Working as the network engineer of an ISP, I have a unique perspective on the trend towards the mainstream use of streaming video services which has been ongoing for the last few years. The appeal of such services is very clear to me, both as someone familiar with network infrastructure and as an end user (I frequently stream iTunes and Netflix content), but I’m also keenly aware of the stress streaming video is putting on ISPs and more recently even large transit providers. I’d like to see consumer broadband network capacity grow rapidly to accommodate the specific increased demand for streaming video, but it seems to me that the typical predictions and expectations currently expressed by many pundits and commentators regarding the future role of streaming over IP networks and the supposed radical effect it will have on traditional broadcast distribution are straying increasingly far from technical reality.

Distribution of TV and movies over IP networks has now clearly broken out of the realm of geeks and early adopters. There are now a number of mature streaming set-top boxes in production which have helped spur the mainstream use of services such as iTunes, Amazon VoD, and Netflix. The original Apple TV has been around since 2007, and was recently replaced by a second generation device, much cheaper and oriented even more towards pure streaming than the first. Total sales of the first-gen Apple TV were estimated at between 8 and 10 million as of mid-2010. Apple recently announced that they’d also sold a million second-gen units in the first three months of availability. Although Apple TV is the most well known and by a fair margin the best selling device in the streaming set-top box category, there are now several established competitors. Roku makes a line of streaming STBs with similar functionality which have sold roughly a million units in the US, and there are other platforms in the market that have yet to gain much traction as standalone devices but which have strong backing and may end up widely integrated into future TVs or provider STBs in the future, such as GoogleTV and Boxee. Aside from dedicated STBs, streaming video on PCs and mobile devices is increasingly easy and popular. There are also many Cable and Satellite VoD customers who may not even realize that they’re streaming video over their broadband connection using their provider’s set-top box, but nevertheless do so regularly. Although Cable and Satellite providers are in most cases just beginning to dabble in IP network delivery of their VoD offerings, there are now numerous third-party streaming services which have become quite popular. Apple iTunes movie/TV purchases and rentals have been around the longest, but now there’s also Netflix and Hulu, Playstation Network and Xbox Live, Amazon VoD, YouTube, Podcast networks, etc.

In particular, the emergence of flat-rate subscription-based Netflix and Hulu have seemingly provided a glimpse into a future in which many people might be satisfied with only such services and could thus drop, or at least greatly curtail, their cable or satellite service. This is the mantra I hear repeated so often now – many folks are considering getting rid of their broadcast TV provider because most of the programming provided is a waste for them – unwanted and unwatched. Better to get Hulu+ and stream only what shows you’re interested in. Optical discs are also relics of the past, so the theory usually goes – forget about Blu-rays and DVDs…just stream rentals from iTunes or watch Netflix. There is usually some discussion of the caveats – online selection is not as complete as with traditional distribution channels and often new releases become available only some time after they’ve been released to DVD and traditional VoD/PPV. Sometimes, there’s a brief discussion of the bandwidth requirements for all of this streaming, but it’s rarely in-depth or particularly accurate. All too often there’s simply the assumption that the bandwidth necessary to support all of this new media consumption via IP streaming is a given – it must be there as a basic prerequisite, but it’s easy enough to get and streaming works fine for everyone with a consumer broadband (usually DSL or Cable) connection, right? Wrong.

Popular streaming services are at this point putting significant strain on Internet infrastructure during the timeframe of heaviest usage – roughly 6 to 11pm, local time – the “evening surge”. At Indra’s Net, our residential userbase typically requests aggregate ingress traffic at levels 75-100% higher than our normal midday peaks (which is more than we can afford to provide). Almost all ISPs which provide residential broadband are now experiencing their highest daily traffic spike within that evening window, at a level which deviates too far from normal daily traffic curves to be sustainable. Even larger tier-1 transit providers are feeling the pain, as the evening surge rolls across the US every night. It’s been recently estimated that Netflix alone is responsible for some 20% of total downstream Internet traffic during the evening streaming surge. Despite continually increasing demand, downstream traffic levels in the evening have begun to flatten out, indicating just how far things have gone – last mile circuit providers and end-user ISPs are simply saturated in the evening. The same basic situation is seen across most providers, across the US, and in many other parts of the developed world now as well. In my next posting, I’ll detail the reasons why this traffic pinch can’t simply be fixed by infrastructure upgrades, as others have in the past, and explain the fundamental problem of converting a significant amount of broadcast TV demand into unicast TV demand.