A new video streaming platform achieving real-time latency without the use of WebRTC is undergoing early deployment by nanocosmos, which has built its business supporting low-latency streaming needs of enterprises typically not involved in mainstream media services.
While the new strategy is based on what nanocosmos refers to as “the new next-generation streaming protocol MoQ (Media over QUIC),” it actually has nothing to do with the IETF MoQ standardization initiative that we’ve been tracking, most recently in this article. Instead, as described by nanocosmos CEO Oliver Lietz, his company’s version of MoQ is its own adaptation to use of the QUIC protocol.
“We see great potential with the addition of MoQ using the underlying UDP (User Datagram Protocol) stack as a better approach to connectivity for real-time streaming,” Lietz says. That’s because adding its version of a MoQ option to streaming over its global CDN can overcome some of the latency-inducing issues that occur with its proprietary approach to lowering latency using HTTP Live Streaming (HLS) with the Transmission Control Protocol (TCP).
But Lietz makes clear the company remains focused on one-way streaming instances targeting up to 100,000-200,000 users, with no aspirations at this point to achieve the higher scalability with support for multidirectional real-time streaming enabled by some WebRTC platforms and the new MoQ protocol under development at the IETF. Instead, the goal is to create an end-to-end streaming environment within the current use-case scenarios targeted by nanocosmos in order to lower customers’ latencies from current one-second and sub-second levels to sub-500ms.
“We don’t have a lot of customers in the traditional media space,” Lietz says, noting, “eighty percent of our streams are on mobile devices.” While nanocosmos isn’t pursuing multidirectional video streaming, he stresses that instant data-linked messaging is a very important component of interactive use cases driving a large share of customer engagements.
That includes “anything that requires real time engagement like auctions, live gaming, sports betting and other applications where you need to support interactive communications,” he says. Corporate training scenarios are an important niche, he adds, citing as one example a finance trainer app that allows trainees to deliver advice on stock trading in sync with what’s happening on the exchanges in real time.
The new MoQ streaming mode was made available to “select” customers in different parts of the world in conjunction with the Fraunhofer FOKUS Media Web Symposium that took place in Berlin June 24-25. Lietz says the company intends to make MoQ streaming ubiquitously available on its CDN in time for IBC.
Once all the CDN servers are upgraded with the new software, nanocosmos will upgrade all customers’ client players to enable the shift to the lower-latency protocol with no big changes in targeted use cases, Lietz explains. He says there won’t be any impact on pricing.
The main driver behind converting its CDN to MoQ transport has to do with the fact that once streams leave the TCP-based cloud CDN environment they are subject to impediments over the open internet that trigger a lot of packet retransmissions and attendant buffering delays. “In theory TCP may work fine, but line blocking that may not be present in test environments is very much an issue in real-world operations,” Lietz says.
“Under optimal conditions, there’s not much difference (between use of TCP and MoQ),” he adds. “But in a hostile network environment, using UDP with QUIC makes it easier to overcome the roadblocks, which reduces buffering delays.”
QUIC was developed by Google and now has been incorporated by the IETF as the default transport for HTTP/3, the latest version of the HTTP standard. HTTP/3 hasn’t yet been widely adopted with video streaming but is already in common use for other types of HTTP use cases with support in most of the leading browsers.
As described in IETF documents, QUIC employs advanced forward-error-correction (FEC) and other mechanisms designed to enable stream continuity that matches or exceeds TCP performance, depending on whether TCP is configured to focus on resending just the blocked packets that really matter. In previously standardized iterations QUIC cuts end-to-end latency to the 4-second range, well below what’s typically achieved with TCP.
But, as previously reported, Comcast at last year’s IBC Show demonstrated latency over TCP connections can be reduced to 1.8 seconds. They showed that through reductions in latency contributions from encoding, packaging, buffering and other points in distribution this could be done over HTTP connections that are interoperable with the dominant MPEG DASH and HLS streaming modes. The Comcast team said they could use QUIC to shave off another 100ms or so but felt that didn’t justify making the switch.
The Comcast group was focused on demonstrating a solution that can be used for mass media distribution of live content at 4K/HDR quality. nanocosmos, through use of its own CDN backbone incorporating its proprietary approaches to reducing latency over TCP and QUIC in what amounts to a closed distribution environment, is providing a platform that goes much further in latency reduction for customers pursuing the targeted use cases. “We don’t worry about connecting to other CDNs,” Lietz says.
While the primary reason nanocosmos is bringing its MoQ strategy into play has to do with cutting latency on connections between the company’s CDN edges and end points, Lietz says, “We’re very interested to learn when we role this out everywhere what the impact will be over the middle mile. We suspect we’ll see benefits where the bandwidth connectivity between servers isn’t stable.”
As for the IETF’s version of MoQ, Lietz acknowledges that would enable extension of real-time connectivity to millions of people over multiple CDNs. While the nanocosmos version works “very well” with the limited user concurrency targeted by its customers, IETF’s MoQ could “enable certain benefits to our users,” which, though he didn’t spell them out, would include multidirectional streaming. “It’s an option we might consider as we continue to embrace new standards, but that’s not decided,” he says.
Right now, he adds, it’s important to understand that demand for real-time streaming doesn’t require a one-solution-fits-all approach. The niche targeted by nanocosmos is vast and growing, Lietz notes, which means his company’s challenge is to provide the most cost-effective way to satisfy those needs as they continue to evolve.
