Product Documentation

TCP Flow-Control Acceleration

May 23, 2013

Ordinary WANs have very poor responsiveness at high link utilization and at long distances. A widely used rule of thumb for ordinary, non-accelerated WAN links is, "once link utilization reaches 40%, it is time to add more bandwidth, because performance and reliability have degraded to the point where the link is largely unusable." Interactive performance suffers, making it hard for people to get work done, and connections frequently time out. Accelerated links do not have this problem. A link with 95% utilization is still perfectly usable.

CloudBridge appliances become virtual gateways that control the TCP traffic on the WAN link. Ordinary TCP is controlled on a per-connection basis by the endpoint devices. Optimal control of link traffic is difficult, because neither the endpoint devices nor individual connections have any knowledge of the link speed or the amount of competing traffic. A gateway, on the other hand, is in an ideal position to monitor and control link traffic. Ordinary gateways squander this opportunity because they cannot supply the flow control that TCP lacks. CloudBridge technology adds the intelligence that is missing in the network equipment and the TCP connections alike. The result is greatly improved WAN performance, even under harsh conditions such as high loss or extreme distance.

CloudBridge flow control is lossless and transparent, and it implements a broad spectrum of speed optimizations. No configuration is required, because of autodiscovery and autoconfiguration. You might, however, have to tweak your firewalls if they block the TCP options used by the acceleration algorithms.