Dead Giant: Preparing for Life After Sun
IBM may have walked away from the table, but it's clear Sun is on the block. What's your next step, and where will open source software fit in?
Despite recent speculation, there is now no question that Sun Microsystems is positioning itself for an acquisition. But IBM isn’t buying. Regardless of whether Sun gets acquired, your server and storage strategy needs to be proactive.
Without getting into Java and MySQL (which will both survive in the absence of Sun), we will offer a few issues to think about if you’re a current Sun hardware customer. We used to worry that Sun would completely fail, leaving servers unsupported and unprotected by warranty, but that prospect was not very likely, so nobody took action. Before large firms fail they are usually acquired. We now know as a matter of fact that Sun is open to bidders, but we have no idea how a successful bidder might integrate Sun’s product line. The time for action is now.
Hard Decisions About Hardware
If IBM had bitten, would it have kept SPARC alive to compete with its existing RISC line (POWER processors)? No, only one would survive. RISC processing is a niche market full of historical holdouts. Current trends show that very few SPARC customers run “general” applications on Sun’s SPARC servers. By general applications, I mean things like the Apache Web server, Oracle, or anything else that runs on Linux as well. SPARC processors are reserved for the rare application that runs faster on them, customers relying on a vendor that only provides SPARC binaries, and the people who just haven’t switched yet.
The simple fact is that you can get a huge performance gain by moving the vast majority of applications to Xeon or Opteron processors. That is why Sun has focused more heavily on x64 servers recently, and also why OpenSolaris has reinvigorated Solaris x86 support. Whether Sun admits it or not, there are very few reasons to run SPARC. Surely Sun’s PR mafia will contact me about that statement, but the company’s own actions speak thunderously louder than words.
Now, do realize that SPARC server hardware is excellent. It is amazingly stable and feature-rich, but raw processing power is so much more important these days. Its long-standing reputation for mission-critical servers is why many IT organizations stick with SPARC servers, even when their applications could run on Linux or even x86 OpenSolaris. As mentioned, the holdouts are shrinking every year. Multi-core, multi-threaded processors just aren’t important enough to skimp so much on raw throughput. Sysadmins can scale applications horizontally, so they just add another server instead of trying to rely on SPARC marketing.
A quick note about Sun’s storage lines: They are mostly negligible. The products inherited from the StorageTek acquisition are very lacking in features and offer no benefits, so most IT shops stick with EMC or Hitachi storage. Their x4500 and 7000 series storage servers (or NAS, or whatever they are) are good products, however. Copycats already exists, and NetApp already competes in this space, so even if those products disappear it isn’t the end of the world. Their main benefit is ZFS, which is open source.
Many companies, especially in the financial industry, certainly do rely on Sun’s servers. The trend, not surprisingly, is that software companies are porting their applications to run on Linux. There are still many holdouts, though, so the disappearance of SPARC servers could be detrimental. That is why companies need to put pressure on their vendors to support Linux or OpenSolaris on x86 hardware. For example, after years of pestering, integrated circuit layout software giants Cadence and Synopsis now both fully support Linux.
Now is the time to start planning. Perhaps you don’t want to pressure your vendor to support open systems, because you feel the products are lacking in other ways. This is a good time to justify the move to another vendor. Of course that path is dangerous and a massive undertaking when talking about migrating to a whole new ERP system, for example. Peace of mind that open standards-based systems will never be eliminated should go a long way toward mitigating the risk of retooling.
Some applications and systems just rely on Solaris, not the SPARC architecture. This is good news, as OpenSolaris will surely live on without Sun Microsystems.
Open source software in general, and operating systems specifically, do not depend on any one company. People often talk about companies corrupting the spirit of the software, but these days it is possible that your proprietary software could just cease to exist.
Carr Was Right?
Nicholas Carr did have some insight to offer in his tragically titled 2003 essay “IT Doesn’t Matter.” His elementary comparisons of public utilities to IT systems generated the most critical responses, and rightly so. But here is a case where the comparison holds: the actual processors. If your business relies on SPARC, its value is linked to Sun’s strategy, and I think we can all agree that nobody wants that.
Computing, especially at the level of CPU processing, is a commodity. Treat is as such, and find a way to escape vendor lock-in. There is no reason for lock-in at this point in the game, especially if you’re locked up with Sun’s hardware.