Deploying an application automatically to over 1000 blades with 3000 to 5000 virtual machines is a big deal, particularly when a failed system can stop millions in daily revenue as is the case with large financial services company.
When you are an architect within a financial services company and responsible for such a massive infrastructure like this, your technology architecture decisions also cost millions of dollars.
VMware had the opportunity to open the “private diary” of a cloud architect responsible for such a system at one of the most well-known global financial services brands in the world. In our dialogue, we were able to better understand why and how they use VMware’s vCloud Suite and vFabric Application Management in their data center.
Here are the highlights:
Starting with the Problem – Spikes in Demand
Demanding system up-time is one thing, but keeping five nines (99.999%) of uptime while supporting major spikes in revenue-related demand is another. In financial services, the volume of transactions can increase drastically during holidays and global events. For example, the volume of trades for a single stock could go up 400% in just one day. For this financial services company, the peaks were the really expensive part. Though peaking weeks in a year were few and far between, the business had to have the hardware infrastructure available immediately in the event of a spike. According to the architect:
“It used to take months to requisition a new server and get it up and running. So, many times the business unit would request equipment and services six months in advance just to ensure it was ready. Then, the hardware would idle 90% of the time consuming time and money until it was needed.”
Obviously, there is a lot of wasted capital in this model – and a large infrastructure like this is not cheap.
|>> For more information, check out the websites for vCloud Suite and vFabric Application Management and vFabric Application Director
The High-Level Solution Requirements
The architect set out to reduce costs two key ways:
- By virtualizing hardware to maximize the value received
- By providing fully automated, self-service deployment that supports elastic scale
In focusing on these goals, the IT team was also able to build a more responsive and agile process for the business. As well, a fully-automated, self-service PaaS (platform as a service) model allows IT to manage chargebacks and services levels internally. Here are the top-level requirements shared by this architect:
- Self-service, fully-automated application provisioning
- Application configuration and catalog management
- Performance monitoring that was automatically built in and initiated in each application deployment
- Consumption monitoring and reporting
“It was a simple concept, we needed a tool to give us automation and monitoring as well as to track who was using what services and hardware. However, finding a single product to do it all was difficult.”
As the company looked across the cloud vendor landscape they chose VMware. The architect cited that the vFabric Application Management Suite 1) was the only solution close to “out of the box,” 2) had the highest score in their evaluation, and 3) was supported by a team who wanted to truly partner with them.
How vFabric Application Management Suite Fits
vFabric Application Management Suite will eventually be deployed on over 1000 blades (with 3000 to 5000 VMs) and support the key, mission-critical financial transaction application we’ve described as well as other important business applications.
“We expect to see a 30-50% annual CapEx savings by better utilizing what we already own and operate.”
Here are the benefits they are already seeing:
- Infrastructure and support costs going down with better hardware utilization
- Dynamic scaling for peak demand
- IT running more like a business with the ability to show infrastructure usage and value
- Applications and updates getting to market faster
>> For more on this Cloud Diary entry and others check out the Cloud Management Diaries