Interview with Alan Arolovitch, CTO of PeerApp


Recently, we had the opportunity to interview Alan Arolovitch, CTO of PeerApp, the leader in the Transparent Caching market. PeerApp which started in 2004, is growing like gangbusters. They have a total of 500 customers, signing up sixty in 2014 alone. That has to be a record in our industry. Alan is a visionary, very well spoken, and has an incredible grasp of web infrastructure, CDN, and Network Operators. The interview was important to many, because some of us in the industry have always had some form of doubt about the transparent caching platforms. The biggest question we had is how can an entire CDN infrastructure be packaged in a product, and it work efficiently like a CDN, because there are so many different moving software parts in a CDN platform.

This interview has put the nail in that coffin of doubt, and we are now firm believers in Transparent Caching solutions. One point we want to emphasize, is that we did not give Alan the questions beforehand, but we asked him right there and then. Unscripted Q&A is the best way to do it, because the natural self and intellect come out. If we had to grade the quality of the answers to the questions, all we have to say is that Alan should enter academia later in life, and become a professor at Standford or MIT.  A Big Thanks to Alan and the PeerApp Executive Team for educating us on the market. Here we go, enjoy.

Q: How is business? How many customers do you have right now? Are you at liberty to discuss how much revenue you’re generating yearly?
A: In terms of revenue, we’ve seen significant growth in 2014, which is in line with what we’ve seen before. We added 60 customers in 2014 and we are approaching 500 total customers worldwide. I believe we’re increasing our market share as compared to overall market growth.

Q: I read you closed a deal last year with a tier 1 cable operator for 5 million dollars. You don’t have to share the customer’s name, but could you provide details? Was it successful?
A: It’s a Latin American cable operator that belongs to a tier 1 group. We’re talking about roughly a million high-speed data subscribers and deployment on a scale of roughly 200Gbps at the metro level in the cable network. This is really a significant project for us.

The deployment and interest of the buyer has been focused singlehandedly on quality of experience (QoE). Customer surveys they conducted looked for a mechanism to differentiate the DOCSIS infrastructure and investment they made early on as most MSOs did. They were thinking, “What else can we do beyond rolling out a very fast last mile, which packages 100 Mbps to the home? What else can we do to improve quality of experience and make our offering more competitive compared to incumbent telcos?”

That has been our focus. We worked with the product group in the company versus the traditional network-engineering group, which typically deploys caching solutions around this. To us, this highlights a market change. Certainly, in terms of business, there are network optimization savings factors, but the driving factor and buying motivation was primarily differentiation around very fast broadband networks.

Q: Are you incorporating any SDN elements into your product and message? How do you see SDN in your world? What does it mean and will you maintain a bigger focus on that space in the future?
A: SDN (Software Defined Networking) means many things. Most typically, the market talks about SDN in conjunction with NFV (Network Function Virtualization) deployments as a way to support various service chaining use cases like subscriber services that are easily provisioned. You can handle traffic for a particular user in a dynamic way for various services. We’ve been doing a lot of work historically with network equipment providers. We partnered with multiple DPI (Deep Packet Integration) vendors and did extensive integration with them; we work with routers and so on. So, out of this we developed redirection capabilities that enable us to basically act as an SDN application.

We were talking about SDN concepts before the name even existed. Now, we’re looking at use cases that will enable us to intelligently drive traffic to the content platform. That can be based on incorporating that directly into our solutions and building dynamic traffic steering capabilities into the solution itself, and also partnering with existing SDN solutions out there, with SDN controllers, and the whole thing.

Here, I distinguish SDN and SDN/NFV combination; it doesn’t have to be necessarily tailored into bigger virtualization revolutions that are happening. We believe SDN has value in pure traffic steering with or without virtualization. So there is value together and separately. Also, there is value that is not necessarily subscriber-focused, like applying SDN intelligence into selective traffic based on various parameters that a content platform can bring to the table. This is around origin of traffic, devices being used, subscribers, and efficiency. We see SDN as a key component and capability of our next generation platform that we’re launching this year.

Q: When I talk about SDN in the CDN space, I point to an example. It’s clear cut – Akamai partnered with Juniper and they have an SDN solution, which means they can copy and paste their CDN platform to a network operator infrastructure using bare metal hardware. That means they can copy and paste their caching platform, reporting, logging, security and so forth. It’s a very innovative trend and more companies will follow suit. Do you see yourselves doing that and will that have an impact on the transparent caching market?
A: If you look at a traditional network CDNs that focus on managed content, they started as bare metal streamers and they’re now moving to virtualized environments, working with SDN controllers and virtualization. It’s really natural and I think that’s sort of the common step to take.

Where things start to get interesting is in terms of how you leverage intelligent network capabilities and SDN built into networks to send and steer traffic intelligently. Do you really combine, for example, and build CDN overlay and regress routing into SDN? Can you really build request routing and overlay, rather than use traditional CDN methods like request DNS based and anycast? Can you actually use SDN for that? That would be really new stuff and cutting edge.

