Thursday, August 14, 2014
What the industry can learn from Facebook''s successes and stumbles
Facebook, like fellow tech giants Google and Amazon, has a strong preference for developing its own in-house solutions. As one of the world's largest networks, Facebook's constant push to maintain high availability for 1.32 billion users is a unique testing ground for the rest of the industry. Few data centers, save government supercomputers, deal with the logistical issues as large as the Web's leading social network. We can look to Facebook's latest victories and trials for hints about which direction the industry at large is headed.
Energy innovation leaders
Developers of cluster management software can look toward Facebook for inspiration. Data Center Knowledge reports that the company has had great success implementing its latest cluster management tool, AutoScale. The software monitors workload balance on Facebook servers and throttles request volumes to maximize server efficiency.
Traditional workload monitors distribute requests so that all servers accept requests. During low activity periods, servers run on low utilization and consume nearly as much energy as they do when running at peak activity levels. AutoScale ensures that servers only run at medium or high utilization, leaving unused servers idle. As a result Facebook has seen a significant increase in their total energy efficiency. The implementation of AutoScale has cut energy consumption across Facebook's multitude of clusters by 10 to 15 percent, according to Data Center Knowledge. The social network's success with AutoScale will likely inspire other data centers to approach an "all-or-nothing" CPU utilization strategy.
Familiar Achilles' heel
Despite it's reputation for innovation, Facebook is still susceptible to the challenges that impact massive data centers across the industry. Motley Fool notes that the social network has suffered two embarrassing network failures this summer. Though the two network crashes lasted a total of 19 minutes, both apparently affected each of Facebook's 1.32 billion users. This level of error is shocking from Facebook, especially considering the company has successfully limited problems on the user end in previous instances of network trouble.
The incidents remind data center managers that they can never be too careful when securing their own networks. Many companies have utilized hybrid storage solutions to provide more dynamic access to their data and eliminate the risks incurred when network infrastructure includes a single point of failure. Thankfully, the popularity of fiber-to-Ethernet has made it easy for facilities to pursue direct links to private cloud servers.
Perle has an extensive range of Managed and Unmanaged Fiber Media Converters to extended copper-based Ethernet equipment over a fiber optic link, multimode to multimode and multimode to single mode fiber up to 160 km.