Slashdot Mirror
Microsoft Challenges Linux community
AmirS writes
"Microsoft are really pushing for the Mindcraft
benchmark to be re-run, so much that they've put up
a page about it.
They say they've met all the requests of the community (seems
like most have been met) and just require linux people to
step forward for it. "
Like Zeus that is reportedly 3-20 times faster than Apache! Since we're pitting Linux vs NT, we should be able to use any webserver we want.
That page hurts. We are going to have to do something about this. There may be some FUD there, but there is a lot of truth too.
ac.
We've got Microsoft's list of selling points now. Address them one by one, and we've got a roadmap on what to improve.
A few of them caught my eye as being near lies where they compare cost of ownership, etc against UNIX, not Linux. Tricky those Microsoftians.
The most increidble selling point that they missed is the bulletproof nature of Linux compared to NT, and the work:crash ratio. NT is an immature product compared to the years of well-tested software that Linux has had to build upon.
Wish we had numbers like the # of man-hours and such or bugs fixed to show.
Have a look at some of the claims on there
"NO back compatiblity for a.out binarys"
"No international support" when KDE supports more countries than windoze.
"Poor support for java"
"MORE PRONE TO SECURITY BREACHES".. remember teardrop?
...that Linux really could lose.
/. does not count as "doing" :-)
Calling stuff "FUD" won't make the challenge go away, nor will it make the _real_ technical problems of Linux go away.
Linux still has a _long_ way to go with SMP scalability. Linux still doesn't have a journaling FS. Linux still doesn't have an LVM. Linux still doesn't have USB support. Linux still doesn't have a good, comprehensive administration GUI. [1] And unless you yourself are trying to help make these capabilities come to the point of production releases, you aren't helping at all.
The best way to approach this challenge is to say "No, Linux doesn't quite measure up yet, but I'm personally doing everything I can to make it so it does." (Posting to
-- Citizen for the responsible promotion of Linux
[1] All these capabilities exist in alpha form, but calling them production releases would be a LARGE stretch.
How about a third option.
Why can't Slashdot, LinuxToday, or one of the distro companies prepare a set of test criteria that Linux performs better on than NT and challenge _Microsoft_? Let's think outside the box a little and quit letting Microsoft set all the rules of the game.
As a registered member of the community, I thought I'd take a few shots at the MS "challenge."
First, in the sloppy writing department:
Well, uh, no: the consesnus is that when Mindcraft configures and tunes the servers (or not, as the case may be...), things go badly for Linux. Feelings on the second point seem to be the same.
What MS mean to write, of course, is that Linux people want to configure, tune, and be present. I'm sure that the sloppy writing isn't intended to muddle the issue, since it's sort of clarified a paragraph or two down.
Looking at their comparison chart, I note that they claim Windows has turned in the "best" scores on some benchmarks, while also noting that no Linux results exist. Winner by default, I guess?
On Linux, it's "easy to gain root access...". But, they say, on NT:
Does that mean that this exploit no longer works?
Here's a nearly incomprehensible complaint about Linux:
Melissa shows what costs and technical risks come with "integrating" stuff to the extent that MS wants to, but I'm not entirely sure what they mean by the word in this context.
In the damning faint praise department, MS graciously admits that there are "hundreds" of applications available for Linux. Call me crazy, but Unix is, er, "several years" old -- I'm pretty sure there are more than hundreds of useful programs available (whether they're "applications" or not is not terribly relevant, if you ask me). Even if there are only hundreds, well, a comparison of quality, rather than quantity, would be more telling, I think.
Another Linux failing, they say:
I'm sure I don't know what that means. Organizations like The Learning Tree have Linux courses, and there have been a couple of certification programs announced (if inchoate).
More Linux evil:
Or, you could just give the job to some random person and let him/her peruse the manuals. Things wouldn't turn out any worse than they would if the person were told to run NT instead. The fact is that a Gooey WimpyWYG PointyClick screen doesn't change the fact that administering a computer well (let alone a network) requires skill, intelligence, dedication, and plenty of learning. No "Wizard" will get around this fact.
Uh, well, maybe. But you do have to "install service packs" on NT, which comes to the same thing in the end -- downtime while the admin does something that, if it doesn't work, will result in Bad Things happening until it gets straightened out.
Oddly enough, they forget the corresponding item on the NT side: "Most config. settings require editing of binary files." Or, rather, one (the Registry), and if you screw it over, God help you. At least the OS keeps a couple of backups by default.
Here's one of their Big Awesome NT Features:
Unix is Home of the Script. That's all I have to say about that.
