Channels; Video Free Flow – Internet/Web/Online Service Information
Streaming media has encountered some obstacles on the way to widespread adoption, but several multicasting and content providers are pushing hard to turn the video stream into a torrent
For most consumers with dial-up modem connections to the Internet, “streaming” video is probably a misnomer; “trickling” video might be more appropriate.
It’s all about barriers. The first is high-speed access at the local loop: A few have it; most do not. Until there are more broadband access haves than have-nots, streaming video isn’t really going anywhere. The second barrier – also infrastructure-related – is congestion on the backbone. Streaming files are big. Thousands or millions of them haphazardly criss-crossing the entire Internet would clog the backbone so that not even e-mail could get through. Content, the third major barrier, is irrelevant until the infrastructure barriers are removed. But once technology problems are out of the way, the question becomes, “What is there to watch?”
Breaking the local loop dam
Regarding the first barrier to streaming media success, Jeremy Schwartz, a senior analyst at Forrester Research, says, “For consumers, we’re still very much in a dial-up narrowband mode. Even with a broadband connection, streaming video is only occupying one-quarter of your screen.” This kind of niche, low- or mediocre-quality experience isn’t something people will pay for, which is why all streaming content currently is free. “As the quality of the experience goes up with greater broadband penetration, people will be willing to pay,” Schwartz adds.
But when will that penetration occur? People certainly are interested in video on their PCs. As proof, Schwartz cites the more than 100 million consumer downloads of streaming media players by Apple, Microsoft and RealNetworks.
According to Forrester, consumer interest in streaming media will drive a rapid growth in broadband access technologies. As for those carriers hoping to capitalize on that demand with DSL and cable modem offerings, they acknowledge the industry must solve not just the local loop problem, but the backbone problem, too.
Covad Communications and Jato Communications are putting special streaming media servers at the edge of their networks.
Covad also is installing its own nationwide ATM backbone (Figure 1). The result is an end-to-end approach that removes technology infrastructure obstacles for DSL users. “In two years, when [streaming media] really happens, we’ll have the best network for content and applications,” says Abhi Ingle, group product manager in charge of broadband services for Covad.
On the cable modem side, which obviously has strong residential penetration, Excite@Home’s founder and Chief Technology Officer Milo Medin says there’s been no huge demand for streaming video. “We watch the load on the network. To date, we haven’t seen lots of bandwidth driven by video. Users can go anywhere and get video, but it’s not a big deal because people don’t watch TV on their PCs,” he says.
This is perhaps one reason why Excite@Home hasn’t removed the so-called 10-minute streaming video “collar” from its contracts with subscribers. The limit applies only to content created by the cable provider or co-branded with partners such as CNN or Bloomberg; video from anywhere else has no imposed limits. Excite@Home breaks upits branded content into 10-minute labeled segments that subscribers evidently enjoy immensely. Segments of the Clinton testimony were the most popular streaming clips last year, followed by the trailer from “The Phantom Menace.”
Although streaming media hasn’t seen a huge demand, Medin doesn’t intend to get caught short. Excite@Home currently is running streaming media trials in partnership with Inktomi and RealNetworks in Nashville. The trials involve placing content-storing cache servers at the edge of the Excite@Home network. Medin doesn’t believe that it’s possible to put all available content close to users. Instead, “put the most popular content closest to the user on caching servers, which self-populate based on demand. Video servers can be further away,” he says. Assuming no problems occur in the Nashville trials, Excite@Home’s cache-server approach to streaming media will launch nationwide by mid-2000.
Flushing out the backbone
As the broadband barrier at the local loop erodes, a second, much bigger technology infrastructure obstacle looms: backbone bandwidth. The math involved doesn’t require a rocket scientist. If 10,000 – or 10 million – people are trying to dive into a 128 (or greater) kb/s video streaming feed, Internet backbone capacity can get jammed in a hurry.
The result? Massive clogs such as the one that shut down the Victoria’s Secret site last February. During a 14-minute online fashion show, the site slowed and then stopped because the servers couldn’t handle the load. When millions of people try to log onto such a site simultaneously, it also means a huge number of bits are floating across the Internet. And those bits congest the backbone, slowing and then stopping the flow of traffic.
Of course, one solution would be to dredge out a bigger Internet. However, assuming that’s not exactly around the corner, other solutions must be found.
