro!
Yeah!
Both of those products are propietary software. I couldn't care less about who wins.
GNU SQL will beat them all once its ready. Yes, GNU is creating its own SQL database.
Azul
I don't know who's better, but I know who will be better in some months.
A GNU/Linux box with GNU SQL searches and a few Perl scripts.
Azul.
Yeah, but it has right at the top of the site in big red letters that the database is 1.01 terabytes. My point was that they chose the relatively large grain data set (my math shows 300k/image) which minimizes the stress of the 1Tb test.
My hometown isn't covered and the spots that I looked at that were close seem to all get me to "cloudy_day.gif". ')
Left shift 1 for e-mail...
Yes, what a joke that site is. A terabyte of data - but its like 1000 really big pictures. Ooo it
really takes a powerful RDBMS to keep track of
1000 pieces of information. Access might even be
able to do it.
Left shift 1 for e-mail...
just as in the famous Win98 demo, I'm sure sure that M$ will get a wonderful blue screen of death when they try this challange...
after all, do you really thing that they have anywhere near all the bugs worked out of SQL Server 7, and then trying to tweek it on top of that???
perhaps using the new version will work against them.
---
"Everybody knows the moon's made of cheese," Wallace.
This was at http://www.microsoft.com/presspass/features/1999/0 3-16sql.htm
There's also a long piece on how they think Oracle tried to skew the contest. It's interesting, but I don't know nearly enough about high end servers to comment.
This month, as part of the premiere event of the "getting Results" Web cast series, Microsoft released its response time to the same query issued in the Challenge, announcing that it had achieved an execution time of 1.075 seconds on query five, significantly faster than Oracle's original mark and on par with Oracle's recent result of 0.7 seconds. Microsoft's results were achieved using Microsoft SQL Server 7.0 Enterprise Edition for a total cost of less than $600,000.
"Our solution not only matches Oracle's performance, but it does so at about one-sixteenth the price," says Leland. "It demonstrates that Microsoft offers powerful data warehousing and business intelligence solutions at a cost of ownership that is in line with real-world business realities. That's the core of our approach: to provide business solutions that drive down the cost of ownership and maximize return."
My Journal
As you're allowed to precompute so much of it during the load phase, just compare last yers Oracle with this years Oracle results....
:-) :-)
99-02-12
Sun UE10000 (64 400MHz US, 64 GByte ram)
Oracle8i 8.1.5.2
35,878.1 TPC-D (QphD)
Q5 took 0.7 seconds
Database Load Time: 46 hours 59 minutes
98-06-01
Sun UE10000 (64 336MHz US, 64 GByte ram)
Oracle8 8.0.4.2
5660.5 TPC-D (QphD)
Q5 took 1315.5 seconds
Database Load Time: 27 hours 2 minutes 2 seconds
Does anyone think that Suns tapes this year are slower than last years?....
(Both are 1 TByte db size)
I'll laugh whichever way it goes
Erik
Has it ever occurred to you that God might be a committee?
Has it ever occurred to you that God might be a committee?
--- Jubal Harshaw
Oracle is trash.
Great comment, backed up with lots of very understandable and informative facts. Armed with this, I think I'm going to my boss and suggest that we dump all our Oracle-based applications here at MCI and use... What?
Yes it's expensive, but it's very scalable, very reliable, and it's [GASP] portable! I can run my apps on something like 20 different hardware platforms without changing a single line of code. Try that with SQL Server.
SQL Server is trash. [GRIN]
Come on people, try to at least read the article before asking stupid questions. They are going to justify SQL Server as better based upon price. MS is not stupid, they write shitty operating systems but they aren't stupid.
