Put together by a team of former Cisco and Juniper executives under the leadership of co-founders Alon Maor and Dan Sahar, Silicon Valley-based Qwilt, Inc. appears to be scoring strong interest among fixed and mobile network operators for a new type of video-centric caching platform. “More than 50 carriers worldwide with whom we are engaged are telling us that existing solutions to the online video challenge are hard to deploy and use, and they don’t scale well as the network grows,” says Maor, the firm’s CEO.
Qwilt, which just announced its first product release, is off to a fast start with deployments in ten networks in North America, Asia-Pacific and Europe, according to Sahar, who serves as Qwilt’s vice president of marketing. “Most of our team has been selling to operators throughout their careers, so we’ve been able to leverage that experience,” he notes.
Employing what is known as “transparent cache” technology, Qwilt’s platform can identify and capture the videos responsible for a large share of bandwidth consumption and make them available from headends, central offices, hubs, mobile gateways or other local access points so as to minimize the stream volume on the backbone. “Qwilt took a blank-slate approach, so we were able to design a new, single-platform solution that is significantly faster and more powerful than competitors’,” Maor says.
Transparent cache products are designed to complement the role of CDNs by performing various specialized functions at edge locations in the network without impacting user experience or the operations of other network elements, including CDNs. For example, Bluecoat, a company that’s been around since 1996, provides solutions that help enterprises optimize and secure all types of data flows over wide area networks. Another more recent entrant, PeerApp, offers transparent cache focused on video streaming, peer-to-peer transfers and software downloads.
Qwilt brings to this space an extremely compact solution devoted strictly to maximizing bandwidth efficiency for network operators by continually identifying and placing in edge cache those videos comprising about 20 percent of the titles that are consuming approximately 80 percent of the network capacity, Sahar says. The company’s first announced customer is Mitsubishi subsidiary IT Frontier, a Japan-based IT support firm that markets services and solutions to enterprises, including network operators, worldwide.
“We have identified an urgent need in Japan for a solution to ever-growing Internet video traffic and its associated costs for network operators,” says Masashi Ono, division general manager at IT Frontier. “After working with many members of the Qwilt team for over a decade, we are pleased to be partnering with them yet again and believe the company and its products will have a major impact on network operators worldwide.”
Interestingly, IT Frontier also recently announced it would serve as a channel outlet for the transparent cache solution offered by PeerApp. The Japanese firm’s engagement with Qwilt highlights the important nuances that differentiate various iterations of transparent cache solutions.
As described by Sahar the difference has to do with two factors: Qwilt is optimized for streamed video and it integrates everything in one, unified device that sits offline outside the data path, thereby expediting deployment by eliminating the overhead of device integration. “Experience has taught us that if you want to introduce a new product to network operators you have to make it very compact, easy to install and independent of reliance on other network elements,” he says. “In other words, it has to be drop and deploy.”
“We’re a software company riding on commercial hardware platforms,” he continues. “Our entry-level product, the QB 100, analyzes and classifies all data traffic at a rate of 20 gigabits per second and at the same time processes the video segment at 5 gbps. You don’t need a router or a DPI (deep packet inspection) device. The QB 100 does all this in a two-RU (rack unit) form factor.”
Finding the video traffic and delivering it in real time without slowing things down or disrupting non-video traffic is no easy trick, Sahar notes. Qwilt’s Online Video Classification Engine uses multiple techniques such as content templates, fingerprinting and weighting/thresholds to identify video content and provider signatures in the network. The ability to deliver cached video transparently to the end user relies on out-of-the-box delivery policy templates that have been tested and maintained for accuracy, which can be augmented with custom-built templates to fit special customer requirements.
The QB module “sifts through all the bits in the data stream and looks for video packets, which are large in size but represent less than one percent of Internet transactions,” Sahar explains. “Then it looks for repetition patterns and how much bandwidth is being consumed.”
The video may be a popular but very short clip, in which case there’s no need for caching because bandwidth consumption is low, he notes. But out of all that flow there will be a percentage of long-form videos that are repeating a lot and consuming a disproportionately large share of bandwidth. As a general rule, about 20 percent of the titles stored on a QB device will be replaced over a 24-hour period as a result of shifting consumption patterns.
Qwilt uses a multi-tiered storage approach consisting of SSD (solid state drive) and RAM (random access memory) that’s been purely optimized for video. “We architected this ourselves rather than using a generic storage system,” Sahar says. “It’s tuned to support both long-form and short-form titles and to be responsive to adaptive bit-rate formats.”
Where adaptive bit rate is concerned, the system is not just looking for how popular a given video is at any moment in time, he explains. It’s also looking for which versions of that video, in terms of streaming rates, are popular so that, for example, if the HD segment crosses the high usage threshold meriting storage in cache but the low bitrate video version being consumed on handhelds doesn’t, only the HD version will be stored.
Qwilt asserts its little boxes can produce huge savings in capital spending. “By cutting back on the volume of video traffic you can reduce cap ex over the parts of the network between the Internet and where you place this device by roughly 40 percent,” Sahar says.
He stresses that the Qwilt platform is not meant to supplant the use of private CDNs by operators but rather to extend management of traffic to over-the-top video that otherwise would not be affected by the operator’s CDN. “CDNs are good tools for solving managed service problems,” he says. “If I’m Netflix or some other Internet provider, I’m more likely to take advantage of Akamai, Limelight or other public CDNs rather than strike a deal to use the operator’s CDN. This is the area of concern we’re addressing.”
However, he quickly adds, that doesn’t mean the Qwilt platform might not become an integral part of the managed as well as unmanaged network strategy. “You might want to start out using our platform to reduce unmanaged OTT video traffic,” he says, “but later on you could tie the QB platform into the CDN operation for use with managed services as well.”
Presently, he adds, the platform is not designed to support tiered QoS service offerings to OTT suppliers. “The QB operates as a totally democratic device to support more efficient handling of the most popular video traffic,” he says. “Currently in the U.S. operators are hesitant to use these types of [tiering] policies for regulatory reasons, but if we were asked to have that capability, we’d develop it.”
But that’s not to say operators couldn’t use the Qwilt platform to strengthen their CDN offerings to OTT providers, who would gain not by having traffic prioritized but simply by having it benefit from the forward caching capabilities of widely deployed edge devices. “Our thinking about how to monetize the platform is to have the Qwilt device as an extension of the existing CDNs,” Sahar says. “Operators’ first priority is to optimize use of their networks with deployment of our devices. Once that’s done they can start thinking about commercial agreements.”