Intel algorithm tackles networking bottlenecks

Intel is offering a free algorithm to speed up network transfers and reduce the load on CPU resources

Intel has launched a new free algorithm designed to allow faster network transfers of information and encourage wider adoption of the iSCSI storage networking protocol.

Kevin Kahn, director of Intel's Communication Technology Lab, said the new Slicing by Eight (SB8) algorithm offers three times the performance of current cyclic redundancy checks (CRCs), which ensure data integrity during network transfers but can be a network bottleneck, particularly for organisations using iSCSI.

According to Intel’s tests, the new algorithm also frees CPU resources taking up 2.2 cycles/byte compared to 7/byte for traditional CRCs.

Kahn said the algorithm would help servers to keep up with the increasing speed of network connections; encourage greater take up of iSCSI; and prove useful to any IT manager undertaking server implementations.

"There is pressure to make server platforms better consumers of network traffic. Though processors always used to work faster than the networks they were attached to that is currently being reversed," Kahn added. "[IT managers] need to think of how their server systems handle networking in new ways."

The algorithm is freely available here.