At that time, Tong predicted, "Larry's multimillion-dollar computer will beat our $50,000 one by two-tenths of a second, because it's his test. It's likely that the cost of
the Oracle database software alone will be more than the whole thing for Microsoft, including SQL Server, consulting, software, and hardware."
hmmm, just a thought does this test say what the preconditions are? If SQL server is vaugely sensible it will act like oracle and cache query results....
can anyone guess what would happen if M$ where to run the test query 10 mins before anybody comes through the door - and then runs it again having done nothing in the meantime.
maybe I'm a cynic, but....
Tom
I wasn't suggesting that they cache the table - mearly the results... so while the query might need to access 1Tb of data if it only reuturns 64Mb of data the caching issue ain't so tough.
That said there has already been one response that lays out the terms of the challenge more clearly... which takes this sort of bahaviour into account.
There have also been some comments that MS is likely to only scrape through this challenge - I don't belive their marketing is that stupid.
I would expect them to show a very slim margin between themselves and Oracle - probably with them coming out on top. Quite how they will do this I do not know but they would not put themselves in this position if they did not think that they could walk out on top.
the challenge was not extended to intel platform data warehouses, rather SPECIFICALLY to SQL Server 7.0
IBM only showed that they can spank Oracle with DB2.
"All that glitters is not gold"
Who TF can even afford the Sun machine they tested on???
There is an Enterprise 10000 in the building where I work.
One, I think they're doing it for the money =)
Two, Oracle could turn around and point out that MS struggled and strained just to *finally* come within a factor of 100 of Oracle's wares...
Microsoft needs to do more than get within a factor of 100, if they don't get within a factor of 16, then Oracle still 'wins' because that is what Microsoft claims the cost-benefit of their solution vs. the Oracle one is. Even if they can do that, Oracle still wins because their solution is still going to be far more reliable.
What storage system are they using?
As an EMC employee, I hope they're using our stuff. (We make storage systems with up to 9 terabytes, and strive to have the best performance, the best reliability, and the highest prices, or something like that.)
I'm not in the part of the company that deals with such things, but I wouldn't be surprised if we have a partnership deal with Oracle, whereby their software is specially optimized for our storage systems.
> Who TF can even afford the Sun machine they tested on???
:-)
The people who regularly need to make queries to multi-terabyte databases can afford those machines.
-Snibor Eoj
One, I think they're doing it for the money =)
Two, Oracle could turn around and point out that MS struggled and strained just to *finally* come within a factor of 100 of Oracle's wares...
If I'm not mistaken, clock for clock, a G3 is faster than a P2, in just about every benchmark.
It doesn't mean Job's claim that an Apple is faster, but clock for clock, an Apple system is faster than a Wintel system, except for the OS bottleneck that is MacOS8x. Of course, it helps that Apple has standard on their high end machines SCSI and high performance video(not 3d video), so if you're willing to pay 1.5 to 2 times as much, you will get a machine that is 1.5 to 2 times as fast, but that was in the 3000 range last I checked. Of course, the Celeron changes all that, but clock for clock, a G3 is still faster than a Celeron. The only problem being Celerons have hit 450MHz, while the G3 is currently set to about 400 MHz or so.
If you do price performance, then the P2 beats the G3, but then again, Intel also has egg on its face, for the Celeron beats out the P2; in terms of absolute price performance, the AMD K6 beats out the Celeron even, barely...
If all you want to express is Anti-Apple sentiment, feel free, but don't knock PowerPC/G3. Some of the fastest computers in the world use that CPU; IBMs Blue computers, if I'm not mistaken, use a high end architected version, 64 bits and everything.
AS
-AS
*Pikachu*
Well, as of 4pm EST on Mar18, there're no signs of victory on Microsoft's website. Oracle, on the other hand, has a brand new challenge here. They claim $283QphD (whatever that means).
Anon, you should've read the original challenge.
Oracle specified specific query in a certain standardized database benchmark. If memory serves me right, the meaning of the query was: what share of the products was shipped domestically vs. internationally?