While the new strategy is based on what nanocosmos refers to as “the new next-generation streaming protocol MoQ (Media over QUIC),” it actually has nothing to do with the IETF MoQ standardization initiative that we’ve been tracking, most recently in this article. Instead, as described by nanocosmos CEO Oliver Lietz, his company’s version of MoQ is its own adaptation to use of the QUIC protocol.
“We see great potential with the addition of MoQ using the underlying UDP (User Datagram Protocol) stack as a better approach to connectivity for real-time streaming,” Lietz says. That’s because adding its version of a MoQ option to streaming over its global CDN can overcome some of the latency-inducing issues that occur with its proprietary approach to lowering latency using HTTP Live Streaming (HLS) with the Transmission Control Protocol (TCP).
But Lietz makes clear the company remains focused on one-way streaming instances targeting up to 100,000-200,000 users, with no aspirations at this point to achieve the higher scalability with support for multidirectional real-time streaming enabled by some WebRTC platforms and the new MoQ protocol under development at the IETF. Instead, the goal is to create an end-to-end streaming environment within the current use-case scenarios targeted by nanocosmos in order to lower customers’ latencies from current one-second and sub-second levels to sub-500ms.
“We don’t have a lot of customers in the traditional media space,” Lietz says, noting, “eighty percent of our streams are on mobile devices.” While nanocosmos isn’t pursuing multidirectional video streaming, he stresses that instant data-linked messaging is a very important component of interactive use cases driving a large share of customer engagements.
That includes “anything that requires real time engagement like auctions, live gaming, sports betting and other applications where you need to support interactive communications,” he says. Corporate training scenarios are an important niche, he adds, citing as one example a finance trainer app that allows trainees to deliver advice on stock trading in sync with what’s happening on the exchanges in real time.
The new MoQ streaming mode was made available to “select” customers in different parts of the world in conjunction with the Fraunhofer FOKUS Media Web Symposium that took place in Berlin June 24-25. Lietz says the company intends to make MoQ streaming ubiquitously available on its CDN in time for IBC.
Once all the CDN servers are upgraded with the new software, nanocosmos will upgrade all customers’ client players to enable the shift to the lower-latency protocol with no big changes in targeted use cases, Lietz explains. He says there won’t be any impact on pricing.
The main driver behind converting its CDN to MoQ transport has to do with the fact that once streams leave the TCP-based cloud CDN environment they are subject to impediments over the open internet that trigger a lot of packet retransmissions and attendant buffering delays. “In theory TCP may work fine, but line blocking that may not be present in test environments is very much an issue in real-world operations,” Lietz says.
“Under optimal conditions, there’s not much difference (between use of TCP and MoQ),” he adds. “But in a hostile network environment, using UDP with QUIC makes it easier to overcome the roadblocks, which reduces buffering delays.”
QUIC was developed by Google and now has been incorporated by the IETF as the default transport for HTTP/3, the latest version of the HTTP standard. HTTP/3 hasn’t yet been widely adopted with video streaming but is already in common use for other types of HTTP use cases with support in most of the leading browsers.
As described in IETF documents, QUIC employs advanced forward-error-correction (FEC) and other mechanisms designed to enable stream continuity that matches or exceeds TCP performance, depending on whether TCP is configured to focus on resending just the blocked packets that really matter. In previously standardized iterations QUIC cuts end-to-end latency to the 4-second range, well below what’s typically achieved with TCP.
But, as previously reported, Comcast at last year’s IBC Show demonstrated latency over TCP connections can be reduced to 1.8 seconds. They showed that through reductions in latency contributions from encoding, packaging, buffering and other points in distribution this could be done over HTTP connections that are interoperable with the dominant MPEG DASH and HLS streaming modes. The Comcast team said they could use QUIC to shave off another 100ms or so but felt that didn’t justify making the switch.
The Comcast group was focused on demonstrating a solution that can be used for mass media distribution of live content at 4K/HDR quality. nanocosmos, through use of its own CDN backbone incorporating its proprietary approaches to reducing latency over TCP and QUIC in what amounts to a closed distribution environment, is providing a platform that goes much further in latency reduction for customers pursuing the targeted use cases. “We don’t worry about connecting to other CDNs,” Lietz says.
While the primary reason nanocosmos is bringing its MoQ strategy into play has to do with cutting latency on connections between the company’s CDN edges and end points, Lietz says, “We’re very interested to learn when we role this out everywhere what the impact will be over the middle mile. We suspect we’ll see benefits where the bandwidth connectivity between servers isn’t stable.”
As for the IETF’s version of MoQ, Lietz acknowledges that would enable extension of real-time connectivity to millions of people over multiple CDNs. While the nanocosmos version works “very well” with the limited user concurrency targeted by its customers, IETF’s MoQ could “enable certain benefits to our users,” which, though he didn’t spell them out, would include multidirectional streaming. “It’s an option we might consider as we continue to embrace new standards, but that’s not decided,” he says.
Right now, he adds, it’s important to understand that demand for real-time streaming doesn’t require a one-solution-fits-all approach. The niche targeted by nanocosmos is vast and growing, Lietz notes, which means his company’s challenge is to provide the most cost-effective way to satisfy those needs as they continue to evolve.