NT feature:
Melissa. Not all rosy.
Linux liability:
Nope, I'd say MS is the master of forcing people to integrate. (Yes, that was an out-of-context quotation followed by a cheap shot).
NT feature:
And you know who's paying for that -- look at the prices of their OS and applications (particularly the proposed prices for the various Office 2000 flavours).
And then they sum up. It's crapola in the best tradition of election campaigns, such as the one I'm currently enduring here in Ontario. Some highlights:
Well, when you notice that the messenger is full of shit, you don't tend to pay much heed to what's being said, now do you? The test was flawed (arguably fatally), so there's little point considering the results.
No, Beavis, it's not. Even if no Linux person steps forward with brilliant test results in response to this "challenge," the fact remains that the original tests (and thus the original report) deserve the criticism they've received. This statement is about as valid as an assertion that since we have trouble treating cancer, we musn't go around saying how bad it is.
Mind the Gap
Nevermind the bogus claims this page is making about NT's superiority over Linux (Linux inherited UNIX' weak security because of buffer exploits? Like NT doesn't suffer from the same weakness?)--the fact remains that Mindcraft/Microsoft tried to pull a fast one and they were caught with their pants around their collective ankles. This challenge is nothing but kicking dust in the air until Microsoft and Mindcraft admit that they set the tests up to show Linux in a poor light. I am not willing to give Microsoft a pass on this one, and participating in their benchmark without forcing them to acknowledge the real reason for our dissatisfaction, instead of dismissing it as "attacking the messenger," would do a disservice to the Linux community. This would be like the archetypal battered wife returning to her abusive husband because he promises to clean up his act.
It should be noted that one of the side effects of the Mindcraft benchmarks was that some very effective optimizations for Apache were identified and, last I read, were going to be submitted to the Apache group. I imagine the benchmark rules will prohibit their use in the proposed benchmark.
The Linux community should respond in kind with a challenge to Microsoft admit that the first benchmark was a sham, that Mindcraft lied about the extent of their efforts to find help tuning Linux and Apache, and that Linux and Apache were de-tuned on the test machines. Last, but not least, they should fire Mindcraft.
slashdot broke my sig
Wow - that was harsh... I guess the Microsoft battle is now firmly ON - big time. That's probably good news - it means they're a little bit scared.
,hacker Perl another Just)'
So let's try and address these points.
Spec Web
The Spec Web figures are generally put out by hardware and software manufacturers in cahoots to produce high figures (often using slightly modified server binaries). Linux has no hardware vendors who are yet big enough to produce these figures. VA are getting there but I don't know if they have plans to produce SpecWeb figures.
The same goes for SAP and TPC figures.
Netbench
The kernel developers are aware of some issues here, since Solaris and Irix don't exhibit this behaviour. Also the clients have so far tended to be '95/98 clients. Where I work all the desktops are NT Wks (thousands of them).
WebBench
These figures use Apache, instead of Zeus. Let's see some real figures with Zeus before making judgement here. Apache is meant for complex systems developers who need flexibility, not raw speed.
Also, WebBench's dynamic benchmarks cover ISAPI on IIS and CGI on Apache. Gee, that seems like a fair test eh? It's not, and until WebBench provides an apxs module it will continue to be unfair to Apache. Why not compile that ISAPI module on Zeus and see how it fairs? I think we know the answer.
Reliability
Microsoft are most scared of Linux's reliability (hence it's at the top of their non-performance list). OK, so no OEM guarantees Linux uptimes. Big deal. NT's 99.9% uptime guarantees are based on clustering solutions - not single servers. And these guarantees are expensive. NTFS is not a true journaling filesystem either, although they may be talking about a commercial filesystem that I'm not aware of.
Scalability
I think they're probably pretty close on this - although they still make some glaring mistakes - like Synchronous I/O - only on the driver mindcraft used. And if pthreads aren't kernel level threading I'll eat my shoes. Yummy. Oooo and NT has an integrated file cache... Linux has one of those too. Wow.
Security
I don't even have to touch this section. Wow... stunning marketing going on in there.
TCO
Comparing to UNIX, not Linux.
Ease of Use/Admin
I just loved the bit about "Scriptable administration tools for automated local and remote administration". OK, so wsh is now available. How much is it used? What about non-automated remote administration? For most administrators of NT boxes, wsh and SMS just don't cut it - they have to walk to the box usually.
Actually, I'm bored with breaking this down. What can I achieve - leave me to just use Linux and be successful with it, and not have to suffer reboots.