And they have been. “There’s been a whole movement in the last six months of moving content to the edge of the network,” says Alex Benik, an analyst for The Yankee Group. “It’s the same philosophy with caching – you can distribute the stream from the local service provider’s [point of presence] rather than all across the network. The content is cached at the local POP.” It’s akin to building sites for water storage in every town along a river so that instead of constantly draining the river, thirsty users can tap local resources.
One of the best-recognized names in this new group of service providers is Akamai. The 1-year-old Internet content provider has placed 1700 servers in POPs across 100 carrier networks in 30 countries and constantly is increasing the size of its network.
Akamai divides its 1700-server network into two parts. FreeFlow hosts high-bandwidth non-streaming objects (such as graphics or downloads), whereas FreeFlow Streaming hosts streaming content. In both cases, content is re-hosted from its original Web server to Akamai servers at the edge of the Internet. Users then tap into that server on demand without ever knowing they are being detoured.
“Streams are very heavy bandwidth files,” says Jami Axelrod, Akamai’s product manager for FreeFlow Streaming. “As they traverse the Net, there is a high possibility of loss of data. The resulting quality isn’t good – there’s no synchronization of audio and video, the video will be jerky, etc. Akamai places the content closer to the end user so the file need not traverse the whole Internet. This improves the quality of the stream.”
For a live stream, Akamai serves as many streams as needed to maintain a high-quality event. One example of this is the recent episode of “The Drew Carey Show,” which was broadcast simultaneously on TV and Webcast with live Web cameras across the Internet. While the TV broadcast followed Carey, the Web cam streamed live events from his apartment while he was away.
Akamai served the content and maintained high quality for the Drew Carey event through this multistream method, called SteadyStream. “Akamai’s SteadyStream automatically sends another stream if one is degraded, so the content is delivered in its original quality,” Axelrod says. “SteadyStream can select and recombine the streams to get the best quality. The event may start with just a few streams and then increase the number if the quality is degrading.”
Via its 100-carrier network, Akamai can use the entire Internet for live streaming. “If one carrier fails, we are on 99 others, so we can reroute around the problem,” Axelrod says “We are carrier-agnostic.”
Sandpiper, which has merged with Digital Island, also has a two-segment network that re-hosts non-streaming and streaming content on servers at the Internet’s edge. The two segments are termed Footprint and Footprint Streaming.
Of the more than 1200 servers in both networks, about 110 are devoted to streaming media. Using proprietary software developed by Sandpiper for Footprint Streaming, Sun Microsystem’s Solaris machines monitor the network, track traffic, measure connectivity and determine each user’s geographic location. Once the optimum path is selected, content is delivered to the end user. According to Sandpiper, Footprint’s performance is significantly faster than typical origin servers (Figure 2).
Sandpiper uses various first-tier backbone, DSL, cable modem and ISPs to host the multicast-enabled Footprint Streaming servers and ensure distributed service. However, the company’s merger with Digital Island will give it a competitive edge, says Pat Greer, director of streaming services at Sandpiper. “Digital Island has a private backbone, so this will enhance what Sandpiper already has. It will also allow large files to be transferred without going over the public Internet.”
In business since 1995, Intervu was the first content service provider to build an intelligent distributed streaming network. Intervu’s rack-mounted Media Data Centers, co-located with top-tier ISPs and backbone carriers, are sited at Internet hubs such as Dallas, Seattle and Washington, D.C.
“In all these locations, we try to use gigabit fiber uplinks using Cisco [Systems’] 12000 or 7500 routers,” says Scott Crowder, vice president of Intervu operations. “The fiber is from the racks to the routers.” Via the routers, the entire MDC rack plugs into an Extreme Networks 7 Gb/s, non-blocking backplane switch with fiber and Ethernet interfaces (Figure 3).
The MDC servers essentially function as intelligent domain name servers, receiving content and constantly optimizing delivery by reducing the number of routers between the MDC and the subscriber. “We are doing an extreme amount of analysis on the Internet every few seconds, communicating all that between MDCs and then using it to optimally route the traffic,” Crowder says.
Intervu is distributed for the same reason anything on the Internet is: to avoid poor quality of service (QOS), to avoid ISP failures or, as Crowder puts it, to avoid a “bad day on the Internet.” However, Intervu also is building an overlay ATM signal acquisition network to provide extra redundancy and high QOS. Finally, Intervu even can pick up live satellite feeds, digitize and encode them, and deliver the resulting video stream to end users.
