How Are Data Centers Coping with Increased Internet Traffic?
For years, the data center industry has seen a phenomenon of creating excess production capacity and underutilizing it. Until recently, it was not uncommon around the world to see a facility prepared for electricity consumption by IT solutions at the level of 10 megavolt-amperes (MVA), while its actual load was only 2 MVA. Even more common was the use of server computing power at below 10 percent of capacity, and this applied not only to individual devices or IT racks, but to entire IT systems.
The community focused on energy efficiency and its optimization spent many years trying to implement IT infrastructure at customers' sites "made to measure." Some hope was offered by the strategy of modular development of IT solutions, for which—in excess (in line with plans for further expansion of the facility's computing power)—only critical physical infrastructure, such as power and cooling, had to be provided. However, in such a situation, the demand for IT equipment was often significantly overstated, which left a gap between design expectations and reality.
Power in Reserve
It became accepted that in traditional, corporate data centers, one server (or one processor in a server) is responsible for handling one specific application. Only the introduction of an additional software layer between hardware and applications—that is, virtualization—ensured more optimal use of servers. – However, the situation is still far from ideal. Large IT environments are still built with reserve capacity to meet user needs during peak load, for example on days such as Black Friday, Cyber Monday, Christmas or the post-holiday sales period for online stores – notes Bartłomiej Raab, country manager at Vertiv Poland. – To protect against the effects of a possible disaster, redundant systems are built, where a large number of them are replicated in a backup data center—just in case.
Organic Data Centers
The solution to this problem appeared only in the case of very large data centers, which can be compared to trucks with the performance of super-fast cars. – They are one hundred percent managed by software, and thanks to the unification of the hardware used for data processing and storage, the operator of such a facility has at its disposal tens of thousands of processors and storage devices connected to the network, which effectively operate as one "organism" – explains Bartłomiej Raab. Moreover, several data centers of this type can be connected to each other, so that stopping the operation of one of them (for example, as a result of a natural disaster or planned maintenance work) causes only a slight drop in performance, rather than a loss of services.
Vertiv experts point out that queries sent through the Google search engine can serve as an example. Each of them goes at the same time to three different data centers located in different parts of the world. The facility that is fastest delivers the search results. And if one of them is out of operation, there are still two left.
Similar rules apply to the networks through which queries are sent. However, the installation of fiber-optic cables must be planned with redundancy. The procedure for placing cables in the ground and on the seabed is very expensive. When planning new routes, a network provider must therefore think in a minimum ten-year perspective. When part of its infrastructure is already fully utilized, it simply activates additional, previously unused fiber-optic connections.
Unexpected Loads
Unexpected Loads
Today we are dealing with an unprecedented situation in which a huge number of people—specialists and students—perform their duties from home, participate in video conferences and share the results of their work via the internet. Telecommunications operators are noticing this as well—in a recent press release, the chief technology officer at BT stated that until now, in this largest British telecommunications network, peak traffic intensity was usually observed for a few hours on Sunday evening. Now the amount of data being transmitted is greater than ever before and practically does not decrease.
And even if connectivity with some people is a bottleneck, the entire ecosystem still operates very efficiently. Operators are also constantly monitoring line quality and expanding it when necessary—the current situation is a perfect opportunity for them to verify the flexibility of their own network's operation. Where greater bandwidth is required, it is provided within a few days. Fortunately, this is not a major problem for companies thanks to the fact that they previously invested in their data centers with a large reserve.
About Vertiv
Vertiv (NYSE: VRT) offers solutions in the areas of IT infrastructure, software, analytics, as well as a range of services that ensure uninterrupted and optimal operation of its customers' key systems and their development as their business grows.
Summary
Data centers have been designed for years with excess capacity, which was sometimes inefficient—for example, a facility prepared for 10 MVA with an actual load of 2 MVA. The solution turned out to be large, fully software-managed facilities operating as one organism. In this model, Google search queries go simultaneously to three centers in different locations, which guarantees service continuity even if one of them fails.
The current increase in traffic, related to remote work, has verified operators' flexibility. According to the chief technology officer at BT, data intensity is now greater than ever and practically does not decrease, and infrastructure reserves are proving crucial for the smooth operation of the network.