Matt.
perl -e 'print scalar reverse q(\)-:
Matt. Want XML + Apache + Stylesheets? Get AxKit.
The hardware chosen certainly won't give Linux an advantage, whether it be the four processor box, or the low end system (although it looks like Linux would have a better chance on the low end box). Apache sure isn't gonna kick butt as a high performace web server (which isn't what it was designed for anyway). Of course, I wouldn't expect Microsoft to pick hardware that would give an advantage to anything but NT.
Unfortunately, the "Linux Community" has made a whole lot of noise about the original Mindcraft benchmark and its problems. I fear that in doing so we have played right into Microsoft's hands.
This benchmark business has already made so much noise that I've had people who have a hard time figuring out the Start Menu in 9x/NT asking about Linux, NT, etc. Microsoft is certainly doing all they can do to make sure the world knows that the "Linux Community" has disparaged the original benchmark. Now, they're raisin' a ruckus because we haven't (seemingly) done much to respond to their challenge.
I think (and I'm probably not 100% right, I may even be wrong) that the "Linux Community" only has two options now:
1. Accept Microsoft's challenge (however we're supposed to do that) and have as many experts on tuning, etc., on hand as possible to make sure Linux gets as much of a fair shake is possible given the hardware.
2. Do nothing and let Microsoft trumpet to the press that the "Linux Community" didn't have the balls to accept their challenge because Linux really was inferior to NT (which we know good and well it isn't).
Seems to me that our best bet is to accept the challenge and take our beating like a man, if that's what it comes down to. Of course, if Linux loses, MS will plaster the numbers on every bulletin board and window on this side of the Milky Way, but at least it will be a well tuned Linux running against a well tuned NT, which is much more than can be said about the previous benchmark. Even if Linux does lose, it's not the end of the world for us, even though MS will work to make sure it is. They can't kill us, and what doesn't kill us will only make us stronger. If this turns out to be a failure for Linux, then we can learn from what went wrong and work to improve performance in whatever areas it is necessary to improve performance in.
And yes, I intend to help, as soon as I'm competent enought to do so. In the mean time, I'm learning.
Now for a question: I was under the impression that NT pretty much choked on more than two processors. Am I mistaken? I know we've got an NT box with two PII (233s, I think) and 128 megs of RAM and it's slower than molasses).
Yes, it is a bizarre strategy to say things like "Linux has yet to post SPECWeb results", when "Linux" is really nothing more than a mailing list and some FTP mirrors.
I guess you can file this with the No Roadmap FUD - non-sensical to anyone who understands what Linux actually is.
It would make more sense to post "RedHat has no roadmap" or "RedHat has yet to post SPECWeb results", because RedHat is actually an operating system vendor, who at least in theory competes with the big boys, and therefore is going to have to (at some time) start doing the same kinds of marketing.
--
Business. Numbers. Money. People. Computer World.
MS says:
Overall, 37% less expensive to set up and operate than UNIX.
26% less expensive to set up and integrate than UNIX
27% less expensive to administer than UNIX
Note that this may all be true, for regular commercial Unix, but the difference for Linux is certainly not that large.
I'd like to see a real (objective) comparison between the operating costs of NT versus Linux in various roles. Note that I wouldn't be suprised to see NT come out on top (even with the licences), because the most expensive part of any server is still the system admin.
(If you've got a good Linux admin working at your site for the same pay as a good NT admin, count yourself very lucky!)
--
Business. Numbers. Money. People. Computer World.
Reading some of the posts here and the Microsoft Article leads me to question what the goals of the community are.
.. "hey.. do I need 400Mbps performance? Do I need a Quad Xeon?" If their answer is NO.. then perhaps with some positive press that Linux has received they should be asking themselves the question, "Why should I pay a couple of thousand to MS for NT when I can get Linux and support for a couple of hundred dollars?"
.. time to go to bed.. excuse the rambling :)
Linux will not be able to dominate the world today in its current state or perhaps in any state. Let us accept that, live with it and perhaps smell a few roses while we're at it.
What are Linux's strengths. Let us use those to get at Microsoft. (This is from my perspective as a poweruser and a suit.)
* Linux is a kick butt entry level and departmental server right out of the box. It provides VERY decent connectivity to Windows '95 and NT boxes out of the box with Samba. There are FAR more business using low end NT and Novell servers than there are companies using Quad Xeons. This is also where Microsoft makes its real money.. selling low end NT servers to departmental and workgroup users. Heck, my office in Hong Kong still has a 166Mhz Pentium as its main file server for 20 users running NT. We do not see any need for a P2, let alone an SMP box.
