Radiance optimizes corporate networks

RADIANCE TECHNOLOGIES IS a provider of software that optimizes networks for corporations that need to send large amounts of data across geographically dispersed organizations. CEO Ash Munshi — who counts Fidelity, Bank of America, and BEA Systems among his company’s customers — is about to launch Version 2.0 of Radiance’s TrueDelivery System and is working with Tivo on that company’s next-generation set-top box. In an interview with InfoWorld Editor in Chief Michael Vizard and Test Center Director Steve Gillmor, Munshi explains why the proliferation of storage and limited network bandwidth means IT will always require a system for managing data caches inside and outside of the enterprise.

InfoWorld: What problem is Radiance trying to solve?

InfoWorld: How does the system work?

Munshi: The basic technology allows companies to deploy essentially an overlay network on top of their network, so it’s a complete software-based solution that they deploy across their network. It handles bandwidth throttling and management so you can do all the appropriate things relative to bandwidth assignment. You can do both reverse and forward proxy caching so you can cache objects, you can move them around, and you can populate specific caches. We basically do what we call managed delivery of these objects, where managed delivery connotes two things. One is a guarantee of delivery and second is a quality of service associated with that delivery. When we talk about guarantee of delivery we actually make sure that every bit got transferred across [the network], whether it’s from one location to many or one to one, depending on how that needs to get done with essentially a transaction. The second part of it is the quality of service, which is a commitment … in terms of the time frame that it will take a particular piece of data to be delivered to all the different endpoints or specific endpoints or server-to-server, as one example. The system is completely fault-tolerant, it is completely software-based, and it doesn’t rely on doing anything at the packet level. It really is done at the application level, if you will.

InfoWorld: How is that accomplished?

Munshi: It is all part of a caching mechanism. The way the system works is there is a central repository, essentially, which can interface through XML or Java or a variety of different mechanisms to pull either the data itself or just the metadata associated with the data. We need to manage the metadata. We don’t need to manage the data itself. On your network you basically put caches which act as proxies. These caches are completely centrally configurable and you can manage the policies on these things. We have a lot of algorithms that we’ve worked on that manage both the bandwidth assignment and allocation along with the storage, because those things are actually coupled when you deal with large data files. They’re not independent entities because the delimiting case is that what you do is you dump lots of storage and you move everything out to each one of these caches and then you move it around. The problem is you never have that case. You want to use cheap storage and you want to make sure that you don’t fill up the network along the way. In addition, each one of these caches — and in fact any node in the network — can be configured from an availability perspective relative to bandwidth. So, for example, if you’re doing database backups in the middle of the day or in the middle of the morning, you don’t want those caches to be used because you’re doing something else on the network. You can configure the system to have whatever bandwidth profile you need to have. And then, when we give you a delivery schedule across that entire network, we integrate over all those possible routes and all those constraints in a dynamic fashion to be able to give a delivery time. Then we make sure that delivery time is obeyed. So it’s a complex solution that happens every time a new query essentially is put into the system.

InfoWorld: So once you add this software into someone’s network, what else will Radiance be able to do?

Munshi: Obviously we’re going to move up the value chain. Our intent is, as we get into some of these companies and we understand a lot more about how they move data, we believe that we can actually provide other services over and above the other applications … that take advantage of the data that we’ve collected as a result of being the transport mechanism. What’s interesting is there is nothing inherent in what we’re doing that really forces us to be applied to a large file or long latency. It just so happens that that’s kind of where we’re focused right now because there is a pain point in some very large companies around that specific problem, particularly over the wide area network. We think we can build a substantial company on that basis alone, and at the same time, once we get a foothold into that, we can easily go to the other venues in a very reasonable way.

InfoWorld: How much security is layered into your offering?

Munshi: The transport that we provide is completely secure. You can, in fact, wrap it around any digital rights management scheme that you want. You can have your proprietary licensing scheme and your proprietary authentication scheme, and it doesn’t really matter to us what the mechanism is that you use to do the transport.

InfoWorld: Won’t broadband be essentially free some day, thus obviating the need for your service?

Munshi: We don’t think there will be such a thing. And the digital logistics problem that [everyone] needs to solve doesn’t go away when the highway gets bigger. Even with such bandwidth, the desire to manage the movement of the stuff by prioritizing and playing traffic cop is critical.

InfoWorld: What makes your approach different from service providers that specialize in this area?

Munshi: The interesting thing is we’re not a service. It’s stuff that you deploy inside your network. What we are talking about here is guaranteed delivery and quality of service that’s completely controllable by you in terms of how you want to be able to get this stuff across a network.

InfoWorld: Why does this problem require a dedicated solution vs. relying on a network hardware vendor to tackle this?

Munshi: If you think about networking guys, they work from the bottom of the stack up. As they go up toward basically becoming application-aware, they really don’t have the entire understanding of what the application requires. They understand only small bits of the data and they can basically prioritize. Just like if you think about the freeway system: Somebody can prioritize that the express lane … is something that gets higher priority because there are two people or more in the car. I’ll let them in faster because I’m aware that they’re going to use the freeway more efficiently. Similarly, Cisco can go look at it and say, “This is a packet that’s video, so I need to really prioritize it higher, and therefore I’m going to do differential scheduling of this stuff.” The problem is it’s still [done] down at close to a packet level. What we’re doing is we’re really doing it at the application level. It’s a very different layer problem, and I would contend that no matter how hard a Cisco tries, they will never have the same application-level awareness that we will by being down at the switch.

InfoWorld: What challenges will Web services and XML present at the network level?

Munshi: There are essentially two types of transaction on a network. There [are] essentially what I’ll call synchronous, or relatively small-latency, transactions and then there are the long-latency, asynchronous transactions. The part that we focus on is not the synchronous or low-latency stuff; it’s the long-latency stuff that we worry about. I believe what’s going to happen is that for those applications that really require the handshake that is relatively synchronous, they’re going to have more efficient mechanisms inside the network — and probably through the switch itself — where you’re going to be expediting using the fact that you know these are XML transactions. You might even know the structure of the packets and therefore you’ll be able to do the handshakes much better. But we’re talking about the stuff that’s really asynchronous from that particular perspective.

InfoWorld: So at some point will Radiance try to set an industry standard in this area?

Munshi: I actually don’t believe in standards bodies and I don’t think they’re actually terribly effective. I’ve been in the computing business now … for a number of years. I was involved [while at Oracle] with SQL 2 and all those kinds of things going on. I have yet to meet a standard that truly is a standard that everybody really does believe in and something that actually solves the set of problems that needs to be solved. I’ve played roles. I’ve had people go to standards bodies and participate in the discussion, but the purpose of it is usually not to participate, it’s to hijack it.

InfoWorld: So what’s the one technology trend that you are most excited about?

Munshi: I don’t thing networks are ever going to be fast enough to handle the kinds of data that [are] out there. Caching is going to be there for a long time. I think distributed storage will become much more prevalent, and being able to do the right kinds of things with it is going to be very interesting. If you think about all the amount of storage that’s going into homes, that’s a hell of a lot more storage than inside the enterprise. What are the kinds of things that you might be able to do with that? That’s a fascinating problem. What’s interesting is networks will never keep up. You’ll be able to get the data where you need and computational power is now sufficient that you’ll be able to do some very interesting things with that particular data that’s there. People always want more bandwidth and it’s not all the same to all places, so there’s a major logistics problem that gets bigger over time as storage is everywhere. The whole reason why you have caches is because you want to be able to control the quality of service effectively.