Reading the various pundits talk about a slide from Sun featuring ‘Sun DB’ as a possible forthcoming offering, I’m struck by how far off the mark they seem to be. For starters, PostgreSQL, MySQL, DB2 and Oracle are out. They’re in the slide. They’re also big and successful, and have no need for Sun. Sun may have an established relationship with CA (Ingres anyone), but such relationships and dependencies are fraught with strife, and I just don’t feel that Sun are a company that will want to strike up such a dependency deep within their stack offering. And anyway, Ingres is already open sourced
So the external options are out. That leaves two options: Build or Buy. Building a database product from the ground up is a hefty process, especially given that the bar is now so high. Databases have to offer more than just SQL92 compliance: Businesses need solid repositories for their data, and they need to be able to integrate with the developers chosen tool set. This takes time, money and effort. A lot of it. Time isn’t something Sun have a lot of. And all that just to match what is currently out there and proven. Proving a solution takes time.
Which leaves buy. There are plenty of database vendors out there, not least amongst them Sybase, the database choice of much of the City, and the bedrock upon which Microsoft SQL Server was built (at least initially). Sybase is - seemingly - perpetually in a dire situation, with dwindling market share and a ropy financial position. It’s a good takeover candidate for Sun.
But I don’t think that Sybase fits with Sun’s strategy. They’ll want to own the product, and influence it along their proclaimed strategies of Grid. Sybase is a monolithic database platform, that doesn’t wed itself to anything other than big-iron boxes. It’d also take a long time for Sun to turn it around and get it doing what it wants. Time they don’t have.
No. What the pundits have got wrong is that they seem to think that Sun are going to be buying something NEW. What they seem to have forgotten is that three years ago, Sun picked up Clustra. Clustra had designed a database platform that scaled linearly (by merit of some very nice technology), ran on nice 1U boxes (the more the better), offered better than 99.999% uptime. Sun were early investors, and feared the product getting bought by somebody else as they raised further capital.
That’s right: Sun already HAVE a database product, and an excellent one at that. They just don’t sell it at the moment.
Clustra’s v4 release of the software boasted a SQL-92 compliant interface, plus they were making solid inroads into JDBC support (a type 4 driver was just coming out when Sun nabbed them). I know because I provided the UK support for a customer working to implement it for a new release of their web platform (funnily enough, one of the big drivers for them was the Oracle license fee).
On top of the actual software, Sun acquired a world-class development and support team based out of Trondheim in Norway. Always happy to listen to the issues that I raised, they worked to solve issues and provide support. It was a wonderful example of how keeping the developers and support close to the customer can pay dividends to both sides (CA might want to pay head regarding their support model). Their execs who provided the steering for the organisation just “got it”. All of them. I’d give a lot to work with such a team again: There was mutual trust, excellent delivery and a fun attitude.
Regrettably, Sun didn’t seem to (then) realise what they had bought, and simply pushed Clustra deep into iPlanet as a back-end persistence layer. Gone was the SQL interface. Gone the compelling scalability solution. As with so many ground-breaking technologies, it takes time to build leads, but we had more than a few clients ready to move forward, just as Clustra was snapped up.
Time passed. Despite a few rumours that the management at Sun might give Clustra the lease of life it so richly deserves, nothing materialised. Until this week, with the news of a possible open-source SunDB. The open-source idea is intriguing, although I’m suspicious they’ll go that far initially. There are some very distinguishing technologies inside Clustra that others would be interested in. MySQL in particular: It bought NDB - a similar concept to Clustra, but from Ericsson’s labs. NDB as a product is but a shadow of Clustra’s offering, and needs a lot of engineering to bring it (and MySQL) close to what Clustra had produced.
But let’s not forget that Sun are big on Open Source these days. So I might be wrong. And Sun are caught in the headlights right now, and desperate for a way forward. They need new revenue streams, and Clustra offers a compelling pricing model with reliability competitors (including Oracle) can only dream of. Oracle is putting it’s efforts at Linux, which will only hurt Sun in the long run.
Therefore I’m convinced Clustra will make an appearance in some form. It’s a strong response to Oracle. It’s too compelling a sell for Sun to leave festering in the bowels of iPlanet. It needs work on the SQL environment: People want and need more than just SQL-92 compliance and a JDBC driver. They want Stored Procedures, triggers and point in time recovery: All things that were on the planned feature list for Clustra when it disappeared into Sun’s bowels.
What it had out there was compelling. Better than five 9’s availability, linear scalability, excellent licensing model, cheap hardware requirements, and excellent recoverability and replication technology. You really could add two more boxes and increase the systems processing capacity in direct proportion (Oracle’s RAC is an 85% increase for an additional node (ie. 20% overhead), plus it’s focused on big iron kit, not small 1U boxes as Clustra was). Plus a world-class support team that’s still actively working on the product in Trondheim (proof enough that Sun have some inkling of what they have). It all fits with Sun on AMD/Linux, and Sun on cheap UltraSparc boxes. And with three years of existing infrastructure development, I’d imagine the product has actually moved on some.
So this techie for one doesn’t believe the pundit line being put out at the moment. I think Sun already have a world-class DBMS platform that’s ready to be re-released to an expectant world. But do they have the guts to go against the establishment, the relationships they’ve forged over the years, and put out a database of their own? I certainly hope so.
Entries (RSS)