Overview

In a stringently regulated market our clients are required to action perform more calculations than ever before. The increased demand on the compute infrastructure during peak periods leaves the business with a stark choice: suffer reduced performance or invest in hardware that will stand idle for long periods.

With the help of Microsoft Azure, we were able to offer our client with a third option: deploy compute power for short bursts when demand is high. This gives you the power of a larger data-centre with costs directly linked to usage. No unutilized compute hours, no maintenance fees - pure on-demand power.

Solution

We chose to partner with Microsoft to provide a solution for the client. It was decided to use a combination of Microsoft Azure and Microsoft HPC Server 2012.

Our implementation used both on premise and cloud resources, which allowed us to take advantage of larger compute power when we needed it without having to pay for unused hardware. We combined this with Logscape monitoring, which automatically provisioned Azure instances when the workload required it and removed them once workloads subsided.

Benefit

Provisioning Grid machines on the cloud has a number of clear advantages for our client:
- Scalability - Client’s ability to respond to variable workload patterns is not only improved, but it is also automated, making it less prone to failure.
- Cost Savings - Azure provides a cost model based on minutes used. Combining this with Logscape and the Excelian HPC Automation means that we can keep costs to an absolute minimum, whilst guaranteeing sufficient resources to meet the changing demand.
- Security - Azure is much safer than a traditional in-house solution. The ability to target different Azure data centres significantly reduces the Operational risk faced by such an important piece of infrastructure – whilst Azure instances have multiple redundancies built in to reduce the risk of single component failure.
- Integration - HPC Server is incredibly simple to set up and use.

Full case study (175 kB)

Download