At UUNet, content also is being moved closer to the user, but not through special video or caching servers. Instead, standard Cisco Internet routers, upgraded for multicasting, provide streaming video to dial-up users via the UUcast service.
UUcast is based on IP multicasting, but where IP multicasting is many-to-many, UUcast is a one-to-many service. UUcast’s services also depend on moving the stream closer to the user. “When a user requests a multicast, the network routes the source to the user,” says Kevin Gatesman, UUNet’s manager of emerging technologies. “The stream travels from the source to multicast routers in the network. The multicast routers manage all the replication and the dial-up subscriptions.” For example, if one user is receiving a stream from the router, as other users sign up, they also jump into this stream instead of having new streams sent across the network.
The primary technical challenges in enabling multicasting are a lack of standards and the need to upgrade Cisco routers to support it, Gatesman says. UUNet and others have been working for two years in the Internet Engineering Task Force to establish the necessary standards, but the work still is in progress. As for upgrading Cisco routers, Gatesman says it generally requires new code releases from Cisco. Cisco comments that its most recent two-router operating system versions support streaming video but require three software commands to enable it.
More than half of the streaming video transmissions across the Internet are still unicast – in other words, sent from the source all the way across the Internet to the user. “Multicast is far more effective and efficient for a few hundred to a few hundred thousand users,” Gatesman says.
A barren stream?
With local access barriers on their way to being broken and smart work-arounds promising to alleviate the Internet’s backbone limitations, what’s left to hinder casual, everyday use of streaming video? Just one thing: content.
“People don’t adapt new technology unless there is content to drive it,” says Forrester’s Schwartz. “Content is the driver for this, just as it always has been on TV. Bonanza drove color TV, and MTV drove cable.”
Schwartz envisions an “Internet tier” as a third branch of TV. “We might get narrow, niche markets with only a few thousand viewers. Perhaps there will be tens of thousands of these types of programs.” Schwartz also sees a continuing quality gap between TV and PC-based streaming video. “By 2004, Internet TV will be where TV is now. But by then, broadcast TV will be better. The quality gap between the Web and TV will be the same.”
Excite@Home’s Medin has his own idea of the form Web-based streaming media content will take. It won’t be long-duration entertainment such as movies, it will be short pieces such as video clips and it will be interactive. “People don’t use their PC the way they use TV,” he says. “The PC is very interactive. It has a keyboard, a small screen – you’re close to it. TV is passive. It has a big screen, and it’s a group experience.”
On2.com must have been fishing from the same pier as Medin. Instead of offering movies, the content provider’s On2Movies channel will feature video clips, interviews and other bits such as wrap-arounds and intro pieces. “We will show material about movies. We won’t deliver movies. We will run a mix of short original content of about 15 minutes in length, plus other content,” says Dan Miller, On2.com’s CEO. In 2000, other On2 channels will make their maiden voyages, including On2Music, On2Travel and On2Games.
On2.com isn’t providing content for the low-echelon users of what it calls “slow-as-mud” modems. Its proprietary TrueMotion VP3 technology delivers full-motion, full-screen, video over the Internet at speeds of 250/300 kb/s and up. “The product was focused from the ground up on supporting broadband,” Miller says. “Others like RealNetworks will have to continue to support low-bandwidth users.” Interactivity will be built into the streams in the form of “hot spots” that users can click.
Two-year-old Into Networks (formerly Arepa) also is focused solely on broadband customers, but not for the delivery of entertainment. “Broadband is the wave of the future and the place where all software will be moved,” says Vincent Grosso, Into Networks CEO. “We have invented software to stream software over a pipe, which allows the software to execute in real time without installation, download or reboot. For the user, it’s like using a TV. There are no waits for downloads, no problems with disks and no conflicts with other software.”
Initially focused on gaming software, Into Networks’ business model now also includes business software. “What RealNetworks is to streaming audio and video, Into Networks is to software,” Grosso says. “So [Into Networks] can rent software. It can preview software. It can run whole channels for business, games, learning.” Into Networks’ first channel is called PlayNow, but there is also a business channel in development. The company is working with @Home in Union, N.J., to deliver content and is actively discussing partnerships with U S West, Bell Canada and other companies worldwide.
Into Networks servers sit in the central office of a local exchange carrier (LEC) or POP of an ISP, usually on the backbone side of the network through a switch to the DSL access multiplexer (Figure 4). In the case of @Home, servers are located in the regional data center. Into Networks operates and maintains its own servers. The company hoped to sign 35,000 subscribers by the end of 1999 and an additional 100,000 subscribers in the first 45 days of this year, Grosso says.
