IBM To Purchase Informix Database 228

Who the fuck said MySQL was the open source community's flagship database?

Honestly -- we have PostgreSQL (which at version 7.1 is actually a passable database for smaller implimentations), InterBase, and many others. The state of a single product says nothing about the OSS community as a whole, and it's quite improper to suggest otherwise.

That said, I utterly agree that Oracle and DB2 beat the cr*p out of their OSS competition -- but that doesn't stop me from using PostgreSQL when my needs are modest and my budget slim.

I would certainly agree that Oracle on Solaris is more scalable, bulletproof, karma-riffic, etc. than MS SQL server, but you only need an aircraft carrier when you are fighter planes at sea. If you are just going fishing, a rowboat is a much more useful craft.

While it certainly is true that Microsoft has hyped their database as being capable of things it really isn't capable of, for most projects it is perfectly adequate. Of course, in that same vein PostgreSQL would probably work as well, and it is a heck of a lot cheaper than either Oracle or SQL Server

"50,000 ATM-style transactions per second logged to disk with full recovery on a single cpu. This was against a database of over 100,000,000"

"on the 35 megabyte OLAP APB-1 benchmark queries, Kdb ran 12,000 queries per minute with no precalculation."

"TPC/B benchmark that achieves 25,000 transactions per second with full recoverability and TCP/IP overhead on an (167 Megahertz) UltraSparc I."

"K is also small. The entire runtime system, written in C, fits in 300,000 bytes."

More to the point, it's playing in a completely different market to all of the others. It isn't, and probably never will be considered a replacement for Oracle, because it's not SQL based. It is, however, a fully fledged database, supporting transactions, fine grained locking, online backups etc. Also, anyone that thinks MySQL or PostgreSQL are players in the database leader struggle is dreaming. Sure, they're fine databases in their own right, and in time, they well gain some of the features that they're missing. They're fine for small to medium businesses, but for enterprise use (which is where Oracle and DB2 reign supreme), they're just not even close.

Where did you hear that?

NASA uses Oracle, MS SQL, Access, DB2 and a bunch of other databases. NASA is not a monolithic organization that dictates what software can be used. Each project makes its own decisions as to what software is the best fit for its needs. It could be Linux with MySQL, NT with MS SQL or Solaris with Oracle. If you name a software package, there is probably a NASA project that uses it.

Of course, it's a turn-on if you don't want to wait around for SQL parsing. To each his own, I suppose.

Doesn't one of these seem a little out of place?

Mark: "We're going to need SPEED! Let's use MySQL."

Bob: "This is going to be a high transaction session database! We're going to need transactions and rollbacks! Let's use PostgreSQL."

Dennis: "We'd a crummy little dot-bomb. We don't need speed, and we don't need transactions and rollbacks. Let's use DBM files."

-James

It can't be the software, which was crap. In 200 lines of code, I wrote two different test cases, (only one of which was multithreaded), which crashed the Informix server.

It can't be the support organization. Getting help from Informix support was a surreal experience. There was the time I had to instruct one of their support guys how to unzip a zip file. I had to explain to another one the concept of a client, and introduce the fact that Informix was accessed from one.

It can't be the advanced R&D: The aforementioned Illustra was surpassed in all ways by IBMs research out of their Santa Teresa Labs, and some of this research has already found its way into DB/2.

Customer base? I didn't think Informix had that much of a following.

So what is it? What? I just don't get it.

IBM deserves any and all DB misfortune (including pissing away a billion dollars) for the crime of destroying the far superior QUEL query language from Ingres with 'the EBCDIC of query languages', SQL!

But on the other hand, IBM deserves any and all relational database glory for employing E. F. Codd, who wrote the innocently titled paper "A relational model of data for large shared databanks" in 1971, which started the whole field. Given IBM's previous monopolistic tendencies, it's sweetly ironic that they end up spending a billion dollars to gobble up *any* other RDBMS provider when they used to *own* the field, lock, stock and barrel, starting with their own System R. Indeed, IBM Japan used to brag about it:

ALL YOUR DATABASE 'R' BELONG TO US!!

(Sorry...it just had to be said. :-)) Meanwhile, with their purchase of Informix, IBM has probably stomped out the last possibility that any form of QUEL would ever make any comeback, given that Informix had bought Illustra which had commercialized Postgres, which originally spoke Postquel, the follow-on to QUEL after Ingres had gone commercial. That is, unless the developers in the PostgreSQL project miraculously resurrect it themselves...

As has been pointed out by others Postgres, MySQL and Berkley DB aren't players in the same area as DB2 and Oracle.

