Quality of Service in Integrated Voice, Video, and Data Networks

A Asce Networks White Paper

Introduction

Advances in today's technology enabling integrated provisioning of Voice, Video, and Data services offer much promise to enterprise customers in enabling them to both save money on their telecommunications costs and offer new services on their network enhancing productivity. Digital Signal Processing (DSP) technology and Voice Compression algorithms have advanced to the point that toll quality voice can be offered over these new integrated infrastructures, with no noticeable degradation of voice quality to users.

These advances, though, are useless if an effective mechanism cannot be implemented for assuring Quality of Service (QoS) for these integrated applications. Today's backbone integrated infrastructures may be Frame Relay based, ATM based, IP based, satellite based, or ISDN based, or any combination of these, and an effective mechanism for assuring QoS must work regardless of how the backbone infrastructure is implemented.

Asce Networks offers a complete, working solution for these QoS issues in today's heterogeneous networks. These mechanisms are built upon field proven techniques, which enable high quality voice, video, and data to be delivered for today's demanding business applications.

Issues Involved in Assuring QoS

There are many issues that may affect the QoS delivered for network based integrated services. Essentially, these issues arise from the fact that different traffic types require different levels of service from the network. Some examples of these are:

Packetized Voice Traffic. Packetized Voice traffic is characterized as relatively low bandwidth (typically 8 Kbps), but requiring a low latency delivery to ensure high quality audio.
Video Traffic. Video Traffic is generally higher bandwidth (128 Kbps to 384 Kbps or more), but still requiring low latency for high quality video images.
File Transfers. File transfers require high bandwidth, but can be allowed to suffer latency through the network.
E-Mail. E-mail is typically low bandwidth, and can also be allowed to suffer some latency through the network.
Legacy applications. Legacy applications such as SNA may require moderate bandwidth, with moderate latency requirements.
If one analyzes these requirements, one can see that a mechanism for ensuring QoS in a network must be capable of: identifying traffic types (even specific applications running over IP); prioritizing these traffic types; and then delivering them over the network in such a way that the QoS requirements for the service type are met.

What are some of the issues that may affect the QoS offered to these services?

1. Large packets delivered from lower priority, high bandwidth applications may affect the latency for higher priority, latency intolerant applications (such as voice). For example, a 1500 byte packet delivered as part of a file transfer over a 64Kbps link will take 187 ms to be transmitted. This means a voice packet cannot be transmitted during this interval. As a result, voice cuts or delays will be heard for voice traffic queued behind this large packet.

2. Different speed links in the network may mean that packets can get queued internally in the network. When packets queue internally in the backbone of a network, latency and therefore quality can be affected.

3. Network based IP applications may not respect the QoS policy set up for the network. In a network of hundreds of PC's, it may be impossible to adequately police QoS policies on each desktop, thus resulting in policy violations which can affect QoS.

A flexible Quality of Service mechanism must be able to handle all these traffic types without affecting the quality offered to other services. In addition, the QoS mechanisms must be designed to operate over a reasonably large set of network topologies and potential congestion conditions.

TOP

Requirements for Effective QoS

In order to effectively manage QoS, therefore, based upon the above requirements, several requirements must be met:

1. Quality of Service at Link Level must be employed.

If the user is running IP over a Frame Relay network, then if the device employed cannot ensure QoS over Frame Relay, then the QoS mechanism employed at the IP layer will not be effective. Therefore, a solution that effectively implements a QoS mechanism must be able to deliver the QoS at the Frame Relay or ATM layer.

2. Mechanisms for Identifying Traffic Types must be employed.

If the mechanism employed cannot tell if an IP packet is important (perhaps Voice) or less important (perhaps FTP), then the mechanism will not be effective.

3. Mechanisms for Implementing a QoS Policy must be employed.

If the device employed cannot police the network and ensure that a rogue user cannot override a quality policy, then QoS cannot be adequately ensured. Packets should be capable of being identified as to specified quality policy, and then enforced.

4. Mechanisms used to implement the QoS mechanism must cooperatively interact with other, third party, network elements.

If the QoS mechanism only works with a single vendors equipment, then the network may become unmanageable if third party devices are added later.

Asce Networks has implemented a QoS mechanism within its Climax- series which meets these criteria, and enables real world management of QoS in real world networks under real world conditions.

Asce Networks' Quality of Service Mechanisms

As outlined in the previous section, the Quality of Service delivered to IP based applications is only as good as the Quality of Service mechanism implemented at the link layer. Asce Networks' primary objective is to enable QoS for IP, Frame Relay, or ATM based applications, and to use link layer (Frame Relay and ATM) QoS mechanisms to ensure IP based applications receive their requested QoS. This is an extremely key point, and the foundation upon which Asce Networks' QoS mechanisms are built. For example, since ATM natively provides the ability to offer different QoS for different Virtual Circuits, an effective QoS mechanism will permit higher priority IP traffic to be sent over higher priority, traffic tuned PVC's, while lower priority IP traffic is sent over a different PVC with different traffic parameters. The advantage of this approach, that is, using link layer QoS mechanisms to ensure IP QoS means that IP can take advantage of the sophisticated QoS mechanisms built in to ATM to assure QoS for IP applications.

Frame Relay based Quality of Service Mechanisms

Asce Networks has implemented a sophisticated, field proven mechanism for ensuring QoS over Frame Relay networks. The mechanism is built upon four basic foundations:

1. Fragmentation.

As outlined previously, larger low priority packets when transmitted can delay smaller, high priority packets. Fragmentation ensures that the larger packet is broken into smaller pieces to make the delay not perceptible. However, excessive fragmentation may result in poor data throughput. It is important to ask your vendor how much fragmentation is required to assure no impact to higher priority applications.

2. Prioritization.

Asce Networks implements four priority queues internally which can be used to sort traffic. Voice and Video traffic, for example, may use the highest priority queue, while FTP traffic may use the lowest queues. Users may decide, based upon their desired QoS policy, what traffic goes into what priority queue. In addition, the user may control the relative priority between these queues in order to fine tune the desired QoS policy.

3. Transmission Scheduling.

This is an important mechanism which is not implemented by many vendors. Packets to be transmitted over a link are not transmitted based upon an "as soon as possible" mechanism, but are transmitted on a "as soon as the remote side can receive it" mechanism. This mechanism is designed to avoid queuing in the backbone network which can cause impact to high priority traffic. This is a subtle issue, but quite important once the network is operational in a real world situation.

4. Congestion Management.

The Frame Relay congestion management mechanisms, such as Discard Eligible (DE), Backward Explicit Congestion Notification (BECN), and Forwards Explicit Congestion Notification (FECN) are implemented in such a way that in times of network congestion or transmission in excess of the Committed Information Rate (CIR), lower priority packets are tagged to be discarded if necessary by the network rather than the higher priority voice packets.

Asce Networks employs techniques that enable a user to specify which protocol uses which priority queue, which means that he can effectively tailor the mechanisms to his desired quality policy.

These four mechanisms, taken together, form a sound foundation for delivery of QoS at the Frame Relay level. As you will see later, they also support the delivery of QoS at the IP level. In addition, they are interoperable with public or private Frame Relay backbones, which means that they will work even if the backbone of the network is not based upon Asce Networks equipment.

TOP

Quality of Service for H.320 Video over Frame Relay

When utilizing H.320 Video traffic over a Frame Relay network, the above QoS mechanisms are used to ensure that the video traffic is treated appropriately. Asce Networks' VideoFramer converts the H.320 video stream into a Frame Relay stream, which can then be internally prioritized as above to ensure that it is treated as high priority traffic.

ATM Based Quality of Service Mechanisms

For ATM based networks, Asce Networks has an equally strong set of mechanisms to support QoS. These mechanisms take advantage of some of the inherent mechanisms that ATM has to support QoS, but extend beyond them as well.

Fundamentally, ATM enables Virtual Circuits to be defined with a set of traffic parameters that control the flow of data through the ATM network. Once a set of these parameters has been established for a virtual circuit, these parameters can essentially be guaranteed through the network. Thus, a strong building block exists for delivery of Quality of Service in ATM network.

Asce Networks implements several standard ATM traffic types (such as Constant Bit Rate (CBR), Variable Bit Rate (VBR), Variable Bit Rate Real Time (VBR-rt), and Unspecified Bit Rate (UBR). These traffic types enable a user to match a circuit with the desired type of traffic that flows inside the circuit. For example, since video traffic is generally constant rate, a CBR circuit can be defined to carry the video traffic. LAN traffic may use a UBR circuit to take whatever bandwidth is left after higher priority applications transmit.