ZoomTown.com, the largest provider of interactive streaming content in the U.S., is – not coincidentally – a division of the LEC with the largest DSL penetration in the U.S.: BroadWing (formerly Cincinnati Bell). “ZoomTown is an interactive high-speed community with the highest penetration of [asymmetrical] DSL customers in the world,” says Steve Stephenson, ZoomTown.com’s project manager. With thousands of subscribers in the former Cincinnati Bell region, ZoomTown offers a whole raft of content, from local radio and TV stations, to customized local interactive fare such as a weekly three-hour talk show with Cincinnati Bengals players and announcers.
“We are aggressively pursuing additional content from outside sources for end users, such as sports, training and distance-learning initiatives with local educational institutions. We are asking providers of high-bandwidth content to come and play on our network,” Stephenson says.
Some ZoomTown content is specially encoded at 56 kb/s to make it available on the Internet. ZoomTown-only content is encoded for 300 kb/s delivery speeds across the all-Cisco network of routers, switches and giga-bit-switch routers. “Servers are whatever the customer chooses for their applications,” Stephenson says.
ZoomTown delivers more than entertainment, though. High-end e-commerce capability is available, including special procurement applications and a Cincinnati-area online business incubator. Business customers also can create their own content, and ZoomTown will host it, relieving the customers of the need for streaming video or other technical expertise.
The lure of the enterprise
Business customers have other options besides online offerings such as ZoomTown. “Traditional distributed companies like Intervu are now moving to the enterprise for on-demand and live streaming,” says Mark Zohar, a senior analyst at Forrester. Entertainment media was the low-hanging fruit. Now they’re realizing the huge market for streaming and on-demand video for corporations.”
To date, videoconferencing has been a bust in the corporate world because of its requirements for special equipment, special rooms, a lot of bandwidth and the time needed by the participants to take part in any event, Zohar says. “The reason videoconferencing has never worked is because it’s special – you have to put on a suit and tie and go to a special room,” he says. “But there’s a real sweet spot for corporate video-on-demand.” For one thing, it doesn’t require special equipment, the coat, the tie or the simultaneous meeting of lots of expensive employees. But it also can have lots of other stuff added to it, such as smart, high-quality production, content enhancement through digital publishing and editing tools and features added by content management systems such as indexing, coding and the ability to search video content.
“The enterprise will look to live streaming as well, but only for very high-profile events,” Zohar says. “The key thing for video in the enterprise is that it must be just another application on the desktop.”
Intervu, among others, has leaped into the streaming videoconferencing pool with a big splash. “Instead of an bridge, people can tune in from their desktops,” Crowder says. “We’ve done 3000 or 4000 conferences in the last two months. We can deliver education, corporate training and market updates this way, too.” At the beginning of 1999, 80% of Intervu’s revenues were coming from entertainment, Crowder says. By the end of the year, 50% of revenues were coming from business.
Naturally, there are some technical problems on the business side just as there are on the consumer side. “LAN connections are much better controlled [than the Internet], but if you use a lot of streaming video, you will quickly suck up the bandwidth, says The Yankee Group’s Benik. So the LAN infrastructure is an issue there – how will streaming affect it? It’s no coincidence that Cisco is making a big push for streaming.”
Cisco’s big push is mainly focused on its new IP/TV and IP/VC (videoconferencing) products (Figure 5). However, the company considers its standard IP platforms to be excellent for streaming video, says Heather Rose, group manager of product marketing for the video Internet services unit at Cisco. “They let you build from your standard network to include video.” The standard platforms aren’t specific to the enterprise market, but the IP/TV product is a turnkey streaming video solution for business customers. “They are intended to distribute content efficiently and effectively to a large, usually distant receiver base,” she says.
But whether for business or entertainment, it remains to be seen whether streaming video will soon break through the dams of local access, backbone congestion and content availability.
“The quality and resolution of the streams will be moving up,” Benik says. “Streaming will begin to break into the enterprise and continue to grow in the consumer market. We can’t stream movies in a year – there are rights and technical issues. But bandwidth will still be an issue at the local loop.”
COPYRIGHT 2000 PRIMEDIA Business Magazines & Media Inc. All rights reserved.
COPYRIGHT 2003 Gale Group