Sun launches their new Eco Center Solutions and shows the world how they implemented a state-of-the-art Green Data Center in Santa Clara, California, USA.
First Sun Microsystems launched Project Blackbox, a portable and energy efficient data center in a shipping container in 2006, and this year they have launched their three new Green data centers in Santa Clara, Bangalore (India) and Blackwater (UK).
In a video about their data center transformation, they note key considerations such as remodelling an old building to become an effective data center and reducing the square footage of data center space through infrastructure and data center consolidation – challenges a lot of IT departments face in their own data centers.
So how did they achieve it? Sun determined there were six core areas of focus:
Sun had to overcome internal barriers between departments, such as facilities and IT. The consolidation of shared services, hardware, space and resources was a given – and it paid off: Sun estimate they have saved US$1.1m in utility costs. However the human element could not be ignored: Sun ensured each department, with an interest or investment, participated in the project to reduce resistance and to help change old habits.
TIP: As an overall strategy, ‘the buy-in’ has both negatives and positives. On the positive side you establish constructive relationships with the wider organisation, building a ‘network mesh’ both within and outside your IT team (and I mean a people network mesh not IT equipment!). This enables IT an easier path for future changes, and creates greater understanding and acceptance. On the negative side, ‘too many cooks spoil the broth’ as the old saying goes. The more people involved in a project, the harder it is to gain consensus and the longer it takes for things like approvals. In this case, it is best to nominate a core team of people (no more than 4-5) who have the authority, background information and knowledge to make the final decisions. This doesn’t mean excluding others but it does mean the core team sets boundaries around negotiable and non-negotiable design to streamline the overall process.
Focus # 2: The Heavy Stuff
Hardware takes up space. Especially old hardware. Sun dumped almost 5000 servers, switches and storage units and replaced them with newer and more efficient models. In doing so their utility bill decreased by 61%.
TIP: This is quite simple to achieve in theory however in practice it takes quite a lot of planning. But virtualisation technology can assist – the best enterprise virtualisation tools today include the ability to take an image of a server as it exists now (called P2V – physical to virtual) and transfer it to a beefier physical server where it will usually reside with multiple other virtual servers.
Sun felt their biggest challenges lay in the design of the data center internals, including configurations, power and cooling. To overcome this, they applied a ‘pod’ concept, where each group of racks had similar requirements (power, cooling and structure). Once they defined the base load per cabinet, they could easily size for the future and essentially clone the pod. In this model the data center is split into small rooms within the larger data center space and within these pods the cabinets are physically rolled in and out as required.
Within the Santa Clara data center, Sun implemented three UPS units that operate at 97% energy efficiency and are monitored for capacity and availability through IP connections. In addition they setup the UPS racks as ½ full to ensure there is no downtime in the future when adding new UPS services.
They used the same concept with the cooling towers, allowing space for an additional two to be installed. The cooling towers themselves have variable frequency drives on the chillers and chiller pumps. This strategy allows them to reduce and increase capacity as requirements change during the data center lifecycle.
In the cooling arena Sun’s first task was to look at the need for raised floors – which would cost them around US$1m to install. After initial investigation they decided on a slab floor construction with two alternate cooling approaches:
1. Overhead spot cooling – this approach used Liebert XD technology for areas that required scalability e.g. servers for support teams that work on shorter term projects. The benefit of this system is a gain in rack space and higher “flexibility” but the downside is that it requires a more complex piping infrastructure and is less intelligent than an in-rack system (i.e. can only be either on or off).
2. Hot aisle containment – in this design racks are placed back to back to create an ‘aisle’ of hot air which is partitioned by ‘walls’ and a suspended roof (which allowed Sun to use any size racks). This system used in-row APC cooling units to cool at the source. This system is more flexible (i.e. adjustment for temperature gradients as racks move in and out of the pods) but is at the cost of floor (rack) space.
Focus # 6: She’s electric….
Intelligent power was implemented through Starline busways – like a “lighting track on steroids”. Essentially with this overhead system users can “insert” new power sources at any point without having to shut down the mains power. Like the UPS, the power source has IP connectivity to ‘forecast and monitor energy use’. And in addition to the existing requirements, Sun scoped the piping to be oversized. This added around 10-15% to the project cost however the payback will be on no disruptions as the data centre expands in the future and reduced cost to install additional cabling.
