Derek Vadala thinks Sun's Jonathan Schwartz is living in denial about Linux:
People who have to administer servers like to run Linux (compared with something like Solaris) because it's easy to get help, easy to find task-specific tools that work without much tweaking, and easy to quickly install a reasonably secure system with a sensible (useful) software map.The whole issue of free software and plentiful "help" via a bazillion Linux mailing lists is a red herring. I used to think that Linux was going to kill Solaris, until someone explained why it hasn't happened yet, and won't happen for quite some time: Sun's large corporate customers are exceedingly conservative, and unlikely to switch from Solaris to Linux for moral or feelgood arguments. When Linux is a cheap drop-in replacement for Solaris, these customers will use it. But Linux isn't there yet -- not on the big iron, not for the high availability solutions that have been running for years on Solaris.There's also the whole free software thing, but if Schwartz would like to pretend that's a non-issue for his customers, I won't argue the point. It's not that important (at least to me) in the Sun/Solaris versus Linux debate.
Many companies are already working Linux into their IT plans. But what do you do about the existing mission-critical services running on Solaris? In many cases, Solaris x86 is the perfect transition technology: all your certifications and training remains intact, all the apps keep running, and you're using commodity hardware without porting to Linux.
These same conservative corporations are planning to use Linux (or *BSD or ...) in the back office sometime within the next few years, when the issues of porting to Linux and running Linux on big iron are better understood. Until then, it's foolish to move away from Solaris or Solaris x86, which are working very capably today.
Is Jonathan Schwartz being simpleminded and pretending free software is a non-issue for his customers? I think not.
"In many cases, Solaris x86 is the perfect transition technology: all your certifications and training remains intact, all the apps keep running, and you're using commodity hardware without porting to Linux."
I find the first argument - familiarity with Solaris as an environment - is about 90% of the argument. Solaris applications can usually be recompiled with no modifications necessary, but that largely depends on the application. There are surprising gaps in the commercial market for binaries for Solaris/x86. And commodity hardware support has always been poor. (The main reason I moved off of Solaris/x86 and into Linux in the first place.)
"These same conservative corporations are planning to use Linux (or *BSD or...) in the back office sometime within the next few years, when the issues of porting to Linux and running Linux on big iron are better understood. Until then, it's foolish to move away from Solaris or Solaris x86, which are working very capably today."
Certainly. There was one point made in Derek's article that I think is almost spot on: Sun needs to invest in getting Linux up and running on their enterprise machines. (Although I don't agree that they need to go the route of SGI and run their own version.)
Production-quality Linux running on their server set can help turn their SPARC family closer to commodity (in the fungibility sense of the term), which may be SPARC's only saving grace (if the UltraSPARCs IV and V perform as well as Sun thinks they are going to perform). A reinvestment of technology into Linux - like SGI and IBM have done - can help bridge the gap in the training differentials. And, of course, their killer app would be a compatibility environment, so that they could run most Solaris/SPARC applications on Linux/SPARC.
As it stands today, I don't hear of them pursuing any of this. They did back away from their own distro to supporting Red Hat and SuSe. But I've been completely unimpressed with the hardware Sun is running them on.
Re:Here Comes The Sun....
ziggy on 2003-05-28T02:16:13
The subtext I found in Derek's piece was "Linux is here to stay, Solaris is dead, but Sun hardware is kinda nice." His tone was more of Solaris being irrelevant (Solaris x86 doubly so), and Linux being the wave of the future because "4 out of 5 sysadmins surveyed prefer it!" That logic, quite simply, is facile. It ignores the needs and desires for conservative IT departments to remain conservative and migrate on a five year IT plan.There was one point made in Derek's article that I think is almost spot on: Sun needs to invest in getting Linux up and running on their enterprise machines.As far as Sun getting Linux running on their big iron, I'm not convinced that this is the best path for them to take. The market for their big iron is pretty darn small; I don't think they will double their sales if they sold a Linux-based StarCat, or offered parallel Linux and Solaris partitions on a StarCat. And it is certainly more work for them to upgrade Linux to run on enterprise grade hardware than it is for them to maintain and extend Solaris there.
Nevertheless, IBM is doing a bang up business with Linux on the zSeries. And IBM is leading their customers to Linux instead of the other way around. But IBM is working a different set of markets with the zSeries, so perhaps these market factors don't apply to Sun.
Of course, this may all be irrelevant if Sun can't keep the price/performance numbers up on the UltraSPARC...
Re:Here Comes The Sun....
Whammo on 2003-05-28T22:41:16
Coincidently enough, I've got a meeting with Sun in the morning to talk about their Linux strategy. (Go figure.)