However, there are a couple of other surprising omissions. Sybase ASE 12 is a pretty nice database, and is very competitive feature and platform wise with Oracle and DB2, and probably has a bigger market share than Informix. MS SQL Server 7/2000 is also a very nice database to work with. It's use is growing quickly for good reason - it's fast (on comparable hardware), cheap and the SQL Server development tools kick Oracle's Ass. Ever used MS Query analyzer? It is beautiful.. and comes free with SQL Server licences. You can get third part equivalents for Oracle (eg, from Quest), and they are also nice, but they cost around $10,000 for a site licence.

No, it doesn't run on non-Windows platforms, and yes Oracle on high end Sun hardware will run quicker. However, there are probably less than 5000 companies in the world that need that much power - and MS is going after that, too with MS Windows Data Center.

I'm not a MS weenie - I like an Oracle DB as much as anyone. However, it isn't as far ahead of SQL Server as some of you seem to think - and some of the bugs in it [sun.com] are just as bad as anything you'll see in SQL Server.

Informix has two main enterprise products: IUS (Informix Universal Server), which is the result of the Illustra integration; and IPS, which is the massively parallel server. IPS is actually pretty widely respected and used. It's a monster for data warehousing, and is very commonly used in large (i.e. bricks-and-mortar) retail installations. In fact, a very common installation platform for large retail is to use IBM (mainframes, AS/400, and POS terminals) for the transaction processing, and IPS to handle analysis.

So in that sense, there's both interesting technology there (the massively parallel bits in IPS) and a very good synergy of customer base and products.

So in direct response, there IS interesting R&D out of the Portland lab where informix did the XPS work (I forget if they're actually calling it XPS or IPS these days, it used to change around a lot), you just didn't know it.

And the research you're probably thinking of comes out of Almaden. There's some interesting DB/2 related work out of Santa Theresa, but Almaden is where the really cutting-edge stuff has taken place IIRC.

The main players in database leader struggle will be Oracle and IBM

<naive> Microsoft & SQL Server? </naive>

I don't really keep up with such things (though I probably should), but does this really mean that "no one" is running SQL Server? I thought it was doing well enough that some naive people -- marketing drones, purchase mismanagers, etc -- see the term "SQL" as being synonymous with the M$ product instead of, oh, say, 'structured query language'.

I'm not even trying to start a flamewar here (though Slashdot is oh so good at that), but I didn't think M$ was a player to be dismissed in this area. Am I wrong?

My half-paying-attention understanding was that, among "serious" server systems (that is, neither mainframes nor desktops), the main databases included Oracle, DB2, Informix, and SQLServer/Sybase (with several others, notably open source ones like MySQL and PostgreSQL, trying to get a foothold).

Given that, I was a little bit surprised to see that SQLServer wasn't mentioned in the article or writeup. I realize that it could be seen as, oh, say, "Access Server Edition", but I also realize that real companies are putting it to real use and are quite happy with it. As a flagship demo, Terraserver [msn.com] runs [msn.com] a pretty serious load [msn.com] on such a Windows [msn.com]/Intel [msn.com] based SQLServer [msn.com] system.

Now I realize that that might be all marketing hogwash, tweaked to hell to handle that kind of load or running an application that may be of no relevance to other uses (e.g. web business etc). But nonetheless, it seems pretty respectable to me. If that's "a toy", it looks like a damn powerful toy. Like I say, I don't really keep up with such things, and maybe the competitors can do even more interesting workloads.

I'm here to learn -- enlighten me, flamethrower.

C:\WINNT\system32>strings C:\WINNT\system32\FTP.EXE | grep -i "copyright"
@(#) Copyright (c) 1983 The Regents of the University of California.

C:\WINNT\system32>

Customer base? I didn't think Informix had that much of a following.

It may not be worth $1B, but Informix does have a fairly large list of prominent customers.

When I worked for an all-Informix+IBM RS/6k house, we were told that the company did some $BIGNUM%, where $BIGNUM > 40 and $BIGNUM < 100, of all the real estate-based credit reporting in the US. As another poster has said, they also hold Verizon and Deutsche Telekom as clients as well. I don't call that small :-).

I do concur about their support being surreal, though, and some of their DB Servers were extremely flakey, and as for R&D, well, you already said everything I was going to say :-P.
--

Then why is it so hard for me to find someone who can tell me what makes Oracle better than Postgres?

Every time I ask someone what makes Oracle so much better they either say "Enterprise features!" and change the subject or say "Oracle has X!" making obvious their ignorance of Postgres since it has X too.

As of PostgreSQL 7.1 Oracle is not longer the faster of the two. Thanks to GreatBridge, Postgres has awesome support. Postgres now has all the foreign keys and inner and outer, left, right, up, and down joins. It has a data management layer where you can stick your indexes on a different spindle than your table. It has stored procedures and triggers. It has full transaction support and a write ahead log.

