Rising use of social networks, rising energy costs, and a need to understand new technology such as virtualization and cloud computing are among the top issue IT leaders face in the ever evolving data center.
The modern data center, which contains thousands and tens of thousands of servers and storage systems, is used to manage an array difficult tasks. Search engines, social networks, cloud environments, high-performance computing clusters require a new approach to the architecture of the system and its maintenance.
Social, mobile trends impact the infrastructure
Organizations are increasingly dependent on information technology. However, a two-hour outage may impact the value of that technology when it reaches 25 million end users. The growing importance of social networks over the years poses a major challenge for the data center, in part, because half of new data centers in the future will be unstructured.
The rapidly expanding use of mobile devices plays a role as well, as both mobile and social trends span the enterprise, and affect the management of current IT infrastructure. Although, most of these trends are for the young, older businessmen also actively developing smartphone, tablet and social networking apps to work in conjunction with social media sites like Facebook, Twitter, Google Plus etc.
A majority of IT executives, according to a survey by Emulex, have said that server virtualization, cloud computing, big data, and the convergence of storage and data networks, are key factors in adding unparalleled demand for network bandwidth in their datacenters.
“Social networks have become a gold mine for users’ preferences, interactions, and attitudes,” said Nick Kirsch, Senior Director of Product Management, EMC Isilon in a response to development of tools that can unlock the power behind social networks consumer data.
“Give a user the ability to communicate freely with those that they trust and they will share dramatic amounts of personal information. That information can then be mined and leveraged to provide goods and services that uniquely appeal to that individual and their networks. The explosive nature of social networks and tremendous amount of data available has focused infrastructure tools on scale. The more data that can be captured and adequately accessed, the more information is available to leverage. Everything is now being designed and built for tremendous scale.”
Social Networks, ready for action
Social networking impacts data center infrastructure with vast message storms. Using Twitter and Facebook as examples, if someone has posted a popular tweet or message, there is a burst of traffic that follows, consisting of tweets, re-tweets and likes on that topic. In addition to the bursts of traffic, the infrastructure should be able to handle a large number of concurrent users.
The success of modern data center is also dependent on the quality of services it delivers, which means superior system response to user requests and constant availability to all social users.
To achieve this, data center administrators need to optimize their infrastructure and fault tolerance as well as performance. This can be done by facilitating result caching, scaling and availability, database distribution across many servers, and data replication across caching and database servers. All this must be replicated again, across geographically distributed data centers, with service fail-over in the event of a server or data center outage.
“It’s all about scale – social networks have shown two important trends: they are explosive and they are fickle,” Kirsch says. “The architecture has to be designed to scale rapidly from day one, with the ability to increase performance and capacity on-demand with little to no additional operational overhead.
“As a social network gains in popularity, it reaches a tipping point where it begins to grow extremely rapidly. That growth can be maintained (until saturation) as long as the infrastructure continues to scale. The majority of failed social networks have done so because of infrastructure limitations that has led to poor user experiences. There is no opportunity to re-design the network once scale has occurred – it has to be designed in from the beginning,” said Kirsch.
Scaling the data center for efficiency
Data centers can implement a particular class of services using differentiated service protocols. A protocol uses policies to decide when to forward one type of packet before another. So, policies would determine when critical social applications receive bandwidth over other social networking applications. A particular class of service groups and certain types of traffic for a data center should be prioritized into classes.
Green data centers are another hot topic to power data center usage. As per a recent report from Pike Research Center, the global “green data centers” market will climb from $17.1 billion in 2012 to $45.4 billion by 2016 at a compound annual growth rate of about 28 percent.
Higher energy efficiency, sharp rise in information technology costs and a constant increase in the sectors carbon emissions will result in more modern governments and companies depending on green data centers. A recent NY Times article published over datacenter power usage argued IT as one massive power hungry source, but the article does not paint a very objective picture.
How social network infrastructure can go green, cut costs
Many IT firms are making strides to reduce their energy consumption. Facebook, for instance, to reduce their carbon foot print, has relocated their data centers to cooler climates. The company is also designing a new data center specially to store huge repositories of photos and videos in a way that’s accessible, power efficient and convenient, but also cost-effective.
The lowest possible amount power consumption is achieved by the unique design of the power distribution at the rack, which drastically simplifies the inherent problem of the lack of a major data center input power. Virtualization is another game changer that can help make significant changes in the way we store data, handle it, harness it, and leverage it for all the perks mentioned above.
Then there’s scale-out NAS architecture, which can boost the performance alongside capacity. The architecture can leverage an Isilon array, forn instance, so if one Isilon array runs out of capacity, you can to plug additional Isilon nodes to increase capacity and performance of data center at the same time. EMC is one such vendor that is creating scale-out NAS based unstructured data tools and services to increase capacity and performance of data center.
“Scale-out NAS is the ideal solution for social network companies that pile more and more data into their data centers – once one Isilon array fills up with user data, or more performance is required for faster data access, the IT admin can simply plug another node in. Isilon OneFS (the operating system) then intelligently distributes data across the new disk drives and nodes, so the IT admin doesn’t have to do any manual reconfiguration. The impact to the user is minimized due to an easy install and intelligent OS,” Kirsch explains.