Hi, I’m Thomas. Welcome to the lightboard session about application optimization. Application optimization is part of the SD-WAN offering of Open Systems. Let’s take a look at where it fits into the SD-WAN stack.

At the bottom, we have the **plate**: the lines that connect all your locations. On top of that is the **muffin**: path selection. With path selection, you can choose which of your lines should be used for which application, ensuring that the best available line is utilized for your business-critical applications.

Next, we have the **icing**: bandwidth control. Bandwidth control ensures that even if your lines become congested due to applications using more than the available bandwidth, your business-critical applications are protected and receive the necessary bandwidth.

At the top, we have the **cherry**: application optimization. With application optimization, we can reduce the bandwidth requirements for your applications, overcome the effects of high latency or packet loss, and optimize the performance of your applications.

What is Application Optimization?

The application optimization service is designed to improve the quality and performance of your traffic within your network, connecting different locations. The methods we use to optimize your traffic include compression, caching, block-level deduplication, TCP optimization, and protocol-specific optimization. The goal of these methods is to either reduce the bandwidth consumption of your applications or improve performance in cases where your lines have high latency or suffer from packet loss.

Compression

Application optimization will reduce the bandwidth needed in your network by applying compression. Whenever a file is fetched from a server, we apply lossless compression algorithms to reduce the file size, lowering the amount of data that must be transferred across your network. On the remote edge device, we decompress the file and deliver the original version to the client.

Deduplication

Block-level deduplication in application optimization can reduce the bandwidth required between your locations. When a client downloads a file from a server, we cache the data transferred between two edge devices. If a second client requests the same file, instead of transferring the entire file again, we transfer a block reference to the data in the cache. The client then uses this reference to look up the requested file in the cache, which can then be sent.

This process works regardless of how the clients access the file. For instance, the file might initially be transferred via email, and the second request might be through HTTP, FTP, or Samba. The cache stores byte sequences rather than the files themselves, so whenever the same byte sequence appears, it is replaced by a reference to its stored location. If the file is updated on the server, only the changed parts are transferred, ensuring the client receives an up-to-date version, with changes also stored in the cache. This drastically reduces the amount of data that must be transferred between edge devices.

TCP Optimization

Application optimization is also based on TCP optimization. When a client establishes a connection to a server, we intercept the session and establish a new session between the Open Systems edge devices. The edge device then connects to the server. By controlling the TCP connection between the two edge devices, we can use a TCP algorithm optimized for performance, even on lines with packet loss or high latency. Optimizing at the TCP level allows us to improve the performance of encrypted applications, where other optimization methods might not be available.

Protocol-Specific Optimization

For applications based on HTTP or CIFS, we apply protocol-specific optimization. HTTP requests typically include large headers, which tend to be similar across different requests. We replace these large headers with a smaller representation. The CIFS protocol is known for being chatty, so we accelerate it by reducing the back-and-forth communication using read-ahead and write-behind techniques. This approach optimizes traffic by reducing the required bandwidth for HTTP traffic or mitigating latency effects for the CIFS protocol.