Interview #2 with Alan Arolovitch, CTO of PeerApp


Recently, we had the opportunity to interview Alan Arolovitch, CTO of PeerApp to discuss their new RAN based offering. PeerApp is a leader in the Transparent Caching industry, and has more than 500 customers, with some of those individual customers spending millions on their platform. The key takeaway from the interview is that PeerApp is not standing still waiting for emerging technologies to become the standard, but they are being aggressively proactive in embracing new innovations and implementing these technologies into their SDN and RAN architectures.

The world of RAN and CDN are converging, and PeerApp has emerged as a leader in this space. Content delivery in the RAN is the next biggest market to hit our ecosystem. PeerApp is in a great position to benefit from this new disruptive trend. A big thanks to Alan and the Executive Team for giving us insight into how RAN and Transparent Caching are transforming the marketplace.

Bizety Team: What is RAN, and what role does is play in the video delivery and transparent caching ecosystem?

PeerApp: RAN stands for Radio Access Network and is a complex and expensive portion of the mobile network that aggregates mobile traffic, from the core network, and transforms and transmits it to a subscriber device over the air interface.

In the traditional mobile network architecture, the content infrastructure is centralized and all video traffic delivered over mobile today must traverse both the mobile packet core and the RAN. At Mobile World Congress, we announced plans for a mobile edge caching solution that will enable delivery of video and other applications from within the RAN, dramatically lowering the cost and improving the quality of delivery. This is an evolution of our current mobile core caching in deployment around the globe.

BT: What is the difference in caching techniques and the way DNS works? When you’re comparing CDN versus the standard PeerApp caching platform and also the new mobile edge-computing platform? Is caching and DNS different, because of the spectrum?

PeerApp: The main challenge with caching at the mobile edge is that you’re dealing with a smaller subset of users and the traditional statistical approach is not applicable. The other main challenge or difference is mobility. You’re dealing with users that are constantly on the move, so are coming into and leaving different areas of coverage. During one large media transaction, like a download or long format video, a mobile user could transfer to areas covered by multiple different towers, and that has to be managed across the network. Our solution addresses those issues, and others, such as scale.  We manage this with a two-tier approach. Our centralized core-based solution brings network-wide intelligence about content popularity, and the edge-based solution tailors that for the subscribers associated with the particular edge location.

Our solution uses network insertion capabilities provided by our partner Saguna Networks, and that helps us manage the mobility challenge I mentioned.

BT: Let’s say we have a mobile network operator like Sprint or T-Mobile; the way it works currently right now is they’ll have cell towers in a region, like LA or Boston. Perhaps they’ll have 10-50 spread across the regions in order to cover the area. Are you saying that each tower is going to act like a POP?

PeerApp: First of all we’re talking about much large scale – tens of thousands of towers across the national footprint for a country like US. Part of the solution design is choosing deployment options, because RAN is such a massive network

What has been called “last mile” is in fact a very complex and distributed network, and especially so on the mobile side as you need to offer coverage everywhere. That requires a lot of nodes covering the geographical footprint and you have to be smart in how you go about the deployment.

The deployment design need to take into account several dimensions, such as: content demand density, cost of backhaul, and distance the solution spans. Obviously you want to consider higher concentration of demand in certain areas, but also the opposite – areas with expensive or hard to upgrade backhaul. Then if it is economically feasible, and makes sense from the total cost and density of demand point of view, you can see our solution literally being placed on the tower. We imagine that for any given operator, it will be a subset of towers that actually deploy the edge solution.

BT: Let’s say we have T-Mobile, and they’ve deployed their mobile compute solution. Then let’s say the user is watching a video while they are in transit, they’re in a car, for example, and they are jumping from RAN to RAN and provider to provider. So that means they are only going to enjoy an accelerated experience when it’s in a T-Mobile area correct? Is there any way around that?

PeerApp: Moving from provider to provider doesn’t really happen in the course of same of transaction. If you are the T-Mobile customer, you’ll likely stay within their RAN throughout the session. The real mobility scenario is that the mobile customer may leave the particular coverage area where edge caching is deployed. Our solution takes care of such mobility within same RAN.

BT: Thank you. Do you have MNOs that are testing this in beta, or any customers that are using this already or looking forward to it?

PeerApp: We are now in the beta process, and what you’re looking at is an extension of our core mobile solution. We’ve been selling mobile solutions to MNOs for 5-6 years; we have multiple large mobile groups as customers. This new mobile edge solution is natural extension of our existing centralized core-based caching solution; we are bringing those capabilities to the edge of the network, allowing MNOs to build on benefits gained in with open content caching in the core. Today, you have a centralized cache in 3G and LTE networks, deployed at what is called Gi LAN network. There is a set of functions there in the mobile network that provides various optimization capabilities like caching, mobile optimization, TCP optimization etc.

With the edge solution, we’re bringing those capabilities – and adding new capabilities over time – right into the RAN. We’re working with current customers and new customers on deploying this in the beta, and will be widely available later this year.

BT: So is this just like a plug-in to your regular PeerApp current transparent caching platform? Or is it a whole separate application? Is it a separate server? And can existing customers just plug it in and are up and running?

PeerApp: From a logical perspective, you need an additional plug-in so people who already have our centralized caching solutions will be able to add the edge capabilities. From a physical deployment point of view, there is a whole process of going into the RAN, identifying the right location that works with operators, and deciding which of the several deployment options is most feasible and desirable. Figuring out what is really important for the operator, what are the capabilities they want to support in the RAN, caching and DNS optimization, all these factors go into the physical deployment of the solution. We are distributing our footprint much deeper into the network, and placing the solution much closer to the end-user than before. The core benefits of caching are improving quality, reducing total cost, reducing latency, and enable broader mobile edge computing infrastructure.

