Building Instart's Web App Streaming Infrastructure

Victor Gartvich

My name is Victor Gartvich and I am Head of Technical Operations at Instart Logic. Over the past year I have scaled out our network to dozens of data centers around the globe. What excited me the most about coming to work here was the chance to develop what I call a “diamond” technology - a valuable crown jewel that is truly different than anything else on the market.

I met with the founders in 2011 to hear about the product and it seemed like a natural extension of my professional experience in building large-scale systems and infrastructure. This is the second time I have scaled out a globally-available Software-as-a-Service offering to meet enterprise SLAs and achieve greater than 99.99% availability to customers. I built the CDN (Content Delivery Network) and operations infrastructure at Cotendo, a web acceleration provider acquired by Akamai last year.

Now, at Instart Logic, I am the one responsible for making sure our service meets the expectations of our growing list of customers, including some very large Fortune 500 companies. After spending a lot of time in the CDN world, it’s exciting to be working for a company that takes web and mobile application delivery to an entirely different level.

It is a big job. Building out a high-availability global network is very different today than it was a decade ago. Back then, in the early days of CDNs, the Internet was a far more fragile entity. The global backbone was weak. The fiber optic networks that formed that backbone were far less reliable. There were many more bottlenecks in the core fabric of the Internet. For this reason, the first CDNs built out giant networks with hundreds of Points-of-Presence (PoPs). The CDN providers sought to place cached content in the same data center as virtually any mid-size or large Internet Service Provider in order to ensure that these static digital assets could be delivered quickly to end users even if problems arose in the global or regional Internet backbone.

Today, the Internet is much stronger. Backbone routes have higher capacity, lower latency, and rarely break. When they do, you can easily switch to any number of redundant networks within seconds in order to failover gracefully and maintain performance, SLAs and availability. You don’t need hundreds of PoPs any more. In fact, a network of a few dozen PoPs that are strategically located at key peering points in the Internet backbone generates a mesh that delivers true enterprise-grade coverage and availability that is barely distinguishable from networks with hundreds of PoPs. For this reason, we were able to scale out Instart Logic’s infrastructure in a year to provide robust, redundant, global coverage.

For redundancy, we didn’t stop at the network and overlapping PoPs. At every level of our application tier, we have baked in redundancy, high-availability and graceful failover. Server, cabinet, power supply, router and switch, data center, database, fiber - at every level we have taken great pains to eliminate any single points of failure.

A lot of our network is built using a foundation of Open Source Software products. We are firm believers in the transparency of OSS and will be contributing code to various OSS projects in due time.

Thus far, we have maintained 100% uptime and not have suffered a single outage that impacted customers. We have done that with millions of users accessing Web apps through the Instart Logic network. We understand that our customers are trusting us not only with their revenue machines and user experiences but also with their brands and reputations. I am confident Instart Logic’s network and service will not only meet but exceed their expectations. You can read more about my thoughts on operations at my personal blog and follow my posts here for more insights into how we are building and operating Instart Logic’s network.