Peribit gear boosts WAN performance

Peribit is introducing a set of equipment and features to boost site-to-site WAN performance by compressing larger chunks of data, overcoming delays caused by high-latency links and utilizing the better WAN connection when a site has an alternative.

The company's new appliance, called SM-500, caches up to 500G bytes of data that it can refer to when scanning WAN traffic. It then looks for large repetitive patterns it can send as shorthand to another SM-500 across a WAN link.

Compression devices work in pairs to seek data patterns that repeat within a datastream and replace these patterns with smaller patterns before sending the data. The receiving box replaces the smaller pattern with the original data. Peribit competitors include Expand, ITWorx, Packeteer and Riverbed Technology.

To continue reading, register here and become an Insider. You'll get free access to premium content from CIO, Computerworld, CSO, InfoWorld, and Network World. See more Insider content or sign in.

Peribit is introducing a set of equipment and features to boost site-to-site WAN performance by compressing larger chunks of data, overcoming delays caused by high-latency links and utilizing the better WAN connection when a site has an alternative.

The company's new appliance, called SM-500, caches up to 500G bytes of data that it can refer to when scanning WAN traffic. It then looks for large repetitive patterns it can send as shorthand to another SM-500 across a WAN link.

Compression devices work in pairs to seek data patterns that repeat within a datastream and replace these patterns with smaller patterns before sending the data. The receiving box replaces the smaller pattern with the original data. Peribit competitors include Expand, ITWorx, Packeteer and Riverbed Technology.

Traditional compression devices have limited memory to store these patterns and look only for relatively short patterns. Or if they recognize large chunks of data, it is as entire files, so even small changes to the files make it impossible to abbreviate them with compression.

SM-500 views streams of bits, including those that represent data within files. This lets the device recognize large patterns within files so if small changes are made, the device still can compress large chunks of the files. The company says it can store large patterns on disks, keeping recognized patterns in its library indefinitely. The new device is expected to be available in August and costs between $9,000 and $40,000, depending on the size of the WAN link connected to it, ranging from 256K to 20M bit/sec.

Electric utility Ameren Services in St. Louis is looking at two of Peribit's new features in particular: forward error correction and TCP spoofing that Peribit calls Active Flow Pipelining. Ameren hopes the Active Flow Pipelining will improve throughput on some of the firm's finicky T-1s, says Roger Luechtefeld, senior engineer in Ameren's network engineering department.

Lost packets force retransmissions, and when the link is compressed, "we're losing a larger chunk of data," he says.

Peribit calls its caching and compression capability Network Sequence Mirroring. It is meant to augment its earlier, short-pattern recognition technology called Molecular Sequence Reduction (MSR). SM-500 also supports MSR so it can interoperate with Peribit SR appliances that only perform MSR.

Peribit also is introducing an SR device called SR-100 for speeding connections between data centers. It supports 155M bit/sec OC-3 connections while the previous fastest SR was the SR-80 at 45M bit/sec. It does this by aggregating the compression engines in up to six SR-50 devices to create one compression device. SR-100 costs between $24,000 and $100,000, depending on how many SR-50s are added to it.

The company is releasing a new version of its software called SR Version 5.0 that adds the ability to monitor two connections - say a frame relay link between sites and the Internet connections to those sites. Users can set policies that describe the latency, packet loss and jitter on the connections and route applications to the appropriately performing link for each application.