Come on! Will one of you Oracle advocates PLEASE tell me what the hell makes Oracle so damn good. I'm not trying to pick a flame war, I just want to know! Has Oracle really fallen so far behind that the only people who advocate it are those who are drinking too much of the Coolaid to form a real argument? There must be another reason people like it so much!

You still can't even make a simple nested SELECT in this thing !

Its only grace is that the little it does, it does fairly fast.

OSS has proven the old tenet that "you get what you pay for". Which is buy business users pay for Oracle, DB2 and SQL Server. (Well, that and the cool schwag the marketing people bring. "Well Oracle brought me a travel mug, but Microsoft brought this cool bomber jacket! I'm going to make a business decision** and go with SQL Server."

** "Business decision" is dronespeak for "random yet biased stupidity".

--

That being said, I have found some very useful and informative and correct information, many of them coming from the community that make up Slashdot. I find it disappointing when the editors who are responsible for deciding which stories to post and for adding editorial content don't get their facts straight, and depend on the community to correct them.

Do you know where they derive their profits from? Do you know how much profit did IBM turn on Linux specifically? I'm sure IBM can make Linux more profitable for them in the future than it is now, and it's just a matter of time. I have no doubt that open source and profitability are not mutually exclusive goals either.

In fact, I think the best way to make open source and Linux profitable is what IBM is doing, not necessarily what RedHat is doing. IBM provides more solutions. Whereas very often, RedHat just provide more questions (ok, it's just me).

Oh, and don't mention the Transaction Processing Performance Councel, by Performance or by price/tpmC (a hint: MS has 10 of the... top ten), or heck, just overall!

Yes, MS has made some mistakes in the past, but they are learning from them and are making a quiet comeback. Nothing comes close to touching thier data mining/warehousing product.

Sorry to burst your bubble, but in terms of raw performance, MS SQL server 2000 doesn't cut it. Those performance figures are very interesting, so lets take a look in more detail.

In third place in raw performance is DB2 UDB, running on 128 700Mhz PIIIXeons. This manages 440879.96 TPC-C throughput.

In second place in raw performance is MS SQL2000, running on 192 700MHz PIIIXeons - 50% more processors than the DB2 UDB result. And the TPCC throughput? 505302.77 - a mere 15% more throughput. Not impressive.

In first place in this raw performance chart is another MS SQL2000 result, running on 280 900MHz PIIIXeons. Oh dear - they added another 50% more processors, upped the speed to 900MHz per chip and still only managed another 36% in TPCC throughput. I reckon that a linear fit should have shown about 55% more performance than their second place result to be competitive.

So you see - while MS has the money to buy lots of equipment to get impressive TPCC scores in raw performance, they need far more grunt from their hardware to provide equivalent performance to DB2.

Cheers,

Toby Haynes

DB2 developer and therefore biased :-)

That's in the commericial space - obviously SleepyCat, PostGres and MySQL and others aren't going away.

Ummmm....Sleepycat is a commercial embedded database. Sure, it's Open Source, but it's still commercial. The two adjectives "Open Source" and "Commercial" are not mutually exclusive.

You pay money for support. When your entire company rides on the sanctity of a huge database (like a bank, for example), newsgroups just don't cut it for support. Oracle offers time-assured support (i.e., you database will be up in x hours under this support level).

DB2 is actually a fair bit cheaper than Oracle, especially when you start running it on fast SMP machines. Oracle charges you based on the number of CPUs in your server, multiplied by the CPU speed in MHz, multiplied by a $ amount. You also pay an additional premium if your CPUs are RISC, rather than plain-jane Intel.

DB2 is priced on a flat per-CPU basis, irrespective of CPU speed. Basically you're penalised for running Oracle on fast CPUs, whereas with DB2 you aren't. Run Oracle on newer high MHz CPUs (like an Alpha, UltraSPARC III or PIII-Xeon) and your wallet starts to bleed pretty badly.

...I worked on the frontline re: RDBMS for nearly a decade now. And the support I partially experienced from the big boys can partially only be described as shoddy (to put it mild).

Of course that doesn't invalidate your argument. When something goes wrong, as it inevitably does, you're sure better off when you went for Oracle or so. However,: support per se is certainly not better (I'm convinced that e.g. Great Bridge provides good support). It just looks better in your memo to the vice president that explains why the database fscked up big time.

Currently, the best version I've heard is JBase, which allows coding in C, which addresses one of the great weaknesses of Pick, having to code in Basic.

--

The great downside of Pick basic is all the crappy habits bad programmers leave in it and that variables can be of any type (only two firm types, file variable and everything else!) Petty annoyance... having to do this check:
IF (Y.Q1.AMT # 0) AND (Y.Q1.AMT # "") THEN ...

Multivalue fields are actually pretty cool and easy where Unidata was concerned, you could create parenthetic selection clauses.

--