Google, PDPs and OpenSolaris put readers to the test
Can you make it to the end?
We clearly have no shame or problems with full disclosure, and our story OpenSolaris ready to power up on IBM's PowerPC requires a helping of both.
First off, we fell into the trap of thinking the OpenSolaris port was aimed square at IBM's version of PowerPC. Not true. As we amended the story to say, it's Freescale country first.
Second, we weren't quite explicit enough on how long it will take for an actual living and breathing version of OpenSolaris to run on this chip.
An unnamed reader put us in our place.
The "Polaris" announcement was that they got the thing to compile. Further, they claimed that this constituted a port, but that is risably far from the truth. There are been little to no code actually written and there's a huge amount of PPC-specific code that would be required for something that could plausibly be counted as a port.
Several of you had similar feelings.
SunLabs did the port to PowerPC for v2.5.1, so it's a mixed effort of FOSS folks and SunLabs (SunLabs released their code to Blastwave). To date they have compiled the operating system (I am not entirely sure if it's Sol10). That's it. It has not been run.
They need to get it to boot, run reliably on modern Power architecture, they have to get a set of useful device drivers. Dtrace will have to be substantially written for Power PC if they want it. The whole thing will have to be QA'ed. If you are running this in a datacentre you are looking for reliability on a par with AIX, Linux on the same kit. And then (if you did want blades) you have to tie in some observability tools wth the hardware. And if you want to support decent IBM hardware, SMP.
This is a long and winding road, and is discussed in the dev groups. I really do hope they are successful, but please be realistic about the speed at which this will happen.
Read the article with interest, Sun originaly ported Solaris 2.5 to the PowerPC (For Motorola and IBM) back in the late 90's. We couldn't get customers or ISV's interested, even when we showed the speed of PowerPC vs SPARC and x86 vs SPARC (SPARC always came last).
Now IBM will have an OS strategy.
And the folks behind the project had this to say.
In our opinion, the Sun approach makes sense:
1. Tap into an open source community. They started with the OpenSolaris Pilot Project and they kept their eyes open. As the opportunities arrived, they channelled them. In this case, it is Blastware/PowerPC (vs. Blastwave/x86-SPARC). The Blastwave Community already existed, had a proven track record and do a good job. It was easy for Sun to leverage Blastwave. They just watered the plants that grew. Blastware was one of the unexpected plants that popped up.
2. The ODW is an open, flexible platform. The developers themselves chose it. Sun saw we had a community around the platform and Sun management channeled opportunity toward us - it other words they honed in on the opportunity as it proved itself by itself with time. It was probably just as hard for Sun management to do this internally as it was to do so externally. They have done a great job on both counts. Don't forget the some of the Sun engineers leading to Solaris open source charge today might have been the same ones that five years ago would have thought that unthinkable.
3. Once the ISA and platform decision was made, the question became where to head first. Embedded markets make the most sense - higher margins, less competition, and even some complacency when compared to the supercharged-everyone-is competing-against-everyone in the commodity oriented Intel space. Embedded development is more specific and more targeted. The 32-bit PowerPC of Freescale has followed a more traceable evolution than the jump to hyper-space with the 64-bit G5 that is just finding its legs of IBM. Solaris 2.5.1 was built for PowerPC. In contrast, an IBM offering would need to be supported by a more complete package than what could be useful to Freescale sooner. So, Sun started there. Instead of maybe needing all things possible in the IBM environment, Sun could focus on some specific opportunities around Freescale chips that would not require a full suite of OS support and this would allow them to move into markets where they were not currently. It is a growth opportunity!
In our opinion, Sun has the most compelling big company vision for the future of this market. We respect Sun because they act on that vision. Being a leader is not waiting for your customers tell you what they want, it is understanding and anticipating what they want and being there with it when they ask for it.
Raquel Velasco and Bill Buck http://bbrv.blogspot.com/2006/01/why.html
Well, my conscience feels unburdened. Have a great weekend. ®
Sponsored: Hyper-scale data management