Virtualisation is a major trend in modern datacenters as it allows to optimize the server consolidation wave that is taking place today. Also at Belgacom. This article goes into why and how we did it and what the results are.
Belgacom is not only a service provider to other companies, but is also having their own IT department and infrastructure for their own use. Belgacom runs over 1000 servers with a Linux OS and over 1800 servers with a Windows server OS. The rapid growth, amongst others through the acquisitions of Telindus and Proximus, made that the number of servers outgrow what is easily manageable.
Our first virtualisation experiments started in 2004, whereas virtual servers are the default standard since 2008. The standard virtualisation platform is VMWare’s ESX software. Clusters are split over two physically separate datacenters and are managed by VMware vCenter Server and VMware Site Recovery Manager for disaster Recovery. Different clusters have been created to correspond to logical zones : DMZ, internal, virtual desktop, …
The VM installed servers grew 62%, year over year from 2008 to 2009.
The virtualization efforts proved to be successful, since we were able to reduce the energy consumption of our own datacenter utilisation, reduce the investments in new hardware and software and improve the service towards our internal customers with and improved flexibility and shortened time to setup a new server.
The Total Cost of Ownership (TCO) of our servers has been reduced with 85% (on a 4 year period), with the following split. Our hardware cost has been reduced with 80%, due to less server hardware. The workload of our servers has been improved. Systems used to have a 10% workload before the virtualisation happened. Today, we have a ratio of 30 instances per server. Together this brings a 90% electricity consumption reduction and saves 3.4 MWh/year. For 300 virtual servers, we have a datacenter occupancy decrease from 600 racks to 20, and a cabling reduction of 1200 to 100 cables. The manpower to maintain all these servers has been reduced by 90% as well. On the other hand, we had an increase in storage needs of 30% per virtual machine.
Beware : the calculation of the TCO reduction has to take many more parameters into account. It is always very hard to calculate it exactly, since the parameters may be different from case to case. You cannot, for example, always combine 30 instances per server. Depending on the physical size of the server cabinets, the rack space savings are completely different. We therefore always suggest to analyze the potential savings via a consulting phase prior to the execution of the project. Careful planning and dimensioning has to be done in the design phase. A precise transition plan is also needed to ensure the continuity of the servers and the services.
The conclusion is simple : we would definitely do it again, as the savings and the improved user satisfaction surpass the additional cost of software and storage, the additional level of complexity and the previous complexity to manage a large IT datacenter.
Want to know more ?
Visit Belgacom’s solution pages or contact your account manager.
Most Comments