Datacenter Architecture For The Next Decade

“Datacenter Architecture For The Next Decade” with Howard Ting, Senior VP of Marketing at Nutanix, Inc.

By Pushpa Chandrashekaraiah

March 2015 Event

This was a very different topic from most of the SVPMA monthly events, in terms of the topic itself and depth of technical details covered in this session.  Howard Ting talked about how new data center businesses, with a total market capitalization of a trillion dollars, have emerged in the last decade.

With VMware’s virtualization revolution, everything including storage and networking are virtualized and are running on commodity hardware.

Earlier, organizations were buying storage based on their estimation for 5 years or so, but these days, fractional consumption of ‘pay-as-we-grow’ is becoming more popular.  Consequently, the storage business landscape is changing drastically.

Furthermore, modern business requirements that drive modern infrastructure requirements demand much higher scalability and reliability.  For example, Salesforce.com will go down for maintenance but web giants like Google and Facebook cannot afford to do so.  Therefore, they built their data centers to scale and be flexible – they fail in place and can be upgraded in place without huge changes to infrastructure.  To some extent, the type of business drives this difference in datacenter architectures. Salesforce.com or financial institutions cannot afford transaction failures.  On the other hand, if social media sites’ transactions fail once in a while or their data becomes stale on occasion, it does not significantly affect their business.  Companies in the former category still use legacy infrastructures and maintain their own data centers.

Also, centralized storage quickly runs into scalability problems, so distributed storage systems have becoming a norm for quite sometime now.

Another important thing to keep in mind while building data centers is that it’s very labor intensive. Traditional enterprises have one administrator to manage 400 to 500 servers.  By contrast, at Facebook one administrator can manage up to 12,000 servers. This is again because Facebook’s infrastructure is designed to scale easily and fail in place.

When it comes to vendors that are selling storage, the pay-as-we-go model has lower margins because of limited services requirements.  In the future, companies like Nutanix are predicting that 1) pay-as-we-go will become even more mainstream 2) there will be more and more data centers built across the globe because of cheap real estate and power and 3) there might be a shift in datacenter architects’ skills, expanding to include overall data storage, data processing and data analysis, all under the same umbrella.

Pushpa Chandrashekaraiah is a Principal Product Manager at RSA.  With 14 years of PM and engineering experience, she has successfully launched several enterprise products in multiple industries/verticals.  Her interest is in big data enterprise market and security.  She can be contacted at: pushpa_nc@yahoo.com, http://www.linkedin.com/in/pushpa.

#Technology #Tools #Design #Development #Market_Analysis #Programs #Initiatives #Strategy #Planning