Microsoft cuts repetition

Microsoft Research India has developed a kind of compression technology which eliminates redundancy and operates like a host service in enterprise information systems avoiding the use of accelerators over a Wide Area Network.

Coconet stands for Content Compression in Networks. The developers of this system studied access network links of eleven corporate locations for a few weeks besides watching the access link of the Wisconsin University, as it students were collaborating in the said project.

The Microsoft developers calculated that around 75% of the bandwidth was saved by using devices meant for eliminating redundancy on the Wide Area Network (WAN); this was done by splicing repeated byte-strings from the client’s traffic, as explained by Ramachandran Ramjee, the project leader of Coconet via a telephonic interview.

Ramjee explained that the Microsoft developers describe this as intra-users redundancy which simply means that the same user got varied versions of the same files from the server or ended up going to the same Website for updates repetitively.  Ramjee further explained that the emergence of this pattern allowed them to use the redundancy elimination function of the developed software which is run on the end hosts of the network, this eliminated the need of placing expensive accelerators or use middle boxes on the WAN.

Ramjee said that many multinationals have their offices around the world; this technology will help them amalgamate their Information Technology resources thereby saving regulatory costs. With this technology these multinationals do not have place servers on each location, they can centralize their data centers to a handful of locations, Ramjee explained.

According to Ramjee the initially inexpensive LAN traffic between the user and the server has now burgeoned to the more expensive WAN traffic. Further, this has lead to the increase in the need for products like the Wide Accelerator devices which are installed to shed the redundancy in the traffic on the network.

The processing will be done at the server therefore the software evolved by the Microsoft Researchers will be asymmetric in design. The process is very smooth and does not interfere with network’s encryption and the redundancy is removed before the data is encrypted on the server.

This software is run on regular servers where a cache of 10 MB per client is required, at both ends, that of the server and the client. The best part about this technology is that it can be used even on smart phones which have limited resources compared to a computer, clarified Ramjee.

Redundancies as small as 32 bytes packets can be detected and dealt with at the host service, disclosed Ramjee. The intelligent system matches the data requested and previously sent data, if there is any overlap the program points out the commonalities and informs the user that it has already been sent.

The technology needs to be developed still further as it does not deal with inter-user redundancy which develops when many users are accessing the same information. These redundancies are then handles by the BranchCache feature of Window 7 and Wide Area Network Accelerator devices and other technologies, informed Ramjee.

The Microsoft technology can work around the restriction of the TCP which breaks down packets before sending ensuring the success and further use of this technology.