Slashdot Mirror
Red Hat Enters The Database Market
tekBuddha writes: "It looks like Red Hat is trying to make its way into the database market. This article on Yahoo! says Red Hat is about to announce its own product ingeniously named 'Red Hat Database' next Monday. "
Is this the most advanced open-source database available now?
Does sap-db give you the ability to roll-forward a transaction log in a recovery situation? Can you do something like this with Postgres right now?
You mean MySQL. Microsoft SQL Server wasn't mentioned, and it is an ACID database.
In fact IBM or DB/2 weren't mentioned either. What article did you read?
Well, yes and no. First of all the different PostgreSQL backends are already quite adept at reading the binary formats, and the changes (from what I understand) have generally not been that large. Binary formats certainly are more fragile than human editable text, but clearly it is possible to manage binary formats fairly safely, otherwise any database would be a poor place to put your information. Second of all, since the postmaster would be off the tool could safely do a
tar zcvf backup.tgz data/
Before running the format munger. The tool would have to guarantee you had enough space before hand, but that isn't a big deal (and if you failed you could simply exit with an error message).
Besides, buman readable is only useful to a point. For example, the backups on my PostgreSQL database are big enough that I run them through split so I don't end up with a single file greater than 2G. If something happens to one of my backups recovering that data is going to be a pain (though it would probably be possible).
When upgrading I generally shut down my PostgreSQL server for an entire day as it takes a long time to upgrade (during which time the database is unavailable). When I upgrade I generally dump out the database, load the database into a test machine running the new version (while making a tarball of my data directory just in case), I then verify that all is well, upgrade my production machine's version of PostgreSQL, and then load the database into my production machine. This takes a long time. Updating the binary files would almost certainly be faster, and with proper precautions it wouldn't be any more dangerous.
Not that I am complaining. PostgreSQL's structured text file dumps (like most everything in PostgreSQL) work quite well. But it's not very fast or convenient.
No, it is inherently suspect. If it includes enough information to reproduce the results, and they are reproducible odds that it is useful go up. If they lose odds go up more. If it is simple enough that you can see if it merely played to the strengths of that software or if it is a real test of how it might be used...well that helps too.
After all SPEC benchmarks are run by the venders (on programs and data sets that a group of venders all agreed on).
Yesterday's Tux benchmark wasn't worthless (at least not if you serve static content).
Most benchmarks are not done by impartial entities. Many are run by magazines that get advertising money from only some of the products. Some are run by the potential users that have a vested interest in picking the cheapest one.
The same thing happened with the product called "SuSE Email Server." It turned out to just be Postfix, Cyrus, and some other things pre-integrated and shipped in a nice-looking box bundled with a support agreement. Considering the magnitude of writing a database, I'd expect Red Hat's offering to be something similar. They didn't survive the Linux company shakeout and turn a profit by being stupid.
--
Tired of FB/Google censorship? Visit UNCENSORED!
I don't understand this at all.
OK, so, you can't use Oracle, because its proprietary. Fine, use MySQL. No, wait, that's largely banked by VA (IIRC). Also rule out Postgres, and Interbase for similar reasons.
So *develop your own*? That just makes no sense. "NIH" (Not Invented Here) kills companys, it blinds their thinking from the best tool for the job.
Why can't Red Hat just partner with NuSphere/Greatbridge/etc? Is it NIH, or something about business I just don't understand?
Last point: This has (apparently) appeared out of nowhere. Some press, but that's it. No beta program. No white papers. No conspicuous hiring of RDBMS gurus - hell, check the Borders in Durham, I doubt they've even bought a textbook on databases. What gives here? Are they going to launch a product without first having running code?
ZOMG I WOULD LOVE TO KNOW ABOUT YOUR FEELINGS ON MACINTOSH VERSUS WINDOWS, VI VERSUS EMACS, AND HOW YOU'RE NOT A DORK
While the text format of XML is what most people think of when they talk about XML, this isn't what people mean when they talk about XML databases.
XML is also a way of structuring data; trees, parents, children, attributes, etc. When people talk about XML databases, they mean a way of storing the XML data, without storing it in the XML text format.
Presumably, this sort of approach would be much faster than using the text format (and far fewer files!) but would also be much easier when to use with XML data than relational databases are. It would also allow the database to prevent races.
That would be a fantastic move... as far as I know there are no other players in the databse-market, so Red Hat should be able reak in heaps of profit!
I would also look into the databae-market, which is also unexploited.
:-)
The article states that Red Hat is specifically going after small to medium sized business, who don't see Oracle as a reasonable expense.
Oracle is expensive and charged on a per CPU (actually per Mhz) basis. So if you're building something that is expected to scale, you're looking at a lot of money.
For example, if you're building a cluster of web servers on stock x86 boxes, you'll find that the price of a single Oracle license on a single box will probably cost you in excess of five of those plain white boxes. For a small company, that is unacceptable.
I can't believe they listed MSSQL next to DB2 and Oracle! Sheesh!
Yes, MSSQL runs on expensive hardware. Oracle and DB2 run on *really* expensive hardware. And pass the ACID test. And so on.
Guh...
