Intelligent Infrastructure

Mass data and next-gen workloads Intelligent Infrastructure

Cloud Data Centers of the Future – Part 2

Icons_Teal_SERVER

Welcome back, this post is part two in a series of two posts about data centers of the future. Click here to read part 1.

Location, location, location.  Like in real estate, building your mega data center at the right spot will become more and more important.  Data centers will be built in cooler climates, which will reduce cooling cost because outside air can be used to chill those data centers.  And while data latency and qualified personnel availability issues will not allow data centers to migrate to the North Pole or to Greenland, climate considerations will become even more important for the mega-data centers of the future.  Also, for the DC to operate cost-efficiently, access to affordable and reliable power will be required.  But this could be of lesser concern as more companies will follow Apple’s “green” example: Apple’s Maiden, North Carolina data center is now 100% supplied by renewable energy, with 42 million kWh coming from an on-site solar array, and the rest from bio-gas stored in nearby fuel cells.  Rural areas with affordable land and room for expansion will be favored.  Financial incentives, available water supplies, and proximity to customers, major cities, and airports –will all be important considerations.

As expected, cloud data centers will continue relying on low-cost hardware and will utilize architectures designed to tolerate failures without loss of customer data or service interruptions (this is what the “cloud architecture” is all about).  Focus on cost reduction will be absolute. Thus, power saving solutions such as low-power processors will gain popularity, especially, in less performance-centric storage-oriented applications.  Commodity components will be scrutinized for quality and reliability in the persistent effort to find the right trade-offs between lower acquisition costs and lower maintenance and service interruption costs.  Overall, the cost of computing, networking and storage in the data center of the future will relentlessly continue to fall.

It is difficult to predict how quickly, if at all, the adoption of the disaggregated rack-scale server architecture will happen. This approach, promoted by Intel, among others, could appeal to the future designers of mega data centers.  The rack-scale server architecture – if implemented correctly – will enable independent provisioning of compute, network and storage subsystems in the data center and holds the promise of high cost-efficiency.

Migration from replication-based data redundancy schemas to large-scale erasure-coding solutions will gain popularity. Faster network fabric and powerful processors will reduce the performance penalty of erasure-coding making its reduced data storage overhead irresistible. This is especially true for “colder” data storage applications.

Open-source software, such as OpenStack, will dominate.  Associated cost savings will be tremendous, which will be recognized by most DC operators.  The issues of software maturity and support will be addressed in the near future much in the same manner that Linux maturity and support is addressed today.

Exotic architectures like immersive oil cooling will gain popularity but mostly in high-performance computing applications.  It is likely that all the mega-sized storage silos will still rely on air-cooling.

Software-defined networking with “dumb hardware” and smart software will be displacing today’s “all inclusive” expensive networking gear. Much like the battle has moved to the software layer with smart-phones today, we can expect networking to witness the same paradigm shift.

Designers will experiment with new architectures such as low power 64-Bit RISC based CPU’s for compute.  Potentially, new storage architectures such as Seagate’s “Kinetic” product line that fundamentally reduces the number of components required in large-scale storage systems will garner a high level of interest among those looking to optimize systems at scale.

Finally, advanced, complete, and accurate design tools will be available for quick and easy prototyping and testing of the entire data center or storage cloud.  In fact, we are close to completing one implementation of these future design tools right here, at Seagate, in the Cloud Modeling and Data Analytics organization.  If you are interested in more details please contact us!

The future of cloud data centers and cloud storage/computing is exciting!  Many more improvements and technological trends beyond the ones mentioned above will be implemented.  It is a privilege to be a part this technological revolution.  As is usually the case with forward looking analysis, an inherently low return on investment proposition, only time will show how many of these trends mature in the next 5-10 years.

Author: Andrei Khurshudov