Why is (the) Sun trying to peak out from behind the clouds (of computing)

Maybe later on this week I’ll write up a more detailed set of things I’ve been thinking of, while I see another “fad” grip the computing world.
Until then, John West at InsideHPC.com asks a question of why is Sun coming back to the cloud table? John points* out

Anyway, you may recall that Sun has been down this road before, first with SunGrid and then the Network.com rebranding, which culminated in a shut down back in December. I think it’s very odd, and disturbing if you are a Sun customer, that they aren’t using their own datacenter, especially since they have spilled so much ink talking about how great their datacenter guys are (oh, say, here, here, and the 0 datacenter dust up here, and let’s please not forget the whole Redshift “our datacenter will be one of five one the planet” thing).

I’ll save my “cloud” comments until the future article. Though I might point out that for a new investment in technology to build a cloud computing facility, the business model generally needs you to spend as little as possible per system and per gigabyte of storage.
I don’t see how that’s possible using Sun gear.
On the server side, its hard to out-Dell Dell (or out-HP HP), and on the storage side, our list prices for JackRabbit and ΔV are well below their sale pricing.
The economics of real utility cloud models (ala Google, Amazon, etc) require you drive your costs into the ground. Pull out every last un-needed element, every high cost item. Low cost, highly replicated, easily managed environments. Marginal costs have to be driven toward zero for management, per unit costs have to be driven as low as possible.
* I should also note that any time I have run a post any bit critical of Sun, its business, or financials, or market, or technology, or discussing competitors, or better products, we have been DDoSed. I hope John’s servers can handle it.

4 thoughts on “Why is (the) Sun trying to peak out from behind the clouds (of computing)”

  1. if ruling the cloud is really about driving costs into the ground why isn’t there a vendor making wafers for the purpose? Intel should be dominating the cloud computing market just by plugging in their inventory.

  2. Driving costs down isn’t sufficient to rule this space, but it is necessary.
    Each clock cycle will have an effectively fixed cost to it. Clock cycles are (largely) clock cycles. You may get a doubling of efficiency per cycle or a reduction of efficiency per cycle, but rarely are you likely to find clock cycles which are otherwise identical to each other with significant price differences attached. They higher cost clock cycles will be largely unused.
    Or you can sell the clock cycles at a loss. make everything the same. Then you need to work to minimize your expenditure to maximize the utility.
    Cloud computing is all about utilization rates. The business models (that may work) depend upon high utilization to keep the marginal cycle costs low. This is why VMs are so popular in these models … multi-tenancy at the server level, with “machine” level isolation.

  3. It strikes me that in the near future there will be accusations of “overbooking” resources, like airline seats.

  4. @Peter
    Heh … We see this now with quad core on memory intensive HPC codes. Its pretty painful. John West at InsideHPC had a link to a study at Sandia indicating that it’s only going to get worse with more cores.

Comments are closed.