So the purpose of the WAN optimizer is to quickly or efficiently transfer some TCP data over "long, fat networks" (LFNs aka elephants). The specific way they operate depends on the vendor implementation, but includes the optimization methods Bilal discussed. Their TCP stacks are tuned specifically for this scenario. Now between the optimizers we still have the problem of latency, but they are programmed to handle this. this allows the optimizer to quickly send ACKs and allow the conversation to happen quickly (the client and server get the ACKs at LAN -like speed rather than having to wait). An optimizer is on each side of the connection close to the client and server.For the client, the local optimizer pretends to be the server (and vice versa). WAN optimizers work by "proxying" the connection. That is just takes that long to get the signal from point A to B. After each window of data, it takes half a second to get the ACK.
Doesn't sound like much but it becomes the dominant factor determining throughput on the link. 'The technology can be deployed as a complete SD-WAN service stack that provides our customers with the best user experience.' 'With Riverbed's SteelConnect and Microsoft's Azure Virtual WAN, we are able to deploy locations in minutes easily in a few steps, control the access, and remove complex and expensive scenarios.
The one-way latency can be 250ms or more. But consider a long-distance link like a satellite hop. ON a LAN, this typically happens quickly. The sender waits for an acknowledgment after sending a certain number of bytes, called the "window". It requires the destination to "acknowledge" packets so the sender knows what has been received. Remember that TCP is a "connection-oriented" protocol.
0 kommentar(er)