* Linux is a VERY good platform for providing Internet services to the same businesses. Again, with a little bit of tweaking, the same Linux box can turn into a mail, web and news server. Try getting the same server described above to run exchange and IIS. Not possible. (Ironically our Exchange 5.5 server runs on a P2-450 with 256MB RAM).
Performance is not everything. The type of box MS has constructed is used perhaps only in the largest of the large enterprises. Would any of the target audience seriously consider running NT on a $100K box? No, chances are they would opt for a SUN or an RS/6000/HP type solution, which ran their application/solution well.
Let us look at Microsoft's Annual Reports and see where they make money. Is it by selling NT on a Quad Xeon? Or is it by selling Workgroup / Departmental Servers with '95 clients and Office bundles?
My suggestion.. fight the FUD, but do not get consumed by it. The market is a LOT more than what MS has made it out to be. And look on the bright side of things... Linux is not featured on MS's pages. Departmental Managers, Small VARs looking at that page would go
Ok
This FUD is designed to focus the community into a producing small group to answer all their challenges.
DON'T fall for it. DON'T answer their challenge too directly. DON'T provide a focus for their attacks. If MS has a single target for their attack, then they can set the rules for the fight.
Keep diffuse. Attack from everywhere at once.
Let them swipe at a swarm of bees with a sword, until they get too tired to fight.
If big businesses are interested in Linux, they should come on our terms. Don't bow to pressure to soften the open source stance. We'll be here regardless of what they do, if they want to join up and help, well and good. If they want to fight, eventually they'll lose.
Choice of masters is not freedom.
Microsoft talks about the Linux community as if it's something monolithic like a competing company. I understand that that's probably the only way they know how to respond to something, and even if they did understand us it's still in their interest to denigrate linux I guess.
It's mostly interesting to see how threatened they apparently are by linux -- the OS that just won't die!
-- It only takes 20 minutes for a liberal to become a conservative thanks to our new outpatient surgical procedure!
So NT beats linux on a particual hardware config (4 CPUS, 4 ethernet cards) in a particular test (static web and SMB file serving). I am sure MS went to great lengths cramming support for this down in the NT kernel so they could get those results... but at what expense? Just like a dragster, they now have a machine that goes fast in a particular setting... but sucks even more for general use.
Listen. We could no doubt hack the Linux kernel to do the same tricks as NT, do them better, and trounce all comers in these benchmarks... but why!? People don't buy servers to win benchnarks, they buy them to do real work. We need to cut through the FUD and remind everyone of that. We should continue to work on improving linux performance (SMP does need work), but not at the expense of flexibility or stability. We need to counter these slanted benchmarks with our own tests that more closely appoximate the real world demands placed on an enterprise server. This is where Linux mops the floor with NT.
Thad
The Bolachek Journals
Hi Everyone:
I skimmed the Mindcraft report and the first thing I saw was a highly optimized hardware package specifically selected to favour Microsoft.
I have run the identical Microsoft software package on a client's P200 with 64MBytes of memory and an 6.4 Gigabyte IDE HD and it was a pig. The software was replaced by Slackware 3.6 with a 2.0.35 kernel and Apache and it runs
like a champ. That box NEVER sees the kind of throughput that is contemplated by the Mindcraft/Microsoft test but most applications don't.
If you can consider that most small commercial webhosting sites won't be seeing 1000 hits an hour, let alone ~3800 request per second.
I have run into overworked and underpowered NT4 sites and it is pathetic because they are so ungraceful under pressure.
Sure, it is easy with an unlimited hardware budget to get NT4 to run but that is not the real world!
The real world is $5,000-$10,000
hardware/software/consultant budgets running on 10Mb/s Ethernet connections. Linux and Apache are about the real world, not the fantasy world concocted by Mindcraft/Microsoft and their team of engineers working behind closed doors.
I suggest to you that with hardware/software budget Mindcraft/Microsoft used and a free hand given to Linux/Apache techs would result in a faster, more reliable, more graceful solution (four distinct servers perhaps).
Here is my challenge Microsoft/Mindcraft:
Who can build the best HTTP server (hardware & software) with a real budget of $5,000US hardware/software/consultant fees total!
Just one guy's opinion.
SubDude
Did any one notice on the last row of the table they state that:
PC Magazine (e[commerce - SSL) - 250 Requests/sec (680% faster than Linux)
But on the linux side:
PC Magazine (e-commerce - SSL) - 1950 Requests/sec
Wait a sec here?????????????