Instart Logic, HTTP/2 and You


The previous blog posts in the HTTP/2 series covered the background on HTTP/2, the functionality and how it differs from HTTP/1.1; in this post we will look at how these changes will impact Instart Logic customers today and in the future.

From its inception Instart Logic took a different stance to traditional Content Delivery Networks (CDNs) and web performance optimization. Instead of doing traditional front end optimization which can be time consuming and has the potential to break applications, we focused on what optimizations could be achieved with the use of machine based learning, leading to technologies such as SmartSequence and SmartVision which powers our HTML Streaming, JavaScript Streaming and SmartVision Image Transcoding features. This is great news as many of the front end optimizations that companies implemented may actually hurt web application performance in an HTTP/2 world; while our optimizations will continue to provide web performance improvements for web applications.

Many of the optimizations provided via HTTP/2 are network level optimizations which pair nicely with our application level optimizations. Take for example our SmartVision Image Transcoding service to reduce the size of images delivered over the wire. Header compression in HTTP/2 will work in parallel with image transcoding to reduce the size of data to be transferred with a given image even further.

On the network side, the middle mile of our network uses a custom built transport mechanism, Interproxy Transport Protocol (IPTP). IPTP was built to eliminate the limitations of HTTP and speed up content delivery on the Instart Logic network. The benefits of IPTP is it doesn't matter whether the first or last mile is using HTTP, SPDY or HTTP/2 our network can handle it and no changes are needed to the middle mile to get the most efficient application delivery. The diagram below shows what communication looks like under various scenarios.

HTTP/2 Instart Logic Network Deployment High Level Map

Today we are working towards enabling our network to handle HTTP/2 client connections. In the future as HTTP/2 has greater adoption, organizations will want to start experimenting with some of the more advanced features of HTTP/2 like priorities, dependencies and push. The work we have done with machine based learning puts us in a position to help customers automate this process by identifying the dependencies, setting priorities and pushing the right content to users at the right time. Our initial implementation will be similar to our SPDY implementation with communication between the client and Instart Logic servers occurring over HTTP/2 and communication between the customer origin servers and Instart Logic taking place over HTTP/1.1.

We were one of the first in the industry to support SPDY, and have real-life deployment experience. We also uncovered some gotchas which will make it easier for customers to transition to SPDY and HTTP/2. We are working on finalizing support for HTTP/2 on our service.