From a strategic point of view, today we have a best-of-breed – and open – content caching solution that is deployed in both mobile and fixed operators. The edge solution helps to extend this caching function into the RAN, so you can cache in a much more distributed and efficient way.

The caching application acts as an initial use case that drives embedding of compute and store footprint in the RAN. Once you have that, you can build on it by adding more capabilities and other network functions on the same node, leveraging virtualization. There is a large number of applications that could benefit from having their footprint closer to the end-user.

BT: Why did you end up choosing Saguna Networks? Did you go through a elimination process of looking at different players, and Saguna had the best technology?

PeerApp: Saguna has all of the particular technologies that we felt were needed. They really have strong mobile edge expertise and capabilities; with some specific complementary applications that we felt we could leverage to solve the peculiar content delivery issues in the mobile network.   A few minutes ago, we talked about mobility and quality issues.  There are also issues dealing with lawful interception, policy and charging and others. Saguna already had a solution that addresses many of these and was available for trial, and what we really liked is the immediacy of it. It could be delivered and deployed in an existing network, without going through massive network transformation first. Saguna provides a way to deliver a higher quality solution now, and can also scale and work later as new architectural options and new approaches become popular. So what we really like is the immediacy, being able to deliver value, and the future-proof qualities in Saguna.

BT: Who are your two biggest competitors in this space? Is it Ericsson or Alcatel Lucent? Do you come across them? It seems like this is right in their backyard, and they’re pretty big in this area. They also provide soup-to-nuts just like you do, maybe even a little bit more.

PeerApp: There are different soups and different nuts. If you’re looking at what’s been happening around the RAN, you can see a trend around making RAN generally more intelligent. RAN used to be something that was radio-oriented or radio-focused, with no content delivery awareness. Now you have self-optimizing RAN that’s emerged, where the network is smart about how you’ve allocated radio resources based on a variety of factors. The other trend – which we are part of – is bringing computing into the RAN.  This is what is called mobile edge computing. These are happening at the same time and there are several trends that are converging. ALU, Ericsson and Nokia are primarily acting as end-to-end RAN suppliers and focus on broad enablement of applications.

The issue with any enabling platform is that you need a “killer app” or first application to drive adoption of the platform.

We believe that caching is a perfect “seed” application and one that delivers value right now. We are experts in that. We certainly also see room for collaboration with broader ecosystem providers, including the RAN suppliers you mentioned.

BT: So you already have Akamai partnered with Saguna, what happens if other companies like Limelight, EdgeCast, and Level 3 start partnering with Saguna or companies like Saguna? What if they start offering their own solution? Do you think your methodology with caching is a better way of delivering than what other CDNs are using? Why?

PeerApp: The primary question I think is, how do you make sure what you’re offering gets deployed? It’s a Catch-22 situation. To have a footprint, you have to have applications and what have you on it; to be compelling for CDNs you have to already have a footprint. So how do you solve that? Again, I think to deliver value out of the box is something that transparent caching provides. On top of the footprint we can work with multiple partners, whether they are CDNs, video delivery platforms, anyone that will enhance your already existing capabilities. As you may know, we have partnered with several CDNs before, so we see our value and our role being a provider of content delivery infrastructure for network operators.  That is a platform from which we can specialize, and we don’t see other companies really being competition in that particular niche.

 BT: What role is network function virtualization playing in your mobile edge compute solution?

PeerApp: The initial view of NFV has been to deploy centralized solutions, to get to economies of scale by co-locating multiple network functions in a centralized scaled out footprint. At the same time many network functions, and especially content caching and delivery, need to be placed as close to the “eyeballs” as possible. In this context, virtualization can play a different role and be an efficient way to scale down. Virtualization enables us to have a scaled down version of the platform and drive the cost down, and also have more functions within the same distributed edge footprint.

BT: What kind of changes are you seeing in the mobile network market? What are some trends? Is there a possibility that Google may buy some spectrum?

PeerApp: Ok, so we are talking about mobile macro trends I think, there are several things of note.

The way mobile network evolution is traditionally being talked about, it is all about speeds and feeds, supporting faster last-mile. While that is certainly important, it is only about upgrading the radio interface. Doing content delivery and cloud computing at the edge, using virtualization, with location awareness is going to be an important piece of infrastructure to optimize the whole network and enable new exciting applications.

Another macro trend that is difficult to ignore is an emergence of WiFi. WiFi was initially used as an offload for 3G mobile network as a temporary fix, before LTE takes off. The industry data indeed showed that percentage of mobile data offloaded to WiFi decreased in LTE. But WiFi re-emerged and is being used in new contexts as a macro mobile network –community and municipal WiFi, cable operators like Cablevision rolling out fully blown mobile offering based on WiFi footprint and now Google planning to use WiFi offload in their MVNO.

The unlimited WiFi-based data from cable in combination with existing triple-play could bring just the disruption that the US mobile market needs. Internationally, in France and other markets, we have seen new mobile operators enter the market and drop the price across the board.

In terms of consumer choice and options, this is definitely a good thing, and I think Google could play a role with their plans to partner with Sprint and T-Mobile, and using WiFi as an offload – they can be such a disruptive force. The disruption threat coming from several directions – it will be interesting to watch in the next 12-24 months.


Scroll to Top