The first step that I would take is to say, “I run it on bare metal, let’s move into virtualized; I live in the virtualized data center now and can work with SDN, okay what’s next”? Other vendors are doing that. The question is, are you taking and integrating SDN capability into the application layer and making new things possible because you have intelligent networking? That’s what we are planning to do. We’re saying, “If I have an intelligent network, I can do more sophisticated things? And, I’m different from CDNs because I deal with all of the traffic on the network. If I can be intelligent about how it is processed and routed, I think there is a lot of value there.”

Q: Can you talk about any new features you rolled out in 2014 and those you plan to roll out in 2015?
A: In 2014, we announced a capacity upgrade on the platform, continuing the scale of solutions so our product line has been upgraded. As a result, we continue support scaling from 1 node all the way up to a 16-way cluster. Now every node supports double the capacity and therefore we can approach 160 Gbps per system, which is what we are shipping today. That’s what’s been publicly announced. Future announcements are coming.

Q: In the past, I saw transparent caching as a CDN in a box, so you ship out boxes with software and one box does everything – reporting, caching, etc. So I would say to myself, “If you do it in that fashion, it’s very limited compared to the CDN feature set, so why would customers want to do that and will the business as a whole work? Because a CDN can provide a full suite and it’s on-demand and global in scale. A CDN can charge on a per Gigabyte pricing scale and its quick to set up and inexpensive, whereas a transparent caching platform includes buying boxes and shipping them all over the place.” So I always questioned the business model, but you’ve proven otherwise. What are your thoughts?
A: I think there is a misconception out there, and I’m glad that you now see it differently. What people don’t appreciate about transparent caching is what’s involved. Also, the global CDNs generally ‘drop off’ content into the operator’s network. Transparent caching comes into augment the operator’s network and optimize “last mile” delivery. In the last mile you’re dealing with a different situation than a CDN, which in many cases is pre-provisioned with known content. One could look at a CDN as a caching network that has knowledge of the content, the format, and the URLs. Content publishers provision information to the CDN and things are events and loading is to some extent predictable and known.

The world of transparent caching deals with content coming down a pipe and has to – in the best way possible – identify popular content across various sources. Where is it coming from? It can come from CDNs, peer-to-peer networks, somebody’s cloud, whatever. And you have to do the best job identifying that and also supporting the business logic involved.

There has been an evolution of caching. HTTP has transformed from a URL-focused object naming space with clear caching prototypes and tools enabling you to manage freshness of content, and so on. A URL used to be a unique address of an object and various caches in the browser or network caches manage freshness around that using HTTP standard. Instead, today HTTP has become like a TCP and the URL has become like a function call. If you’re looking at the way the Internet is structured today, there are various aspects of dynamic logic that exist between client and server. You have dynamic URLs that will resolve to different objects.

First of all, one-off URLs are now used by various popular content publishers like Netflix and YouTube and others with subscriber tokens in the URL. You have adaption to device where the same URL may result to different objects or redirection to other mobile websites depending on the device you’re coming from. There is various time-to-live (TTL), very dynamic, an object is there and has been taken down from YouTube and is no longer there. So client and server no longer talk in the static URL language, they talk in functions. And caching has to deal with that and do the best job supporting that business logic and not breaking stuff so you don’t end up sending wrong content or delivering stale content.

Ultimately, CDNs and transparent caching are really different. There is a lot of sophisticated logic in caching somebody else’s content without having any knowledge of it; and the transparency aspects – transparency to user, transparency to the content source, and transparency to the application logic – has a lot of sophistication and differentiation. PeerApp has been able to do that very broadly without needing a lot of customer intervention. We don’t need configuration; we don’t need touch points on the cache to teach it how to do this. We feel that we do this better and more broadly then all of our competitors. And again, the underlying technology between both a CDN and transparent cache is caching.

So, some people may say “oh you sound like a CDN” but if you’re looking at how transparent caching is being used, it’s really closer to other optimization solutions – for example DPI or mobile optimization in terms of how it’s used and how these things are deployed. Very commonly we’re deployed in mobile networks in conjunction with DPIs and mobile optimization and other optimization solutions. So, that optimization solution space exists where operators are doing different things on the content and application level to optimize networks for cost and quality and that’s where transparent caching lives.

You can’t use CDNs to do the same, but at some point people believed they could. There was a time, around 2009-2010, in which there was an infatuation with the concept of operator CDNs, with no real sustained traction.

If you’re looking at operator CDN platforms that live and flourish today, owned and operated by network operators, those managed CDNs solutions where operators deliver mostly, and I qualify that, their own content to their own subscribers on their own network. And that’s different and separate from your global CDN space where the Akamais and Limelights live. That’s not an offload, but really a migration of VOD (video on demand) solutions from proprietary networks onto IP networks. But still operators are left with the need to handle OTT (over-the-top) stuff that comes down the pipe. And that’s where CDNs fail to help operators deal with that. That’s where transparent caching has seen a boost and growth in the business, helping operators deal with somebody else’s content.

