We will not be buying any Superdome2's. The technology is not the same as prior generations and we don't trust the reliability of blades. Yes they are cheaper than the prior versions, but if they have twice the reliability as our HP Xeon blades we well have a lot of outages. It's a shame because our prior Superdomes had pretty good reliability. Our HP Xeon blades are a disaster and we don't put mission critical databases on them, only clustered apps and infrastructure.
Looks like we will be going to Power7 as our new CEO came from a prior HP to Power shop.
It's obvious HP is divesting in Itanium as the numbers don't justify the expense of large development. Intel is not helping HP at all with an old fab doing 65nm when their Nehalem chips are 45nm/32nm. We are also not impressed with the SX3000 chipset which only uses 3 of the 5 QPI links on the tukwilas in the SD2. The virtual connect is where HP is making all the profit with seemingly lower priced SD2 blades vs. the SD cell blades. We will put apps that we don't have the time to move to p7 to the blades but only the singe/double wide. The BL890 is missing the SX3000 chipset which it needs since the Tukzila has only 5 qpi links to try to connect 8 chips and I/O. It was funny the HP rep put up the BL890 architecture chart and we all tried to find the lines connecting cpu0 with cpu5,6,7 and they didn't exist. The answer was classic " i guess they must pass thru cpu 3 or 4".
Oh well it was nice while it lasted. Going to the AIX for Unix professionals next week. I would have fought this years ago but there is not future in HP-UX and I have 12 more years till I become a greeter at Disney.
We tested IVM last year and the overhead was prohibitive and Tukwila does nothing to address it as it still does not have hardware virtualization.
I could go on but then I would just be venting for HP dropping the ball...but that was three CEO's ago.....thanks Lew Platt
Crystal Shanda Lear