China looks at ‘throwaway’ modular datacenters

China looks at ‘throwaway’ modular datacenters

Technology News |
By eeNews Europe

To construct mega-datacenters (MDCs), the modular approach uses thousands of servers, set up and wired, into one shipping container, and then connects the containers together by viewing them as large pluggable building blocks. Once connected to power, cooling, and the Internet, the MDC can provide cloud services at any location in the world. Typically, a standard 40 foot shipping container is equipped with 1200-2500 servers, and these systems are already being supplied by companies such as IBM and HP for datacenters run by Google and Amazon.
The researchers at the National Key Laboratory of Parallel and Distributed Processing and the National University of Defense Technology in Changsha, Hunan see a new architecture with containers of sealed servers connected together to provide the cloud services. If a server fails, it is merely cut out of the network, until the sealed container is no longer needed and is replaced.
Their SCautz architecture uses low cost commercial COTS switches, with the intelligence in the servers, to provide high network throughput for various traffic patterns, and achieves graceful performance degradation when failures occur and increase.
Containerization lowers the total cost of ownership for cloud providers, and allows operators to manage the MDC using a particular "service-free" model. A container as a whole is never repaired during its deployment,  typically three to five years and as long as the performance of the entire container meets an engineered minimum criterion, there is no need to replace the servers.
In architectures such as Microsoft’s Bcube, server failures not only decrease the computation and storage capacity, but also destroy the structure of the datacenter. Switch failures decrease BCube’s performance significantly, with the average throughput falling by more than 50% when only 20% of the switches fail.
The SCautz architecture is a hybrid structure for dealing with server and switch faults, built by interconnecting the servers’ NIC ports with a small number of redundant COTS switches. It can run with the switches off, providing high capacity for processing different types of traffic patterns, and behaves as well as BCube. The performance doubles when the switches are used, providing the ability to handle bursts of network flows effectively without lowering the quality of bandwidth-intensive applications.
The switches also have redundancy so that when a server fails, the SCautz network finds a peer server in the same cluster to bypass the failed one. This extra cost of the redundant design is very low, as analysis shows that a typical container with 1280 servers only needs 160 COTS switches.
The next stage is for the architecture to be implemented and examined in a larger production datacenter say the researchers.
National University of Defense Technology

If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :    eeNews on Google News


Linked Articles