Monday I was too busy with helping customers optimize their current data centers with virtualization to write about the (finally) announced Cisco Unified Computing System. And I’m glad that I didn’t have the time, because now I can expand my original thoughts based on the rumors and reactions that IBM is in talks to buy Sun. I’m not surprised that in the same week we see the server vendor marketplace expand we also see possible signs of it’s contraction; honestly it is about time someone buys Sun, it it isn’t done soon their future could parallel SGIs and they just fade away from existence.
I have known about Cisco’s California Project Servers, now UCS, for a number of months now thanks to my role at VMware. Since the moment that I first heard about them and later when I learned a bit more about the architecture, I could see how they could help change the data center as well as the hard struggle Cisco would have to do it. One of the biggest use cases I see for UCS is with virtualized desktops. With all the memory these systems will have in them, uses should be able to achieve higher density numbers for virtual desktops per core than with other servers. Desktops typically are idle most of the time, the limiting factor on gaining higher consolidation density is memory, especially when you consider memory requirements for desktops will only increase over time possible eroding virtual desktop density numbers.
UCS’s promise of a centralized managed and clustered high density computing environment will enable data centers to consolidate physical space in their data center even further than could be accomplished with just virtualization. Making this a reality will still require some changes in how data centers are run. Primarily, not being afraid of running a server at higher than 30% or 50% utilization. While you can’t run them all at 100% utilization as there is no place to failover, I still see many people running virtualized environments in much too conservative a fashion. Let’s get server frugal, pinch that server penny to get all you can out it.
Cisco’s biggest challenge will be convincing customers to adopt these servers in large scale. I don’t know if pricing has been disclosed yet, but considering that Cisco’s classic play book is to operate in key hardware markets where high margins can be maintained, I expect the cost per physical device to be higher than the competition. Which might cause sticker shock for those who aren’t adjusting their thinking to cost per virtual machine from cost per physical server. In addition, most of the big customers that Cisco will need to go after already have existing server vendor relationship in place. And that server vendor relationship is with a vendor (i.e., the top three server vendors of IBM, HP, and Dell – see ZDnet’s Larry Dignan’s post) that sells more than just servers to a company. Most of the large enterprises have pretty straightforward formulas for dealing with their strategic vendors, Cisco gets X-ish% of budget, HP gets Y-ish% of budget, and IBM gets Z-ish% of budget. If a customers is going to increase Cisco’s percentage, where is it going to come from? Another vendor? At what risk to the rest of the operation? Decisions like this are highly political in nature. It will be interesting to see who decides to upset their politics.
I have been wondering for the past year or more “who will end up buying Sun?” Years ago I heard stories from high level Sun sales execs about how Sun’s upper management didn’t get that it’s not about the servers but about solutions. Last year at Startup Camp in SF, Johnathan Schwartz gave the kick off address. After talking for 15 minutes about how people don’t know what Sun does, I was expecting him to wrap up his speech with a clear statement and direction for the company. Nope, he left it hanging out there…as if he was saying “no one really understands what Sun’s goal in life is…including me.”
So, the real question of the day is “who is considering buying Sun?” Many rumors that the IBM leak was just to drive up the price for the real bidder. Some think it could be Dell, which would be a good way for dell to capture more market share but still keep it in the #3 spot. Some think it should be Cisco, which would help Cisco catapult into 4th spot in the server market but hard to do considering all the money they just spent on the UCS development and launch (though some have been hypothesizing about the large amount of money Cisco borrowed a few months back, last time they did that was right before they bought WebEx…). I find it interesting to think about Apple buying Sun, how much longer can Apple continue to grow without putting both feet in the business pool. Apple wouldn’t want to do that on the desktop side since they would then have to start exposing future direction to business customers to allow them to plan for future changes (which ruins their big launch surprises) but doing this on the server side allows them to keep the Apple shock factor…though I don’t see the long term strategy there.
Either way, I think it will be inevitable for someone to buy Sun. Investors want their money back.
Third, Data Centers
It is inevitable that partially alone, but definitively in combination these two announcements will impact the data center of the future. The data center of future running on top of a virtualized data center operating system has to built out of standardized and common hardware building blocks. This means less hardware vendors and more innovation in the automation and services front. Needless to say, the data center strategy discussion I will be having in the near future with one of my customers will be very interesting. Partially because of the vendors that will be in the room and partially because of how the customer will be reacting to each of them based on these announcements.
Even with all the economic woes and recession depression, at least the shock waves from these two announcements will keep the server technology market interesting for the rest of the year. I can’t wait to see how this all shakes out over the next few weeks…
~ ~ ~ ~ ~
If you haven’t seen them, here is a short list of interesting articles I have read on the Cisco announcement:
- Cisco seeks for data center what Apple created with iPhone — a new market that stops the madness
- The good and bad of Cisco’s UCS servers
- Scoble’s video interview on UCS architecture
Favorite quote from the Cisco buzz: “Cisco and Intel are joined at the chip.”
And a second set of interesting articles about the Sun purchase announcment: