See Also:
Today, another new technology is poised to impact business computing in an equally dramatic way. Networked multimedia computer applications will significantly affect users and network managers and have a tremendous impact on computing and network infrastructures.
This paper discusses the structure of the industry that is delivering multimedia, the requirements that multimedia places on a network, and Cisco's products that will enable multimedia both today and in the future.
Figure 1: Multimedia Industry Structure
Table 1: Types of Multimedia Applications
--------------------------------------------------------------------
STORED DATA STREAMS REAL-TIME INTERACTIVE
====================================================================
Point-to-Point Multimedia mail Distance learning
Multimedia notes kiosks
--------------------------------------------------------------------
Multipoint LAN TV Desktop conferencing
Financial Live broadcast
broadcasts
--------------------------------------------------------------------
The simplest type of application involves adding a multimedia data object to
a traditional data object (multimedia e-mail or multimedia notes, for
example). This type of application uses existing point-to-point networking
technology and is not interactive in real time.
A second application category involves sending multimedia data objects to multiple hosts, such as with LAN TV or other forms of broadcast. This category needs to go to multiple locations but does not tend to require much real-time interaction.
A third category is point-to-point applications that may be real-time interactive. Many vertical industries are planning multimedia kiosks, for example. Some examples are the following:
A fourth category is applications that are both multipoint and real-time interactive. The primary example of this category of traffic is deskto videoconferencing. Desktop videoconferencing requires real-time interactive communication among groups of individuals who may not be at the same location.
Today's multimedia applications have a wide range of bandwidth requirements and a corresponding wide range of price points. Network administrators have the luxury of choosing the quality they are willing to pay for.
The bandwidth that is required is most strongly affected by the size of the frame, the frame rate, and the quality of the image. All of these factors are affected by the video encoding scheme that is used. Video encoding schemes can be hardware or software based. High-speed processors will generally use software techniques. Older processors will tend to use add-on cards that include video encoding hardware.
Many useful multimedia applications are possible at reasonable data rates. Among them are the following:
Figure 2: Campus Bandwidth Usage Pattern
There are three different multimedia environments where more bandwidth is frequently required. Individual desktops running multimedia applications often share a LAN segment with many other users. The most cost-effective way to add bandwidth for these users is to microsegment the LAN by putting fewer users on each segment.
This technique is very cost effective. LANs can be segmented gradually as applications usage patterns require. Most multimedia applications today will run very well in these e nvironments.
High-speed servers may require other techniques such as CDDI, 100-Mbps Ethernet or, possibly, 25-Mbps ATM. CDDI and 100-Mbps Ethernet are available now and provide a dedicated 100 Mbps. Two standards for 25-Mbps ATM are currently being discussed in various standards bodies.
The second environment that will need more bandwidth because of multimedia applications is the workgroup or campus backbone.
Today's backbones frequently use FDDI. FDDI provides 100 Mbps and is used today for high-speed servers and as a backbone. Early adopters are beginning to install ATM networks as a mechanism for increasing the amount of bandwidth that is available in these places.
The third environment that will need more bandwidth is the wide-area network (WAN). Bandwidth in the WAN is typically expensive and is often a recurring cost. Because of this factor, WAN bandwidth needs to be managed carefully.
Cisco's Multimedia Bandwidth Solutions
Cisco provides a large range of solutions for providing network bandwidth.
At the desktop, Cisco supports LAN switching with the Catalyst(tm) Workgroup Switch and Kalpana EtherSwitches.
At the workgroup level, Cisco supports both FDDI and ATM. The Cisco 4500 and 7000 family routers have FDDI cards. Cisco provides two ATM switches, the LightStream 100 for workgroup connectivity and the LightStream 2020 for enterprise ATM connectivity.
In the WAN, Cisco supports a comprehensive set of technologies that include dedicated lines, Frame Relay, X.25, SMDS, ATM, and ISDN.
Cisco's bandwidth-on-demand techniques are particularly valuable for WAN multimedia networking. Bandwidth on demand enables a router to bring in additional bandwidth when the traffic requires. This is a very cost-effective way of providing extra bandwidth only when it is needed.
The next step is to move voice and video applications onto this network. Today it is common for a business to have a data network, a telephone network, and, in some cases, a videoconferencing network. This is expensive. There is a strong economic push to integrate all of these applications onto a single network.
However, these different types of applications have different service requirements. The success of networking in the next five years will be driven by finding an economic way to extend the integrated services network; one that can handle data applications, voice applications, and video applications.
Networking application traffic today can be divided into three categories: Constant Bit Rate (CBR), Variable Bit Rate (VBR) and Available Bit Rate (ABR). In addition, some networking applications have strict limits on the amount of latency and jitter that they can tolerate.
Constant Bit Rate
Historically, voice and old video codecs produced a constant stream of bits. These applications are described as constant bit rate (CBR) applications because they require a specific minimum amount of bandwidth. With less than this amount, they cannot function; with extra bandwidth, they receive no benefit. (See Figure 3.)
Figure 3: Constant Bit Rate Service
Traditionally these applications have run in circuit-switched environments where they have fixed, dedicated bandwidth.
Variable Bit Rate
Today, LAN TV and modern video codecs produce a variable stream of bits. This stream is similar in nature to traditional interactive data applications. The stream is bursty; that is, sometimes the bandwidth required is low and other times it is high. These applications are called variable bit rate (VBR) applications.
Circuit-switched networks are engineered to provide enough bandwidth in each circuit or virtual circuit to handle the peak rate required by the VBR application. When the VBR traffic is below the peak rate, the extra bandwidth is unused.
Packet-switched networks are engineered to provide enough bandwidth to handle the average rate required by the set of VBR applications that are running. The peaks are handled by statistical sharing of the extra bandwidth. This technique is known as predictive quality of service. (See Figure 4.)