-grendel drago
Laws do not persuade just because they threaten. --Seneca
Anybody think that they'll actually try to make it protable across all Unices?
--
SecretAsianMan (54.5% Slashdot pure)
Washington, DC: It's like Hollywood for ugly people.
Also note that Greatbridge does NOT SELL PostgreSQL - it sells services. I would bet money that it is PostgreSQL that RedHat produces, and contracts out tech support to Greatbridge.
Synergies exist, there is no direct competition, and the owner is an old friend.
I think the writing is on the wall.
-Mark
I've used Interbase since 1996 and like it. If you follow the uninstall procedures (i.e run the uninstall program) as specified by Borland, then it uninstalls just fine and will not corrupt other database drivers. It's when people try to manually delete files that that they get into trouble. If you were using the BDE, manually uninstalling drivers is a bad thing in general and is not recommended.
Additionally, InterBase is not orphaned and is very much supported by Borland. I suggest you go to their website and see what they are offering. Last I checked, IB was running on a myriad of platforms and is included on the Kylix CD. I assume its also on the Delphi 6 CD...but as I don't have it...it's just an assumption.
If you are looking for a free alternative, I suggest you consider the FireBird project. It's essentially InterBase but based on the Open Sourced code. It runs under Linux and Windows.
Best tool to use for accessing InterBase or FireBird is Jason Wharton's InterBase Objects.
Finally, if you are running Windows, use a utility like CleanSweep to record your installation process. This will make a clean uninstall a piece of cake. One thing about Windows I really don't like is that there are no reference counts to files, libraries, or drivers. But, then again, I don't know of too many install programs or OS's that actually track this information. Oh well.
How does it work if a decision in a big company for a _new_ DB system (usually for a special purpose project) has to be made?
I'll tell you how I've seen it done in telecom. You sit around in one meeting after another, gathering a fantasy list of user requirements, only to discover that the sponsor blew their budget and expected the IT department to cover the costs.
Then, when you get to the sticker shock, you get questions like:
"What's the $250K for? Licenses? For what? Database software? Can't you just write something? After all, what's all that development time for if you're just going to go out and buy stuff from Oracle? Why are you paying all those programmers and database people if you're just going to buy it anyways?"
By that time, open source is looking to be the only option that'll come close to the user's budget. True, you've got the developers, and probably the time (wait in line), but without the funds for the big budget commercial database, open source sure beats writing from scratch.
Plus, I've yet to meet a commercial database that came pre-constructed to do exactly what I need to do for telecom billing, so it's assumed that you'll be conducting a good amount of development on the system regardless of its origin.
But please folks, stop dreaming that RH will eat away the piece of the big guys in the upperscale solution market.
No. Microsoft already has done that. In too many vertical markets, Oracle/IBM are not adding any value. At least Microsoft doesn't charge you nearly as badly based on use. (Vendors: Take a hint. Charge me on use and you'll guarantee I'll be looking for a way to get rid of you as I grow!)
Per the previous example, I've tired of Oracle quotes that demand:
- extra per processor
- extra if you're going to use the data on the Internet somehow (even if it is interfacing a Microsoft SQL server first, which is talking to the web via IIS and Oracle never sees a public IP)
- a huge chunk of change for maintenance (aka bug fixes)
Then again, it's not easy to make Larry Ellison wealthier than Gates...
You're also right that PostgresQL and other open source varients can't touch Oracle's very high end capabilities. Their stuff is "trivial" (thanks Clayton Christensen for the model), and is the only thing the very, very high end customers (5%) of Oracle can use. The other 95% do just fine with the open source alternatives.
*scoove*
oracle already runs on linux. i've run 8.1.5, 8.1.6 and 8.1.7 on redhat 6.2 without any problems. i have a development 8.1.6 server that has been up and getting hammered for almost a year now without a reboot or a restart or oracle.
The article does not mention where they are getting the code. It is probably too big a job to start from scratch - so their only real choice is probably Postgres. Are there any other open source databases they can start with besides MySQL?
What would the impact be, compared to the fact that Oracle would soon be running on Linux and also Informix (acquired by IBM) and DB2 would soon be running on Linux clusters ?
Do you think Redhat could have some clout in convincing the market that it can build a solid database compared to the solutions already out there. Agreed, that Linux is a solid server OS, but that itself wouldnt do. RedHat needs to have a solid solution which doesnt depend on Linux, but be able to stand on its own, in terms of TPS and Performance issues.
Rapid Nirvana
Actually, I think most databases perform ok while their admins are on acid.
Portable versions of Firefox, GIMP, LibreOffice, etc
The article is a little light on the details, but it would sure be nice to see an object database bundled with the next RH release. The flat "tables and rows" paradigm (did I really just say that?) is ok for some things, but in most cases, I often ask myself "why not use XML instead?" An ODB would give me a better reason to use a database at all. It would be sweet to store serialized C or Java objects this way, or access them through CORBA.
Well, your fingers weave quick minarets; Speak in secret alphabets;
std::disclaimer<std::legalese> sig=new std::disclaimer; sig->dump(); delete sig;
Of course, why anyone would pay for something as simple as "rpm -i postgres-..." is beyond me.