The Blog

Dec 15, 2008

StrikeTracker Reviewed by Dan Rayburn on 

by Maxim Porges @ 12:08 PM | Link | Feedback (0)

If you are in the streaming media industry, you will definitely have heard of Streaming Media magazine and their EVP, Dan Rayburn. Recently, Dan blogged about the StrikeTracker console developed by my team at Highwinds. Dan covers a little history on his post so I won't repeat it all here, but it's great to see some of the things that he has written since they are spot-on with regard to our approach.

At Highwinds, we have a strong focus on user experience and meeting end user needs, which is evident by the fact that most of our CDN services and reporting can be provisioned and accessed in real-time by the customer without needing to engage the NOC. So, when Dan released the results of his survey of CDN customer needs earlier this year, we immediately incorporated it in to our product roadmap as a CDN customer "wish list". One of the key areas that we have been focusing on is to leverage the real-time nature of our CDN platform to deliver as much actionable analytics to the end user as soon as possible - usually within 30 seconds of the data being captured. This is obviously of great benefit to customers who are delivering marketing-driven content and/or working in a live environment where the timeliness of the information produced by the CDN is directly related to its usefulness.

To date, our focus on real-time analytics and empowering the end user through the StrikeTracker console is a trend that has won us a lot of favor with our customer base, and has brought over key customers from other CDN players. We certainly have plans to continue to build upon this strategy in 2009, although I obviously can't share the specifics without spoiling the surprise (not to mention tipping off the competition)! Let's just say that if we are successful, 2009 is going to be yet another strong year for Highwinds.

While I'm on the topic, Limelight also made a blog post on analytics recently that I found interesting. The following paragraph was of most interest. (Before I comment on this, I'll remind everybody that this blog carries the opinions of me alone and does not reflect the thoughts of my employer.)

"At Limelight Networks, currently accumulate and process over 100 terabytes of uncompressed log files each day due to the sheer volume of Internet traffic we deliver globally on our network. We offer byte-level accurate reporting on this traffic - not a sample or estimate, but an actual accounting of each bit we deliver. Reporting and analyzing this data in a timely and consistent basis is no small task when you are delivering massive traffic volumes. This puts us in a unique position to not only innovate, but also provide analytics and insights that few, if any, companies have ever delivered to their customers."

The blog post is clearly a generalization about CDNs and how their technology scales, which I find largely accurate based upon the industry players that I am familiar with. As Limelight points out, working at the scale of a CDN is extremely challenging, and it makes sense that this would be the case: the reason that the CDN industry exists is that we agreed to take on problems with Internet content delivery that nobody else wanted to, precisely because they are massively hard problems to solve.

What is interesting to me is that in solving these problems, many CDNs have taken something of a "cookie-cutter" approach, using the same (or very similar) network architecture and patterns as those established when the industry was in its infancy. Indeed, this lack of technology differentiation is a major reason why there is so much litigation currently taking place with regard to patent infringement in the CDN space.

The reason why Highwinds is successful is our core architecture, which is significantly different from every other CDN on the planet. For example, Limelight points out in their post that they accumulate and process over 100 TB of log files every day to deliver byte-accurate analytics. I believe that this accumulate-and-process approach is the same one that 98% of the CDNs out there would take.

What's interesting is that the accumulate-and-process approach is the direct opposite of the way we deal with reporting data on the Highwinds CDN. When you deal with analytics by waiting to accumulate it en masse and then processing it after the fact, you give up valuable time waiting for the accumulation and processing cycles to end. This lag in time doesn't make for a real-time solution.

So, we do something else entirely to process our data, and we can still deliver byte-accurate accounting and analytics. But unlike the other CDNs, we can do it in real-time, and that means you get the information you need when you need it - such as during your live event instead of after it is over. The nicest part of our solution is that it scales directly with the size of our network, so it will always be there as a feature regardless of how much content we are delivering.

Of course, our differentiated approach doesn't make dealing with the engineering problems of content delivery any easier. Every time we consider a feature for incorporation in to our CDN, we have to be able to answer the questions: "How do we keep this real-time?" and "How do we make this scale?" Answering these questions effectively without compromising on feature usefulness is certainly very hard. However, our CDN has the huge advantage of being built on a proven and scalable real-time platform to begin with. In many cases, our architecture answers these questions for us, which is a huge plus when you are working to deliver new and exciting real-time features to your end users that other CDNs can only dream about.

In closing, I'm really excited about 2009. We're taking a step back from StrikeTracker and re-evaluating every feature to determine what we have done well and what we can improve upon, and we have a mountain of constructive customer criticism both positive and negative to use as our guide. We've also got some awesome new features in mind that I believe will redefine what users will come to expect from a CDN administration console. I look forward to demo'ing them to you in person at an industry event in 2009.