Hmmm... it makes me really nervous that the same query in the test took betweem 1.5 and 0.06 seconds (almost 30 times difference). As I understood it, "material views" is a glorified way of saying "we know there will be a query WHERE DAY>#1/1/1999#, so we'll stealthly create an extra column that equals .T. for 1/1/1999 and later."
So, how do you benchmark database engines? By randomizing the query to WHERE DAY IN (#1/1/1999#, "2/12/1990#,...)?
some background on the 1M challenge.
Oracle performed the test on a $16M machine (64 cpu machine if i remember correctly). No way on earth NT could be installed on such a machine (not in the near future at least.) The test was sorta partial because of way data was processed before queries were run on top of it.
How does mySQL measure up to all of this?
couldn't register 'cause the m$ SQL server was full. great PR move guys.
-Lkb
Microsoft OLE DB Provider for ODBC Drivers error '80040e14'
[Microsoft][ODBC SQL Server Driver][SQL Server]Can't allocate space for object 'microsoft_registration' in database 'CreateTech' because the 'default' segment is
full. If you ran out of space in Syslogs, dump the transaction log. Otherwise, use ALTER DATABASE or sp_extendsegment to increase the size of the segment.
/microsoft/gettingresults/event_setup_p.asp, line 67
Read the original post. He said If you throw a G3 chip to the ground, and merely drop the PII (<STRONG></STRONG> emphasis mine).
Newton's law does dictate that the acceleration of both will be the same 9.8 m/s^2. BUT, if the initial speed of the G3 is greater than the initial speed of the PII (0 m/s), and the distance is the same, then the G3 should have a higher speed all the way through, and hit the ground faster (air resistance notwithstanding).
Of course, it doesn't really matter all that much...
--
- Sean
It's a fine line between trolling and karma-whoring... and I think I just crossed it.
- Sean
If this Microsoft SQL Server is anything like the TerraServer it had on the web for satellite imaging maps.. Oracle gets to keep its million.
yacko
I worked at a shop where the Novell/Foxpro solution was becoming unreliable and slow with around 200 users. Time to upgrade to a real database. So we shopped out, and Management chose Oracle.
The Oracle database software is great, but their consulting makes MS look even better. In fact, that's what happened - a MCSE consultant (who was setting up our T1s and Exchange connection to our parent company) got hired as IT department head, after promising to move to NT/MS SQL and rewrite the app in a few months with three VB programmers.
I left shortly afterwards to start my own company.
One year later and the VB programmers had not completely replicated the Foxpro system, but it worked, and the sales force was using it, buggy and crash prone as it was.
Oracle consulting needs to get their act together, or MS *will* catch up on their technology eventually, and MS will be the better solution. MS SQL is just Sybase, anyway... MS can throw enough money at the problem to eventually be roughly equivelent, and marketing will hide the remainder of the gap.
( In regards to the test - is the team working on this the same team that produced the videotape for the DOJ? )
Oh, and anybody acually register and SEE the results of the competition?
"$30 for the One True Ring. $10 each additional ring!" -- JRR "Bob" Tolkien
The guys who need the bandwidth and uptime - banks, insurance companies etc. These users need multi-TB databases and all that power. :) ).
your rationale is flawed - if a 50K$ server gets about 1/100 performance of a 5000K$, it still does not mean that their DB will scale to 5M$-worth of hardware (and MSSQL server _certainly_ does not scale linearly, not to say NT
This means that if they get to within 1/100 (or even 1/50 or whatever) on a 50K$ machine, they've still lost.
>All the consults finbally decided that unix was dead and NT is the feature and I can't wait untill that day comes.
You're probably the only one then.
I can hear the headlines now.
"Following Microsofts recent cracking of the Oracle challenge, Microsoft shares rose 15 points and sales of SQL server have tripled. On a side note: 10 of the worlds largest web sites are down today quoting 'database problems' as the reason."
Oracle blows the doors off of SQL server, 100x is probably an understatment, but it's so damn expensive that people will go with the slower system (which they probably got for "free" in their MSDN subscription) Oracle and IBM are worried about MS in this market and there will be more outlandish marketing ploys like this.
