OpEd Passing shot: Optimising DCs
Written by Wil Cunningham
Wil Cunningham, environments manager at Lloyds Banking Group who is currently working on contract on the integration with HBOS, discusses the best ways to sweat existing systems when budgets are tight, with particular reference to his previous work optimising a data centre at a well known High Street bank based in Edinburgh
In the aftermath of the economic downturn, IT budgets have never been so constrained, yet demand for space and power in data centres is ever increasing. So with less money to continue to fund data centre expansion and buy ourselves out of trouble, what can we as IT professionals do? Get the most out of your existing systems is my advice and I'd like to share some of the things I learnt last year when working on a large Data Centre Optimisation Programme at a major UK retail bank in Edinburgh. I hope the procedures we followed may be of help to you.
Step1: Challenge policies and standards Some historic policies were formed due to previous technology constraints at the bank or what was considered 'best practice' at the time, but these needed to be challenged. I continually looked at why we were co-locating test and production systems; why were we allowing critical and non-critical systems to be located within the same data centres; and why were we providing Disaster Recovery (DR) services for minor systems? The result of this review was senior stakeholder support to change the policies at the bank and create a new 'placement strategy'. By mandating that test beds, production support beds and production should never share the most critical data centre space, I was able to optimise IT spend and efficiency.
I challenged the incumbent standards too, reducing the numbers of backup copies taken, removing the need for minor systems to have DR, and stopping new systems being given six years growth capacity, as it was no longer appropriate in today's 'growth constrained world'. We were then able to re-use the freed up storage rather than annually buying more. Power capping limits on idling servers also reduced power wastage.
Step 2: Premises The Programme had already created a dedicated test centre from a mothballed, then re-commissioned, data centre. This allowed the transfer of all test and production support environments from our critical data centres to this new facility, allowing re-usable production space to be freed up in the main facility and extending its lifespan.
Step 3: The technologies The Programme utilised virtualisation at ratios of 20:1 to 40:1. Vendors helped as they had tools that scanned our estate both for low hanging fruit, but more importantly to check out and assess application compatibility with the new virtualisation technologies that raised server utilisation rates. For other midrange platforms we employed consolidation that provided a great way of reducing power and space and allowed a restacking of systems against their individual criticalities. I knew not to stack critical systems with non-critical systems as by default we would be giving MI systems the same level of DR, resilience and support as payments systems, which is not cost effective.
Step 4: People Individuals on each platform were accountable for their part in the data centre optimisation. This covered decommissioning and potential re-use (part of new installations), new technology adoptions, like virtualisation, which benefited the data centre offering, and general cost reduction initiatives. The team then created inter platform competition by actively supporting and funding the best ideas in respect of power and space efficiency, thereby creating healthy competition.
We created a central log of the optimisation opportunities and we then generically sized them against weighted criteria, covering cost/benefit/risk/ and do-ability (i.e. could the platform easily implement and support the change). We kept it simple with a process that was like a cut down request for information (RFI). The platform reps and DC staff felt empowered to challenge each other, however, the weighted scores showed the most efficient opportunities and we went with them.
All vendors were invited to support and get behind our goals. We used our project partner, Morse, as an agile and flexible support to assist in creating 'downward pressure'. Vendors were naturally keen to sell us new expensive technology but by driving them through the same internal process as the platform teams, we knew we were funding the right initiatives.
Step 5: Funding I was 'vendor agnostic' which encouraged them to bid for each others' business, thus generating savings in terms of 'lost leaders' and bundling packages (almost the equivalent of the classic buy one, get one free offer). We introduced a capital expenditure board that brought these funding elements together and made sure we got the best deals.
Step 6: Processes I mandated decommissioning as part of the project lifecycle so that all funding estimates had to contain a percentage of the new infrastructure spend as a decommission cost. By 'capturing' decommissioning estimates the Data Centre Management (DCM) teams could get the earliest view (along with platform forecast spend and usual capacity plans) of future demand and plan accordingly. Allied to the use of meter readings to assess efficiency this let us optimise space and power. Finally, I developed the mantra for DCM staff that 'nothing goes in, unless something comes out'.
Each efficiency drive is different but I hope the steps I've outlined above are of some use if you are presently sweating your assets.
• If you would like to discuss the above project and optimising DC subject matter please contact Wil at email@example.com