In part 1 of this series we talked about why video traffic is increasing on networks. In part 2 we focused on some key questions to ask in order to get your arms around video apps and their impact on your network. We also left you with a short poll on types of video applications currently being used on your networks. In this last segment, we will look at some specific things you can do in your network to prepare it to be "Video Ready".
Video Apps on Your Network
First, the poll results (keep in mind multiple voting was allowed, so many of you have several different types). Let's group the results so we can talk about them later in the article.
Group A: about 1 in 5 are using Streaming Internet Video, Desktop Video Conferencing, and Room-to-Room Conferencing
Group B: about 1 in 10 are using IP Video Surveillance, Web 2.0 Intranet Embedded Video Clips, and Video-on-Demand (VoD)
Group C: about 1 in 20 are using Digital Signage, Re-broadcasting feeds, Webcams, Personal video clips, and Telepresence
These results are pretty much along the lines we see with prevailing video trends in the networking industry. "Group A" above are the most established video applications, a mixture of streaming video and real-time interactive video. "Group B" represent the "next wave" of streaming video applications as new convergence opportunities are attractive (e.g. IP Video Surveillance) and as Web 2.0 transforms the internet. "Group C" are video applications that are still in early adoption, but quickly growing, again a mixture of streaming (Signage, Personal) and real-time interactive (Telepresence) video.
What is Your Video Strategy? In the last segment, we also posed a question to ask yourself as a network administrator or network architect: what will be your strategy for video? We see different approaches from the companies we work with. Some embrace video entirely, seeing it as driving the next wave of productivity for their business. Others adopt a stance to manage and protect select video applications on the network. Still others are in a bit of denial, and would rather pretend video isn't there yet. What should you do? If we have learned anything from past productivity enabling technology "waves", it is this: if IT doesn't deploy it, users will....and usually poorly. Think about Wireless LAN several years ago. Some IT departments were skeptical of the need, or questioned (rightly so) security, so deployments lagged. Users responded by purchasing their own consumer-grade WLAN AP's and plugging them into your network. Ouch.... Nothing can put a damper on your network security strategy like a bunch of "rogue" AP's in your network, without proper WLAN security...and often without the ability for you to locate them to shut them off. We believe the video app wave will be no different. Its already happening. If you don't embrace an IT-led deployment of video clip production and distribution (think MySpace replaces your employee directory), an ad-hoc server is liable to show up under someone's desk. If you don't embrace an IT-led deployment of desktop video conferencing (think personal High-Def video), people will buy their own consumer-grade webcams and try to do it. We also believe networks are at a crossroads: they will either evolve to be a "Video Ready Network" or network administrators may find themselves in the Wild West, with 50, 60 or 70% of their network traffic being un-managed video apps. Our point is: pick a strategy whether its fully embrace, partially embrace, or strictly control. Having no strategy is not much of an option. What Can You Do to Make Your Network "Video Ready"? After you understand the behavior of the different video applications in your network (or that may be on your network someday soon), draw common threads of requirements between them. Here are a few common ones: 1. High Availability Data apps are tolerant of multi-second interruptions, VoIP and Video apps are not. If you have already been beefing up your network with VoIP convergence in mind, excellent, you are a step ahead. If not, then it's really really time. For VoIP and especially real-time interactive video apps, the user experience matters and to achieve it, the network needs to deliver very low latency (100-150ms end to end), very low jitter (0-10ms) and increasingly, low loss. Loss requires a bit more explanation and may warrant a dedicated article to do it justice. But to make it simple, HDTV video formats take billions of bytes to transmit and they are not practical without compression codecs like MPEG4 or H.264. If you compress the equivalent of several thousand packets into one packet, and you lose that packet, you effectively lose thousands of packets. For data apps we were sometimes happy with 1-2% loss. For VoIP we tightened our expectations to say 0.5-1%. For video (especially High-Def) think order of magnitude reduction, 0-0.05% loss. In short, build your network with built-in low-latency, sub-second failure convergence, and as close to zero loss as possible. 2. Quality of Service (QoS) Again, QoS is not new as hopefully you have applied it to your network to protect your existing critical applications. If you already have QoS in place for VoIP, fantastic, you need to extend it to include Video apps. If you have not, then now might be a really good time to consider doing so. You may say "But my apps have worked fine, why do I need QoS now?" The answer is that video consumes bandwidth...lots of bandwidth, far more than VoIP. If you added VoIP to your network without QoS, and it worked, you may have gotten lucky that a bandwidth hungry app didn't chew up the bandwidth you needed for high quality VoIP calls. Why not just throw more bandwidth at the problem? Bandwidth alone doesn't solve the problem. Think of it like a freeway. Onramps control flow of new traffic, there are slow lanes and faster lanes, and overall speed limits. All these controls work together to move the most amount of traffic safely through the highway. Without those controls, you are likely to have a free-for-all with major pile-ups that more lanes is unlikely to solve. Identify which applications you want to protect in your business, whether they are data, VoIP, and/or Video, and implement a comprehensive QoS service policy to mark and prioritize traffic. 