Australian biotechnology research and manufacturing company Proteome Systems has expanded its hardware infrastructure by integrating blade servers into its protein analysis platform.
The company has a protein analysis platform called ProteomIQ which, according to CIO Warren McDonald, uses IBM RS/6000 servers with DB2 and Apache for the core data management system, BioinformatIQ. The company also has Intel servers for hosting its image analysis tool ImagepIQ and other vendors' software.
“We are installing IBM blade servers as part of the IT infrastucture of our ProteomIQ Platform, and expect to see many systems installed this year,” McDonald said. “In-house we have only installed a single BladeCenter with a few blades at this stage. This was to prove performance on our applications and determine full capacity requirements. I expect to install another eight blades in the next few months as we have several projects on the horizon that will demand significant computational resources ideal for the blade server architecture.”
The blades being used by Proteome are dual 2.0Ghz Intel CPUs with 1.5Gb of RAM and redundant 40GB IDE disks.
When asked what influenced the company’s decision to deploy blade servers over expanding their existing infrastructure or more traditional rack-mounted servers, McDonald said he was looking to overcome frequent scalability and flexibility problems.
“We chose blade servers as they give us the redundant sub-systems (like cooling and power) you get in larger SMP systems but also the flexibility to host different operating systems and environments,” he said. “The integration of many specific capabilities for Proteomics dictates that we run many different server environments, being able to easily provide redundancy and additional capacity for these in a managed server centre was our big motivation.”
McDonald would only describe the investment in the blades as “relatively modest”, however, he expects to see tangible benefits from their use.
He expects savings to come from two areas. “Firstly, lower capital costs for the same compute resources. We found that the break-even point (per GHz of processing power) in relation to buying larger conventional SMP servers was as low as six CPUs, or three blades,” he said. “Secondly, the site management of the servers is achievable with much less cost for rack space for accessories like KVM switches, network infrastructure, remote monitoring and management and ease of replacement. It is difficult to give an exact figure but is most certainly significant.”
McDonald also said the baldes support more users. “We implemented several redundant Linux nodes for our image analysis in just a few days and now have the capability to upgrade to additional nodes in just hours using standard system images,” he said.
Proteome went with IBM’s blades due to an existing global strategic alliance with the company and a history of IBM product use.
“I think the management layer and server density of the IBM product are very good. This combined with the benefits of our existing relationship, led us to IBM.”
With McDonald’s expectations being met with the blade product, he sees the server market permanently changing with the advent of this nascent technology.
“There is still a role for traditional servers, but I see a lot less of them in our future plans,” he said.