Peter Tribble wrote: > On Wed, Jun 17, 2009 at 2:43 PM, Dave Miner<Dave.Miner at sun.com> wrote: >>> Why is wanboot used at all? In addition to not having support in a >>> decent fraction >>> of the installed base, it's simply far too slow in my experience to be >>> useful. >>> >> All things being equal, wanboot should be significantly faster than standard >> tftp-based boots since you aren't stuck with a lock-step protocol. That >> said, there are obviously bugs in some platforms that need resolution. File >> bugs, as always. > > But ideally the tftp (or wanboot) phase would only drag across a meg or two. > Ideally you want to get out of what is essentially firmware and into something > more capable as soon as possible. Which leads to another related question: > why is the initial boot image so huge? >
Because it's the complete booted root file system that's loaded into the ramdisk. It's not using NFS. >> The installed base issue is a rapidly declining problem, because essentially >> all of the non-wanboot-capable platforms will be out of service life by the >> end of 2010, a very few outliers go into the first half of 2011. > > It wouldn't surprise me to still be running non-wanboot-capable platforms in > production at that point. Let alone in testing - I might have been able to > grab > something like a 280R for testing. Sparc hardware in particular is extremely > long lived, helped by a healthy 2nd-hand market and the fact that speed > improvements have been - at best - modest. > It's unlikely we (or any profit-making entity) will design products around the capabilities of platforms that can no longer generate revenue. Dave
