Many Thanks Jesper - Good Points

6. OK, PRIMEPOWER not a Mainframe but certainly there's a clear mainframe design thinking input; cooling, metal, layout, three-phase power, cabling and more of this thinking than SunFire and in practice PRIMEPOWER had good availability. Not mainframe partitioning its true and oher aspects. On my watch PRIMEPOWER performed well in one account zero production errors against around 4% fail rate across 10,000 Sun servers, perhaps SPARC64 was helped by ECC on many paths.

1. Agree also IMHO. Sun's and Fujitsu's problem was a sales one and that ultimately was about effectively getting the message across. At times they were very well placed against competition, IBM, HP and failed to be able to take advantage of this in the market, especially well positioned once Sun had got over Fujitsu's system capability with PRIMEPOWER and the resulting M class systems. Can't comment on M vs. PRIMEPOWER as I don't know the newer machines.

Certainly IBMs Z system appears to be most interesting.

Oracle and Sun uncertainty has not helped in recent months. What's your view on today's clarification? Does this help? You are right about Fujitsu's virtualisation strategy. Again, like many of these things, if there is good growth more resources can be deployed to try and improve. Again is tied in to opportunity and market share.

Separate VM arguably better than Solaris container. Thanks for your performance detail that appears to show the power of POWER6 and POWER7 is very much better.

2. Sure, its a balance. Says who though? I guess IBM and HP with Intel would argue otherwise as they are both market leaders in their respective sections of the market. Overall if they make money on servers, I do not know. Server leadership a key for server vendor survival.

Agree it looks like a two horse race. Oracle and Fujitsu will have to gear shift to compete.

For it to be a three horse race, with perhaps another entrant would help keep the market competitive. It may well be that a new approach gains traction. Markets do tend to a duopoly in many cases with Other often being a large segment, often the largest.

It may already be too late. Agree Venus as is, doesn't look enough to be a game changer. You clearly showed this in your earlier post. Water cooled looks like a retrograde step. We know the problems with distributed arrays. Plus as you pointed out to change fabrication process fraught with risks, costs and difficulties.

Wonder what today's clarification by Oracle has said about this? What Fujitsu's answer would be to this Venus question. How the roadmap and catch-up is now rendered?

SPARC would have to be stronger for it to matter and to make it to the Intel IBM table to compete. Will need radical strategy and action for this to work. Time is clicking on.

Whats your view on Itanium and its future? Eclisped by Nehalem and successors?

What about the google server farm approach versus big centralised servers?

The big boxes appear to fall over in terms of the costs of supporting that type of architecture and maybe moved to water cooled is another indicator. Perhaps Intel architectures are headed that way too?

Its going to be interesting to see what happens as feature sizes shrink, if there is a move to more chip stacking or wsi arrangements.

Its also interesting if it becomes a two horse race IBM and Intel the costs of just staying in the game for both is increasingly astronomical. The margin for error perhaps reducing.

Having read Scott McNealy's 26th January 2010 memo, I can't help wondering and thinking if Sun would have survived and prospered independently had he stayed at the CEO helm. He'd had a long tenure.

Will be fascinating to see if Oracle are able to harness the Sun opportunity. IBM and HP continue to attack the Sun base.

Will need a tightly executed Oracle strategy to secure the base and grow Sun's infrastructure business.

Many Thanks Jesper.

// Ivan