Q: It appears that you’re targeting and going after non-network operator markets. Are you going after different industries or verticals, so you’re not limiting yourself to one specific market? If you are, have you had any success?
A: Great question because that is indeed happening. We cut our teeth on the service provider environment and that’s an environment characterized by a particular traffic mix and particular scale problems and issues. The service provider is a very specific environment. Of course, Internet access is more ubiquitous and applicable to more companies. And the closest so far that we have seen, in what you could call alternative service providers, are universities and school systems offering Internet access and that’s where we’ve seen quite a lot of business and is an area focus. It’s interesting, in view of demographics and behavior, what we’re seeing is that college and university networks, act as service provider networks. There are no limitations, no policy, and people are consuming video and media content at lengths and at very large volumes. These networks face the same quality and cost issues that traditional broadband providers face.

What’s interesting about these networks is that you see generational change. We talk about these new behavior patterns, what we call ‘broadband first’ behavior, where users are switching over to broadband for their primary means of consuming content, any entertainment, whether that’s live television or on-demand, it’s broadband first. Those are the customers that are consuming much more content. There is an interesting statistic that came out from Sandvine in the first half of 2014 in their Internet Phenomena Report. They talk about people exhibiting ”cord cutting” behavior that are consuming media in the volume amounting to 100 hours of video per month. So, whether they still continue to pay cable television or not, these are the users whose share of time is dedicated to the broadband environment and watching on iPads and connecting to televisions and so on. We see that behavior in colleges and universities so much more.

People talk about the millennial demographic and we see in our system just how much more data and media is being streamed in this environment and how that predicts what’s to come for all networks. That’s one segment that is interesting and fascinating and bringing business to us across multiple regions. I think there is also an opportunity beyond that, going after alternative service providers that are not necessarily universities. We see municipal providers offering Internet, municipal WiFi. We see enterprises in certain situations offering Internet access and again at that moment you’re in that situation where there is a lot of content and consumption and you need to manage quality and cost. So, we see a broad fit.

Q: What is the biggest trend right now that will impact your space? Is it going to be SDN, cyber security, big data, Hadoop? What technology is going to impact your space and force you to revamp architecture? Is there any specific technology that is changing the way everyone is operating in your space?
A: With your permission, I will interpret the question broadly. I would say in terms of trends I would point to two that are significant. One is external and one is internal. I think an external trend that we see as very significant is the whole migration of various content services to over the top delivery. If you look at October of last year, there were particularly fruitful announcements from HBO, Univision, Sony, and CBS going direct. And Dish just announced the same. To us, that indicates a quality change in the market. It’s not just more of the same, it will signal more supply of content and drive more consumption. And the problems or issues that we’ve been helping our customers deal with – again with somebody else’s content, how you manage and become part of that, how you manage cost and quality – those issues will be accelerated because we believe this is just the tipping point and we’re standing in front of a massive increase in Internet consumption as more content consumption moves away from traditional paid TV access towards over-the-top and Internet-based. That could be a driver in the US, but in parallel we see same thing in Europe and advanced economies in Asia and beyond. So for us, that’s a big market trend that will move a lot of boulders around here or around paid TV and Internet and content.

From a technology point of view, I would say SDN and NFV are big trends. There are multiple opportunities that open around that, which could be a significant thing for us; and there are a lot of things happening in those areas and operators are all taking different directions. We’re seeing trends around applied virtualization in a distributed manner, taking ‘hyper web’ companies’ practices of scaling out data centers, using virtualization, and using low cost hardware, resulting in scaling out data centers as a practice. If you take this into telco, this translates into scaled out large data centers for everything. And operator networks, if you take it to the extreme, become pipes with large data centers and everything happens in the data center as a result. I think that was the initial trend, but I think now we’re seeing new telco-specific applications where people are looking at distributed virtualization. Cisco came out with an interesting term called Fog Computing, which is kind of edge-based virtualization. Others are looking at the same thing and there will be a lot of transformations and potentially new players emerging in the telco space. More vendor companies not traditionally associated with the telco space will come in as a result. I think that is generally a single significant trend from a technology point of view that affects our business.

Q: Do you see yourself and other companies in the CDN ecosystem starting to use an SDN message within value propositions and stories, as well as making it a key feature in some ways? Do you see yourself becoming a member of an open source SDN project like Project Floodlight or OpenDaylight or beginning to implement that in your architecture?
A: We see the traction and momentum that these are enjoying and we’ve seen large network vendors initially come out with their own proprietary implementations before switching over to OpenDaylight. OpenDaylight has certainly enjoyed a lot of momentum in operator networks. There are other efforts, open NFV efforts that we’re following as well so I would think there is probably more than one effort that is interesting. We’ll probably join some of them, which specifically I think I’m open and fluid. From our point of view, we kind of primarily see ourselves as a VNF (virtual network function) provider in this context, so as such we see ourselves first using capabilities like Open Daylight as just infrastructure. So, it would be natural for us to join it.

Scroll to Top