This is my signature. There are many signatures like it but this one is mine..
Ok, so it's amusing for media coverage.
The real question is what is the uptime of your database?
SQLServer only runs on NT. Ooops.
I can rely on my OS staying up. That's why big wesites run against Oracle not on NT.
Do you even know anything about perl? -- AC Replying to Tom Christiansen post.
Perhaps microsoft has a trick up its sleeve ... (dont leave doctored video/results out of the equasion) ... Maybe their going to post much higher performance than one might expect?
--- ask me about nihilism, I will have nothing to tell you.
Does Oracle have to honor the offer since it is being done on a product that was not available at the time of the challenge?
As far as I understand Oracle, a shop of this size would not justify internal Oracle consulting, instead they would far it out to their registered development community, which is a bit hit & miss.
This model is at odds with MS, who very much support their MCSE's.
ps - MS haven't actually answered the challenge; they've just decried it as irellevant (yeah right, MS would NEVER make an irrelevant challenge...), and come up with some figures that they claim reflects the 'real world'.
j.
In the world of the data center, that kind of hardware is not ridiculous - just rare!.
;-)
The whole thing just highlights yet another shortcoming of SQL6.5/7 - namely, that it has to run on NT. hahaha.
Anyone like to try & benchmark mySQL on that kind of hardware ?
j.
You are incredibly naive if you don't think Oracle did the exact same optimization you're accusing MS . Would any sane CEO set up a challenge like this if he hadn't already tweaked everything on his software to make it as hard to beat as possible?
A much better line of criticism is the fact that it runs on (unstable) NT server.
It looks to me like this will just call attention to the fact that SQL Server is far slower, whatever the factor. Nobody's going to think about the hardware involved. They should wait to take the challenge until they can somehow be _faster_ than Oracle. Heh. If that ever happens.
Dito mate : (
Doug.
I think that *any* MS presentation has the same credibility as one of their videotaped ones... I'm sure that even if what we are shown isn't *technically* the truth it will certainly reflect how MS thinks the world should be- which is, of course, the same way everyone else views it :)
-Chocodile "Thud on top, I ate the chocodile." -from "Disseminated" by Soul Coughing
This could humble Microsoft in the eyes of many.
I agree...
:)
.. and SQL is a Sesna.. but, from what I have seen.. a lot of companies don't need more than a Sesna..
:)
I am a programmer/forced to pretend to be a DBA.. I have evaluted Oracle, and SQL Server was just plain easier to use.
It might not give me 1billion transactions a second, BUT TSQL is a really easy, and quiet dynamic language.. PL/SQL seems to be a bit older and not as dynamic... (I lied, its as dynamic.. just more dificult..)
I felt like PL/SQL was more like cobol.. and TSQL was more like Visual Basic..
I am not a MS Advocate, so please don't flame me. I just need to find a solution that was more on a work-group level, and I didn't need to spend 200k $ on a system, we wanted something that our in-house MIS person could handle, once it left R&D..
Honestly, we have our development server up and running, and maybe we reboot it once a month.. no kidding. (Unlike my NT Dev workstation that gets rebooted like 30-times a day.. )
I wouldn't be happier if they compiled a version of it to run on Linux, or some *NIX, but I doubt that is going to happen.
I guess the point I am trying to put acrost, if you don't have 3 years experience running Oracle, good luck just figuring it out.
I was up and hacking on SQL Server in hours, as with Oracle.. well, if you have used Oracle w/out instruction.. you know.
I am not saying that Oracle is not a better product, because after I evaluted the two of them... really, the Oracle system is a F-16
SQL has some good replication features, and it's a dream to program on. I never have had a problem with corrupt data, and its really low cost as far as maintenance..
For 6 months, we had it running on a AMD - k6 166, with only 64 megs of ram..