3. Bandwidth and Latency As we said, video apps need bandwidth, and also relatively low latency packet delivery. In a campus switching network, make sure you have enough bandwidth at the aggregation points and through the core (such as 10GigE). Don't forget about your branch offices. Many, many people in the typical large company now work in satellite or branch offices away from the main headquarters. These folks will expect no less than the same set of video-enabled apps as your HQ employees. In fact, they may rely on them even MORE because of the need to communicate effectively with corporate. Survey your WAN speeds and make decisions now about which branch offices need to be upgraded to higher speed and/or secondary WAN connections. Some quick calculations based on the number of seats in a branch office can give you a quick indicator about bandwidth needs. For example, suppose there are 20 people in a branch office and your company relies on desktop video conferencing for collaboration, streaming video for training and corporate communcations broadcasts, and plans to install IP video surveillance cameras at all branches for security. Lets further assume a 5:1 over-subscription on conferencing. desktop video = 4 simultaneous calls to HQ x 512kbps training VoDs = 2 simultaneous x 384kbps video surveillance = 2 x 512kbps VoIP = 5 simultaneous calls to HQ x 128kbps data apps = 64kbps x 20
With simple estimates, we can pretty easily see that we may need 6Mbps or more for our WAN speed, and our current T1 is not going to cut it.
One thing we can do is to "harvest" bandwidth, using WAN optimization technologies such as Cisco Wide Area Application Services (WAAS), which using compression and optimization can give us back 20-50% or more of our current WAN bandwidth, without sacrificing application speed. WAAS or any other WAN Optimization technology is unlikely to save bandwidth of Video apps themselves, because of the high degree of compression already "built-in" to most video codecs. The point of implementing WAN Optimization is to "clear" bandwidth from other apps to be re-used by newer or expanding apps.
Many people ask: should I optimize my WAN or upgrade the bandwidth. The answer is: both. Optimizing the WAN will allow the most conservative upgrade path.
4. Broadcast Optimization
Several popular video apps, including streaming video, video-on-demand (VoD) training, corporate broadcast communications (e.g. IPTV) and others have a behavior model of a single or few video sources and many simultaneous viewers. Whenever you have such video apps, you want to optimize these broadcasts so that preferrably you send a single (or few) packet streams on the network that viewers can join, instead of each viewer requiring their own dedicated packet stream.
IP Multicast is a good technology that can be leveraged to optimize such video apps. If you already have IPmc enabled in your network, great, you are a step ahead of the game. If not, it may be time to consider turning it up.
Stream "splitting" is another alternative starting to appear in products aimed at the WAN and branch office networks. Effectively stream "splitting" behaves alot like IP multicast, only instead of a real multicast packet stream in the network, usually a proxy device receives the stream at the branch office over the WAN and then handles "join" requests, much like a rendevouz point in IPmc. Cisco's WAAS product is one such product that has an integrated "stream splitting" capability for certain types of video streams.
5. Visibility
It is more important than ever to understand the applications running on your network, what resources they are consuming, and how they are performing. Whether you are trying to insure a high-quality experience for video conferencing users or trying to understand how YouTube watchers may be impacting your network, its important to have visibility into the network.
Tools like Cisco NetFlow and tools like it can be essential to understanding what portion of traffic flows on your network are your critical data apps, your VoIP apps, your "managed" video apps, and the "un-managed" video (and other) apps. For example, if you discover that YouTube watchers are consuming 50% of the WAN bandwidth to your branch offices, potentially squeezing out other business critical apps, it may be time to put some usage policies into place or even more drastic measures such as network-based policing.
Another important aspect is to understand how the business managed video apps are performing? What kind of experience are users receiving? One way to proactively monitor such apps are using network-based tools such as IP Service Level Assurance (IP SLA), which can be programmed to send periodic probes through the network to measure critical performance parameters such as latency, jitter, and loss. It can be helpful to discover troublespots with long-latency times, for example, and take actions with the Service Provider (or other root cause) to correct them before users get a bad experience and complain.
Summary
Video applications are increasing exponentionally on the IP network. It is best to adopt a proactive strategy to understand how these apps will affect your network now and into the future. By taking an inventory of video-enabled apps and understanding the new and changing requirements they will place on the network, it is possible to successfully manage through this next evolution of IP convergence, and take steps to enable your network to continue to be the converged platform for your company's communications and collaborations.
