Hi Glenn, On Fri, 10 Jul 2009, Glenn Lagasse wrote:
> * Alok Aggarwal (Alok.Aggarwal at Sun.COM) wrote: >> Hi Glenn, >> >> On Thu, 9 Jul 2009, Glenn Lagasse wrote: >> >>> This sounds great. I like the idea of having just one AI image that can >>> perform multiple duties. It simplifies quite a few things and could in >>> certain circumstances save time for deployers. You build just one AI >>> image and then deploy it in all kinds of scenarios. >>> >>> A question. Is there a reason you're limiting the combined AI image to >>> x86 only? >> >> There's no technical reason in limiting the >> combined image to x86 only. Infact some of >> the changes to live-fs-root and such will be >> common to both platforms. Others such as the >> changes to the GRUB menu obviously won't be. >> >> If we know of a use case for a bootable AI >> image for sparc, the design for it can be >> addressed as part of the current work as well. > > I don't know that I have any great specific use cases for a sparc based > bootable AI image other than keeping feature parity between the two > architectures for starters. > > While VMC is a real consumer for an x86 based bootable AI image, there > are other uses I think a bootable AI image would be useful for outside > of VMC that apply to both x86 and sparc architectures. For instance, > having the ability to take a piece of media to a machine, pop it in, > boot the machine and walk away while it's installing is pretty > convenient. It also helps out purely sparc shops with their current > dilemma of how to install OpenSolaris when they don't have an x86 box to > install OpenSolaris on which they can then install AI to (chicken and > egg problem, though possibly less of a concern once we get a text > installer included on some bootable sparc media). If we had bootable > sparc AI media, then they can just boot that on a sparc box, get a > default install and then setup AI and deploy to the rest of their > infrastructure. The sparc audience will find the media based text installer for sparc more satiating I believe. > The other use case which is perhaps even more valid (now that I think > about it) is that subsequent phases of the VMC project are going to > introduce support for creating pre-constrcuted LDOMS, in which case > we'll most likely need bootable sparc media. Now, that is a valid use case. > If the work to do sparc is extensive, then pushing it out to a later > phase is probably ok. But I think we're going to want to do the work at > some point regardless and now seems as good a time as any (modulo effort > and time required in addition to x86 work). The work actually isn't extensive and will mostly fall out from the work on x86 (with appropriate disclaimers applied). The bigger piece is rationalizing the user experience based on section 5.7 of the AI client func spec as that will require sparc specific changes. And, obviously testing will also need to take place. Alok
