The world knows that the world runs on Oracle, DB2 and SQL server—that part of the world that doesn't know that "everything" runs on MySQL, anyway. In fact, of course, life isn't that simple. What the "world knows" isn't the complete picture and there are many other databases (including industrial-strength Open Source relational competitors for the likes of Oracle) with extremely loyal customer bases.
Just now, I want to talk about only one of these alternative databases (sorry, IMS, Teradata, Pervasive, Postgres, Pervasive, SQLBase and others too numerous to mention): reports of the death of Informix IDS as a useful platform for database developers seem to be much exaggerated.
Of course, some of the impression that IDS was on the way out was IBM's own doing—it was going to be merged with DB2. However, as Philip Howard of Bloor has pointed out here; and as was confirmed to us by Jerry Keesee, (an ex Informix employee; now director of Informix engineering at IBM), that was then—the policy has changed.
According to Keesee, who executes the strategic direction for the IBM Informix products, there have been no plans to merge IDS and DB2 since IBM management in the area changed: "Both IDS and DB2 are great technologies", he says, rather as you might expect. However, there is lots of 2-way cross fertilisation going on (compression, for example, is something that DB2 has and IDS doesn't yet, but it's now on the IDS roadmap).
The positioning seems to be that DB2 is for datawarehousing and mixed-mode transaction processing (that is, for systems where datawarehouse and online transaction processing—OLTP—work runs against the same database; which we think will be the accepted norm in future). Informix IDS is for high-performance OLTP, especially in embedded (integrated) environments—because it is scalable, reliable, easy to manage and performs well. This still sounds like overlap to us (DB2 isn't that hard to manage) but presumably it will reassure existing (and new) customers in Informix's traditionally strong sectors: retail, telco, govt, banking and gaming applications.
One of the big IDS retail customers, Keesee told us, started to plan for a move off IDS at the time of the IBM takeover but has now just bought IDS 11—because it couldn't find anything that could replace it. Keesee says, "I believe that IDS was designed from the ground up to support distributed environments where there are hundreds or thousands of branches requiring remote data servers. We have examples of customers running tens of thousands with less than 10 DBAs".
There's a new focus in IDS on Availability, largely using its remote standalone secondary server capabilities. From one to many secondary servers can now either be co-located or operated remotely at a contingency recovery site. Auto or manual failover is available and its shared disk secondary server facility allows one copy of the data to be accessed by multiple database servers. Nevertheless, high availability still needs some advance thought, of course; although Keesee claims that with Informix, this is all in the planning stage, and not a continuing, operational, cost of ownership.
Now to our main interest in IDS; it certainly doesn't appear to be a dead-end as far as development goes. It now has new checkpointing and isolation facilities, the sort of things that are key to enterprise operations. There's a new Data Studio which delivers what seems to be a database-centric Application Lifecycle Management (ALM) environment, supporting both IDS and DB2 and, soon, competing databases (but probably not U2, IMS etc., which is a pity). Worth noting, too, is the Solid in-memory database acquisition. We anticipate good synergy with IDS, rather as TimesTen integrates with Oracle, as a very smart cache for real time applications, but Informix was still evaluating the potential of this when we were originally talking with Keesee. Now, however, he tells us that he expects to see the first IDS/Solid offering in 2008, with an initial focus on the real-time aspects of telco billing (using IDS to integrate this with monthly billing).
There's a new SQL-based Admin API for IDS—for developers, especially ISV partner developers. And a new Open Source OpenAdmin tool is written in php: Informix is really trying to capture developer hearts and minds, which (these days) means being receptive to modern "Agile" languages and Open Source principles. Informix's message to new applications developers and new ISVs is underlined with a free developer edition—that's also available on the Mac. Mac OSX is seen as a cross-platform development opportunity, especially for new developers in the entertainment space.
We think (having an IBM database background ourselves) that's there's a lot to be said for developers specialising in niche enterprise databases. There's usually plenty of work and limited competition; while, if you are properly trained in database fundamentals instead of the cosmetics of a single product, your skills are really very portable (for example, we think our experience with IBM's IMS hierarchical database gives us insights into issues that will affect XML data stores).
Similarly, organisational management needn't be too concerned about being reliant on IDS (beyond the usual levels of risk assessment / management and contingency planning, of course; even being reliant on a market leader involves some risk). IDS is apparently managing double digit growth and increasing its ISV partnerships—check out the Bell Micro City community site. The latest IDS open beta was lunched at Linux world last year and shipped in May. Its "Cheetah 2" release has been in open beta since February 2008 and completes the IDS story—for now....