Figure 4: Variable Bit Rate Service
Available Bit Rate
Traditional data applications such as file transfers and new multimedia applications such as multimedia mail and notes can function with a wide range of available bandwidths. They require little bandwidth to function slowly, and they run faster as they have access to more bandwidth.
Traditional packet data networks are very good at supporting available bit rate applications with best-effort quality-of-service guarantees. (See Figure 5.)
Figure 5: Available Bit Rate Service
Latency
Real-time, interactive applications such as desktop conferencing are sensitive to accumulated delay (latency). For example, telephone networks are engineered to provide less than 400 milliseconds round-trip latency.
Multimedia networks that support desktop audio/video conferencing also need to be engineered with a latency budget that is less than 400 ms round-trip.
The round-trip latency budget is consumed by the sending computer, the network, and the receiving computer. As a rule of thumb, the sending computer will take a few milliseconds to send a packet.
The network contributes to latency in several ways, including propagation delay, transmission delay, store-and-forward delay, and processing delay.
Most internetworking devices receive a packet before sending it out on another interface. The amount of delay that is introduced depends on the size of the packet and the speed of the media.
There is a great deal of difference in the amount of processing delay that is introduced by different networking vendors. Scott Bradner of Harvard University periodically conducts latency tests on submitted network equipment. For example, in August 1994, Bradner's tests showed that the Cisco 7000 introduced less than 50 microseconds of delay for 64-byte IP packets. (For the latest Bradner numbers for Cisco networking gear, contact Cisco in the U.S. at (800) 553-NETS (6387) or outside of the U.S. at 408 526-4000.)
Voice communication requires low latency in order to maintain audio quality. Telephone networks are typically engineered to keep end-to-end latency below 400 ms. Data networks that carry voice traffic will need to be engineered similarly.
Jitter
When a network provides variable latency for different packets, it introduces jitter. Jitter is particularly bad for audio streams, because it can cause audible pops and clicks that can be disruptive to communications.
Multimedia networks must provide techniques that minimize jitter for traffic that is adversely affected by it. The next sections of this paper describe techniques available in the Cisco Internetwork Operating System (Cisco IOS) to minimize jitter.
Cisco's Multimedia QOS Guarantee Solutions
The Cisco IOS(tm) includes several sophisticated queuing
algorithms that lower latency and minimize jitter.
In addition, applications frequently provide techniques that minimize jitter. The most common technique is to have the network place the data into a buffer that the display software or hardware pulls data from. The insulating buffer reduces the effect of jitter the same way that a shock absorber reduces the effect of road irregularities on a car; variations on the input side are smaller than the total buffer size and are therefore not observable on the output side.
Integrated Services
It is not difficult to build a network that can support any of the different classes of traffic: constant bit rate, available bit rate, and variable bit rate. It is more difficult to build a network that can economically support all of the classes of traffic with the quality of service that they require.
There are two technologies under development that can be used to extend an integrated services network.
ATM
One technology for extending integrated services networks is to use ATM.
This solution holds the promise of guaranteed quality of service for
different classes of traffic.
There are three problems with using ATM for quality-of-service guarantees today:
Resource Reservation Protocol (RSVP)
Another technology for building integrated services networks is to use the
Resource Reservation Protocol (RSVP). RSVP is an internetwork end-to-end
protocol that reserves the necessary resources for the different classes of
service by using the techniques available on each underyling network type.
For example, an application signals the network about the class of service it requires. The network reserves the resources end to end using FDDI techniques on FDDI, ATM techniques on ATM, Frame Relay techniques on Frame Relay, etc.
RSVP is a standard protocol that is being defined by the Internet Engineering Task Force (IETF). RSVP will be able to support multiple network layer protocols such as TCP/IP, Novell IPX, and AppleTalk.
Cisco's Integrated Services Solutions
Cisco is working with the ATM forum to define the standards that will allow ATM to achieve its potential to build integrated services networks. In addition, Cisco is working with the Internet Engineering Task Force and with multimedia application developers to bring RSVP to the market quickly. RSVP will extend the quality-of-service guarantees that are currently provided by the Cisco IOS by doing end-to-end resource reservation and by taking advantage of ATM techniques where available.
There are three ways to provide multipoint communications:
Multicast is a very efficient technology. It is easy on both the hosts and the networks. However, in order for multicast to work, the networking devices need to know which computers need to receive multicast traffic, and they need to be able to dynamically build efficient paths to all destinations.
Cisco has worked with the Internet Engineering Task Force to develop a set of protocols that enable multicast routing.
The first protocol is RFC 1112 -- the Internet Group Membership Protocol (IGMP). This protocol allows a computer to inform the network that it is a member of a specified group.
The second protocol is currently an Internet Draft -- Protocol-Independent Multicast (PIM). PIM allows the networking devices to build efficient distribution trees for multicast packets. PIM is expected to become an RFC in the first quarter of 1995.
IP Multicast has been used to build the MBONE, an experimental multicast backbone that runs on top of the public Internet. It is used to carry the general sessions of the Internet Engineering Task Force and other technical forums that have worldwide appeal. It is also used today to do collaborative work by researchers and engineers who need a rich communication infrastructure.
Many host and application vendors support IP Multicast:
Cisco's Multicast Routing Solutions
Cisco introduced support for IP Multicast in IOS Release 10.2, which began shipping in October 1994. Cisco also supports IP Multicast in Catalyst Version 3.0, which began shipping in November 1994. Cisco plans to add support to Kalpana EtherSwitches in a future release.
Cisco will continue to actively work with the IETF to promote the development of scalable IP Multicast standards.
Cisco enables the deployment of multimedia networking today by meeting the following requirements: