We've all heard of the fab-less computer company: now here is the datacentre-less computer company; Sun in this instance. Sun intends to operate without a single datacentre by 2015.
That's the somewhat incredible vision described, in a sign of the times, in a blog post by Brian Cinque, who is a datacentre architect at Sun Microsystems. Given the sensational nature of Cinque's ideas we have checked, insofar as we could, his credentials over the weekend and, yes, he does indeed seem to be a datacentre architect at Sun.
Sun's IT CTO, John Dutra confirmed the vision although not the 2015 goal, in a Forbes interview. Dutra said that Sun will eliminate five of its eight datacentres by 2013 with the remaining three going sometime after that. Yes, Sun does indeed intend to operate entirely on a software-as-a-service (SAAS) model.
The answer to datacentre green house gas (GHG) emissions, ever-increasing electricity prices, and datacentre operating burdens is to do away with them. Stop treating a datacentre as an in-house DIY activity and rent computing power over the grid, having applications run 'in the cloud'.
Nicholas Carr, author of The Big Switch, a book on utility computing published earlier this year, which says that exactly this shift in IT is happening, said: "This is a very real sign of the shift from the private data center to the public grid for computing. It means that Sun expects the utility computing grid will be mature enough by 2015 to handle any type of software that the company runs."
By 2013 Sun will cut in half the total power needed by is datacentres today through a consolidation exercise, moving from eight to just three centres. He writes: "SunIT will move in an evolutionary manor from several generations of infrastructure, both technology and processes, to a model that is far more efficient with enabling factors like technology and IT Governance-like processes. As SunIT reduces datacentres due to higher utilisation functionality, there will be a point proven by metrics, that SunIT can only become so efficient. At that point SunIT must progress from a service-oriented architecture to more of a software as a service."
Why do this?
- Reduced IT costs thanks to a major reduction in datacentre space, power, and cooling requirements.
- Reduced IT costs and decreased time to implement projects thanks to a reduction in operations complexity for the datacentre staff.
- Compliance with Datacentre audit finding.
- A foundation for sourcing components.
- A transition to an “evolutionary change” infrastructure management style.
- Alignment with application, security, business systems, and agent architectures.
What technologies will be used to consolidate the datacentres? Cinque mentions these:-
- Server Virtualisation – Xen/xVM, Solaris Zones/LDOMS (logical domains)
- Storage Consolidation – Tiered storage, NAS, etc
- Application Consolidation – For SunIT its called IBIS
- WAN acceleration – Another interesting technology in which its roots have been around for a while.
- SaaS - ... SaaS plays a huge role ....
The SaaS facilities don't exist yet. That they will is a matter of faith I think.
What happens to Sun hardware sales?
Currently Sun sells lots of low-end, mid-range and high-end servers to business. It is widely distributing its open source Solaris software. It is developing server-boosted storage arrays s as Thumper and Honeycomb. What is going to happen to this business, which, if it goes away, will decimate Sun's revenues?
Sun CTO Greg Papadopoulos, has articulated a 'red shift' computing scenario in which ever-larger datacentres supply computing, as if from a cloud, to other businesses and to consumers. The red shift refers to the supposed accelerating rate in which they will scale out their datacentres. The idea seems to be that Sun will sell hardware to red-shift businesses, ones needing the ability to massively scale out server, networking and storage resources.
It will still sell servers. but in larger numbers to fewer companies.
Sun hosts its own cloud computing service, Network.com and intends to keep operating this, selling its cloud computing resources t other companies, but not using it to run Sun's own internal software. But, of course, Network.com could be out-sourced and run on someone else's cloud.
Secondly, Sun has its engineering IT resources and these may remain in-house and need a datacentre to run them. Cinque's vision could butt up against engineering's need for security of data and a need for its own resources. But then Cinque might say that the cloud wil provide both the security and resources required.
It is a little surprising that Sun CEO Jonathan Schwartz has not articulated this amazing vision. Perhaps he will.
Sun has certainly trumped - a Bridge card game term - HP's massive data consolidation scheme and it will be interesting to see how Sun's marketing messages shift to accommodate this zero-datacentre idea. For example, if it is telling customers to prepare for hollowed-out IT operations with no datacentres then why should customers buy any Sun datacentre kit, unless that is, they are going to sell cloud computing resource to others?
No doubt the Sun story will go on to say that use of Sun kit will best prepare the way, the evolution, to the zero datacentre end-point. Seeing how that story gets flashed out will be fascinating.