OpenStack is an example of an open source cloud stack. With support from open source, the cloud could become a lot more ARM and in general low cost alternative friendly. By design the cloud stack virtualizes the hardware which means it shouldn't matter to a user of the cloud what actual hardware in the datacenter is running the stack. This definitely favours ARM as it gets rid of Intel's x86 compatibility advantage.
From the q3 conference call......
Jim Covello - Goldman Sachs
So my follow-up question stands specific to the Data Center Group, some of the ARM based server players are arguing that they can now address a significant part of the workload from the Google’s and the Amazon’s and Facebook’s data centers. Is that some competitive dynamic that you're seeing in that area or do you think that ARM still isn't competitive in that realm?
Paul Otellini - President & CEO
They need to add features sets like 64-bits and ECC and RAS features to be particularly in those environments to be considered, so that maybe a roadmap planning opportunity that they are pushing but the products that are being shipped today certainly don't have those feature sets.
You can look at some of the workloads, things like Hadoop, Jim that would be conducive to let me say in a ray of micro servers, and those can easily be run on Atom; we've got our second generation of the Atom micro server chips out now, the first one is on 32-nanometers. Now we’re now sampling the 22-nanometer one and what we’ve decided that we are just going to push Atom as hard as possible in this space and have it be a better offering for our customers than having to switch all their software and worry about all the reliability features.
1. The servers will have to be Intel x86 with strong performance that will be crucial with all the extra load.
2. Most likely the client retrieves data from the cloud and runs it on his local machine - security reasons -- again means x86
3. Some non important apps will use ARM - email, basic games like angry birds etc
OpenStack is a project started by Rackspace and they shipped their first release in Oct 2012. The Platinium and Gold (Intel) members do most of the heavy lifting. I don't recognize any ARM representation in the group although there must be someone. All the work was done in Python with Linux defining the API. Getting OpenStack to run on some ARM CPU that runs Linux (Android) should be pretty easy.
OpenStack is not "ARM compliant". OpenStack conforms to the OpenStack specification. OpenStack will run on an ARM chip if the that ARM chip and its version of OS provide the proper OpenStack API support.
If the purchasing manager decides that the ARM based server offers the best solution to his problem, he will buy it. If the approving manager decides that there is a better solution, he will buy it.
He will decide based on performance, power consumption, cost, how it will fit into his existing equipment and future plans.
Will the ARM server be faster? No. It might be for FP workloads but nothing in the OpenStack.
Will the ARM server be lower power? No. Intel power optimization, advanced process and code restructuring will easily match any ARM vendor offering.
Will the ARM server be cheaper? No. Intel cost per transistor will make it easy to set their price right on top of ARM vendor products.