In addition to this, Asce Networks implements eight (8) priority levels for these circuits. These priority levels are used to prioritize the transmission of information for different circuit types. This enables a user to have a high degree of control over how QoS can be effectively implemented in an ATM network.

Again, Asce Networks' mechanisms are flexible such that a user, based upon his desired QoS policy, can control which protocols and traffic types on his network get assigned which priority and circuit type. In addition, they are interoperable and work with mechanisms employed on third party ATM backbone switches to ensure they work over real networks.

IP Based Quality of Service Mechanisms

Asce Networks has designed and implemented a quite sophisticated mechanism for delivery of QoS for IP traffic. The mechanism, as described before, builds upon the QoS mechanisms implemented for Frame Relay and ATM, and is also standards based enabling it to work with third party, standards based routers. The mechanism is based upon the IETF Differentiated Services RFC (diff-serv), which uses the IP header to signal the priority of the IP packet between routers. Asce Networks uses the Type of Service bits (TOS bits) to enable 8 different priorities of IP traffic to be recognized and handled appropriately. The diff-serv RFC, while specifying exactly how to signal priority, does not discuss how each router should ensure the priority of the traffic. Asce Networks implements mechanisms for routing and tagging packets, as well as mechanisms for ensuring that they get treated appropriately.

Asce Networks' IP based QoS mechanisms work in a three-step process. First, packets incoming into the box are identified by their IP header as to priority. Next, according to the QoS policy configured, these priorities may be overridden or set in the IP header. These policies may set the priority based upon traffic type (e.g. FTP traffic, e-mail traffic, Web traffic), source IP address, destination IP address, source port number, destination port number, etc. Thus, completely customized policies may be set up for a network. Finally, the packet is routed and the link layer QoS mechanisms are used to actually implement the quality policy. As an example, if a packet is determined to be a high priority IP traffic, it is sent to the high priority Frame Relay queue as described before, where the combination of Frame Relay prioritization, fragmentation, and transmission scheduling are used to ensure that is appropriately delivered through the network. The user can decide which ATM or Frame Relay priorities are appropriate for which IP packet priorities.

Asce Networks has further extended these mechanisms by enabling different virtual circuits to handle these different priority IP packets. For example, if a user has two ATM virtual circuits, one with a CBR service and one with a UBR service, the user can configure high priority IP packets (perhaps an IP video session) to use the CBR circuit, while all other IP traffic uses the UBR circuit. Thus, very fine control of QoS for IP can be attained through the backbone, built upon the ATM QoS mechanisms.

Quality of Service for H.323 Traffic

Voice over IP (VoIP) poses some special issues for control of Quality of Service. Due to the nature in which H.323 passes voice traffic, it can be difficult to design a basic policy that can assure all H.323 traffic is treated appropriately. Asce Networks has designed mechanisms that enable this H.323 traffic to be handled with effective QoS through the network.

Asce Networks has the ability to internally gateway PSTN traffic to H.323 traffic, as well as to route external VoIP terminal traffic. Each of these needs to be analyzed separately for QoS.

For internally generated VoIP traffic, Asce Networks internally recognizes this as VoIP voice traffic. Since all the traffic is originating from within the Asce Networks unit, this is relatively easy to do. Once identified as voice traffic, it is treated as high priority, and transmitted on the high priority Frame Relay or ATM queues.

For external H.323 traffic, the problem is a bit more complicated. Since it can be hard to identify this traffic, Asce Networks has chosen a different mechanism to solve the problem. The mechanism is called "proxy H.323", and it enables LAN based H.323 terminals to communicate with remote H.323 terminals by utilizing the Asce Networks unit as an intermediary.

Using this mechanism, which is transparent to the end terminals, QoS can be controlled, since the proxied session will then appear to us as an internally generated H.323 session. There are additional benefits which result form this technique including security and bandwidth management, but these are not the subject of this paper.

In the example above, the LAN based VoIP terminal places a call to the remote phone. Instead of the connection going end to end directly, the call is effectively terminated and regenerated again by the local unit. By doing this, it is known to be VoIP traffic and treated as high priority.

Conclusion

Assurance of Quality of Service is critical for proper operation of an integrated voice, video, and data network. The evolution of IP based applications place more stress and require more sophistication in equipment designed to support these applications over real world networks, while delivering services at similar reliability levels to those experienced over non-integrated traditional networks. The ability to have a flexible mechanism that enables a user to tailor the QoS policy to his specific needs is a critical component of an overall integrated network.