Can an IP router replace a video router?
With telco entering in the quadruple play of video, Internet, phone and cell phone, it seems that video over IP is part of a video future. The requirements to deliver acceptable video and audio signals via IP are different from traditional voice and data delivery methods, bringing new challenges. IP is not new to broadcast facilities, as many files get transferred between locations, servers and workstations. However, most of that transfer is not in real time.
As manufacturers develop solutions to support the new IPTV deployment, traditional broadcasters are looking at using some of the IP technology to improve their real-time workflow. One question comes to mind: Can broadcasters replace a traditional video router with a standard IP router?
In theory, any type of data stream can be encapsulated into IP. In most cases, it is a good idea to limit the maximum size of the IP packet to 1500 bytes. The overhead for the encapsulating is small. It's about 40 bytes for 1500 bytes total, which is less than 3 percent. Today's networks commonly run at 1Gb/s (950Mb/s useable), even though there are still a lot of 100Mb/s networks inside existing facilities. The future will bring a 10Gb/s plus backbone and beyond, but this will not be widely spread outside of the large telco companies for some time. Broadcasters can encapsulate hundreds of video streams of H.264 or MPEG-2 SD distribution rate content on a 1GigE connection. There is no defined standard for encapsulating uncompressed SDI, HD or 3Gb/s video in IP packets. So at this time, carrying uncompressed HD-SDI or 3Gb/s around the plant over IP is possible, but not practical. (See Table 1.)
IP networks are different from traditional video networks. (See Figure 1.) IP networks are switched networks, which — with inconsistent signal paths — make it challenging to troubleshoot. The classic example is VOD, which by definition is only active while the user orders and watches the movie. It is therefore difficult to “follow the wire” from point to point.
Delivering packets from the source to the destination is a well-known and controlled process (ATM, SONET). Timing and packet order has always been a challenge but is not critical in the case of e-mail and data. Most IP routers have a buffer built in but not much compared with high-speed video routers.
When delay happens in the network, it can result in underflow and overflow of the buffer at the switch. When data gets lost, freezeframe occurs, or there is tiling on the video output.
The media delivery index (MDI), which measures network jitter and drop of IP packets, is a composite number proposed by the Internet Engineering Task Force (IETF) and adopted by multiple test equipment vendors. It offers a simple but accurate way of measuring the network delivery quality in IP probes.
As IP becomes mainstream, broadcasters are wondering if they could use the momentum of the IP networking to improve their real-time workflow.
The attraction comes from the fact that an IP router is perceived to be low-cost, can handle all kinds of data in real time or not real time and seems to be independent of the codec bit rate. The concept is that if you can equip yourself with an IP router, you should be able to encapsulate any signal available now and in the future.
So the question that broadcasters are asking to the industry is: Can I use an IP router for my core video router?
In a video router, the I/O count is key. A typical video router has a large number of input ports and output ports. Each port can carry one video signal SD, HD, 3Gb/s or compressed ASI in the case of an advanced video router. The port count of a video router can now go to 1024 inputs and 1024 outputs in large facilities.
In an IP router, each port is bidirectional and can support multiple video signals. That is a great advantage to the IP router, as multiple SD video sources can flow in both directions, whereas only one video source can travel on a video router I/O. Typical IP routers have a core up to 48 ports, a low number in the traditional video world. Multiple 48-port line cards can be combined in one chassis. The larger one-chassis design in the IP world is 480 ports. Note as most routers today are limited to GigE ports running at 950Mb/s, it is therefore impossible to switch real-time uncompressed HD-SDI and 3Gb/s because these won't “fit” in a GigE. 10GigE ports are becoming available, but they don't yet have the density of video routers.
The switching fabric of a video router can route any input to multiple outputs. There is no limitation of input load or how many points the same video can be routed.
Video routing in the IP domain is different in concept. In a low-end IP router, the routing is performed in software, allowing flexible routing. This often creates processing conflicts when a large volume of switching is required in real time. New, advanced IP routers offer hardware routing, which improves performance when high bit rate and large numbers of streams are switched.
Video router switching times are typically in the order of one frame, but most importantly, they are deterministic. IP routers can delay routing requests because of traffic or multiple requests at the same time.
Video router controls are fairly simple and unsophisticated. A typical router takes RS-232 commands with X-Y coordinate for input and output and responds to a simple command. More modern routers support IP controls with SNMP commands, but the principals remain the same. All the control is handled out of band of the video using a dedicated port. This method requires the user of the content to have access to the router control, which is typical under automation.
IP routers act differently because most of the control, when the routing rules have been set up, is handled in-band. Multicast and IGMP allow downstream devices to request, through the data port input signal, to the video switch without accessing the core control of the IP router. On one physical port, multiple video can be requested and routed. Note that the time to get the routing execution depends on the IP router configuration. This can take from as few as 100ms to 30 seconds, depending on the settings of the downstream devices and switches. This capability makes the IP router attractive and powerful, but requires the proper use of advanced settings.
Great care is taken in a design of a video router to ensure reliability. Redundant I/O, crosspoint and switching logic is typically a high selling point for router manufacturers. This redundant architecture results in better than 99.999 percent of availability. Downtime doesn't exist, and there is no maintenance window readily available.
IP routers are reliable today. It is, however, difficult to manage load balance and guarantee there won't be downtime in IP ports. It is not unusual to have a two-second redundant switching time. This is not a long time in the data world, but definitely represents many lost video frames. In normal circumstances, an IP router is expected to be taken offline for maintenance windows.
Modern, high-video I/O port, high-bandwidth video routers are considered expensive, and they are typically about 645 per port.
Even if a consumer eight-port GigE switch runs about 16 per port, a large, level-three broadcast-suitable IP router that can support the proper bandwidth will cost about the same as a video router.
Another popular video over IP application is for CCTV applications in a broadcast plant. Most broadcasters today have an RF distribution plan for monitoring internal channels as well as news channels. Each facility often has analog RF channels with a coaxial cable distribution reaching the required offices. At each office, a monitor is used to watch the program. This RF internal channel system is typically analog and doesn't support HD or the increasing requirements for a larger number of channels.
The trend is to replace this aging system with an IP network. The video quality requirement is typically lower than broadcast and is used to deliver the monitored content. The user can employ a standard desktop software decoder.
Using video over IP in the broadcast facility requires a different set of skills to configure and appropriately manage a large IP router. The video/IT engineer needs to understand and be fluent in DVMRP, PIM, MOSPF protocol, Rendezvous, or Flood and Prune protocol. Many of those protocols are completely foreign to video engineers and require advanced training.
Because IP routers are technically and economically viable for monitoring purposes, it is clear that they are not yet ready to replace video routers in large facilities. As 10GigE and 100GigE networks make their way to the market and technology advances, it is possible to imagine a complete video routed network based on an IP infrastructure. By then, the IT department will have more staff than the video engineering department, and every video engineer will be IT-certified.
Stephane Billat is product manager for Evertz Microsystems.