Really, in peak time we would have 5-8 users max on the thing, doing nothing more than Statistical reports..
It's just a matter of picking the right tool for the job, and sometimes you don't need a nail-gun.. a hammer works just fine.
SQL Server 6.5 was a piece of crap though, 7.0 is a 100x imporovment.
I wisht that oracle would do a nice intuitive port of their enterprise managment tools to the KDE environment. (I couldn't believe how nice KDE is.. I always thought that Linux was just a good Server System... KDE changes all that!!!)
hehehe... $.02
Microsoft mentioned that they din't use TPC-D.:
----------------------------------
Rather than try to match a result in a benchmark test that was already in dispute,
Microsoft and Hewlett-Packard chose to craft a solution that would more closely
match actual business conditions, including the need to get maximum value from
any enterprise solution. They also wanted to develop a solution that met the original
intention of TPC-D.
"There was no merit in responding to the aspect of the challenge that was nothing
more than a marketing stunt," explains Microsoft's Douglas Leland. "Instead, what
we found interesting was the opportunity that the challenge raised for us to use our
technology in a way that focuses on what customers really need, which is
affordable, scalable, and powerful solutions that are flexible enough to answer a
wide range of business needs and issues."
----------------------------------
So who knows what they used? They don't define their database. Here's a silly analogy:
It's as if Bill was challenged to eat four crackers and whistle all within a minute. Well he went off, ate and whistled in a minute, but he decided that eating crackers didn't make sense so he substituted something else. "Rather than trying to eat crackers I chose to eat something else. I'm not telling you what I ate, but I ate something."
Well good for you Billy Boy, good for you. But that ain't going to get you the prize.
Esteem isn't a zero sum game
I say we call up Don King and see if he can setup a bout between Bill and Larry. That will settle everyones dispute on whos boss. Of course you can expect the fight to be fixed.
I've been a DBA for a good few years, done a lot of performance work on many different platforms. From my day-to-day work getting reports/transactions to run in a timely fashion I can tell you what this means to me. Squat.
The TPC benchmarks have to be analyzed properly to understand what they really are telling you. Not just that, the real point here is that we have two multi-million companies throwing their weight around and posturing. It doesn't help people trying to admin databases at all. It benefits some egos of some rich guys and ups the profit margin for a while.
Under OSS this kind of challenge would become 'Hey, I just found this neat way to make this box go faster. Why don't you have a look at the code and see if you can use or improve it?'
This whole shooting match just further exposes the true weakness of proprietary software. If knowledge is held under guard, copyrighted and isolated its benefits wither and die. It serves to make money for a few for a short while and nothing more. The progress that has been made in database technology over the past 10 years that I have worked in the industry has been produced by chance from short-sighted industry players following fashions like butterflys in the wind. (what a crap analogy)
ZamZ
Well, a lot of argument about nothing, there are situations when an Oracle/Sun solution would be appropriate and situations when an MS SQL/NT solution would be. it just depends on the situation, their is no definitive answer.
Thers is no doubt that Oracle is the superior product, but thats not the whole story. Yes if you are running a large DB with say a couple of thousand users then you wouldn't look twice at MS SQL on NT. But lets say you've got twenty users with a support resource limited both in size and technical ability, would Oracle on Solaris be your first choice?
> can anyone guess what would happen if M$ where to run the test query 10 mins before anybody comes through the door - and then runs it again having done nothing in the meantime.
If M$ manages to cache 1 Tb of data using a
$600.000 solution, I'd be very impressed.
-- Abigail
> Who TF can even afford the Sun machine they tested on??
Banks. I know banks that have bought 5 E10k's for over $50M, just to run Sybase and one suit of applications. (That of course includes the standby's)
-- Abigail
Whats up cary?
B2
Of course, Microsoft has agrred to take the challenge only _AFTER_ the challenge has already been closed. http://www.uk.oracle.com/info/news/feb99tpcdbench. html Typically Microsoft. David