TABLE OF CONTENTS
Executive Summary 1
Part I Business Factors 3
What is Convergence? 3
Toll Bypass 3
Utilization ? Getting Your Money’s Worth 3
Part II Technical Overview 4
Mixed-Media Requirements 4
Problems: – Delay 4
– Packet Loss 5
– Jitter 5
Network Technology 5
Voice over ATM 5
Voice over Frame Relay 6
Voice over IP 7
Inter-Vendor Support 8
Part III Implementations 10
Types of VoIP Implementations
VoIP through a router 10 LAN Telephones 10
IP PBX 11
VoIP Gateway 12
Appendix A ? Works Cited 14
Chart 1 ? Cost of International Voice Calls 3
Graph 1 ? Long 1-Way Voice Transmission 4
Chart 2 ? Summary of H.32x Standards 8
Picture 1 ? Converged Network Architecture 9
Picture 2 ? H.323 Architecture 9
Picture 3 ? VoIP through a Router 10
Picture 4 ? LAN Telephones 11
Picture 5 ? IP PBX 11
Picture 6 ? VoIP Gateway 12
Picture 7 ? Cisco’s Consolidated Data-Voice Network 13
The universal belief, today, is that IP will become the transport for virtually all communications traffic. Yet, there are still fundamental issues involved with converging voice and data traffic onto the same medium. Many vendors and standards organizations are working on developing solutions that interoperate together. It is no longer desirable to have proprietary products that do not work outside the company walls.
Users expect a quality of service equal to that which they are already experiencing. By making use of intelligent network design, advanced routing protocols and open-industry architecture this dream can become a reality. Umbrella standards, such as H.323, spell out a model that is non-vendor specific for providing voice, video and integrated data.
Merging telephony and data will have two major benefits. The first, and most important to any businessperson, is the impact on the IT budget. Within the IT budget three areas will have savings: IT personnel, network equipment, transmission services. The IT personnel will have to be knowledgeable in both data and voice networking. Thus reducing the need for separate teams. In most cases the need for forklift upgrades has been eliminated. By simply adding hardware components and software the migration can begin.
The second, and more significant than the first, is the new applications that this makes possible. Combining voice and data onto one packet infrastructure enables new capabilities that are not possible with separate networks. Together they produce a synergistic effect that can give a company customer interaction capabilities like never seen before.
The network itself can be chosen for facilitating voice and data. The most impressive of which is voice over ATM. ATM’s high speed, high availability, scalable architecture molds well to the requirements of convergence. Voice over IP is a more general technology allowing a variety of networks to run underneath its mature, sophisticated protocols.
Several implementations allow for a gradual migration that many times uses much of the existing hardware. By properly planning and slowing making the migration, a company can be assured that end result will be a success.
Converging voice and data communications onto the same network is, by no means, easy. The two, although at first seeming alike, they are actually quite different at heart. Networks can be classified in one of two ways. The network is said to be connection oriented when a direct connection, physical or logical, is setup before data is transferred. Connection-less, however, simply addresses information and sends it to the recipient. Every packet is addressed and must be routed through the internetwork, meaning packets can take several different paths to the source. Voice networks are circuit-switching networks. They are connection oriented, whereby the caller and the called party have a connection established before talking. Data networks are a packet switching technology. No setup occurs when data is sent and received. Each individual packet must receive a network layer header with the destination address. When the packet is passed between routers, not all packets take the same path. This is because routing protocols have intelligent route selection capabilities that allow load balancing and other features. It is easy to see intrinsic difference. How do you make connection-less behave as connection oriented?
Voice service has been highly refined for many years. Users have become accustomed to highly available, clear, fast connections when making phone calls. This presents a major quality of service (QoS) hurtle that must be overcome for Voice over IP to be accepted Protocols have been developed that use certain bits within the IP header to define the Type of Service (ToS). Currently, many vendors have used these bits in a proprietary manor but the IETF has decided to redefine them. Another issue arises when defining QoS, what do you do differently with high priority traffic versus low? To this RSVP has answered with the ability to define a route through the network and then have high priority (Voice) traffic routed along that same path.
The leaps and bounds that technology has made in recent years have opened the door to faster routers with much more sophisticated routing protocols. Enabling higher and higher data rates that are necessary for the limited delay requirements of voice traffic. Even network design has been rethought to allow for speeder and more reliable connections. Innovations and education from vendors like Cisco, 3COM and Nortel have lead to lowered congestion on network segments. This enables networks to scale as large as the company and maintain similar features across the whole enterprise.
Throughout this paper it will discuss both business and technical issues associated with migrating towards a seamless voice and data network. It would be unwise to try to implement these changes too quickly. The quality of service users are accustomed to must not change. The object of networking is to increase productivity and decrease cost. A converged network promises both but the migration process must be well managed in order to ensure a smooth transition.
What is Convergence?
Convergence has been a hot topic for many years. The dream spawned by the Internet’s wealth of possibilities, of a combined voice, video and data network has fueled vendors to come up with an industry-wide, non-vendor-specific solutions. More importantly for business this dream spells big savings over the long run. Three areas of the IT budget should see savings.
? IT Personnel ? Rather than having data-network personnel and voice-network personnel. IT staff will be required to be knowledgeable in both areas and therefore cutback to one slightly larger team.
? Network Equipment ? Although at first, in order to establish the technology, cost may be significant. By using Computer Telephony Integration (CTI) the need for dedicated, specialized devices can be reduced. Also packet switching is soon becoming as much as 20 ? 50 times more cost-effective than circuit switching because of its connection-less nature.
? Transmission Services ? Mainly dealing with cost savings from non-US calls.
Convergence is defined as combining voice and data in one media without channellizing. There are basically five ways of doing this:
? Point to Point digital circuits
? Frame Relay
? Corporate intranet
The global market that we live in today demands that businesses conduct calls with foreign countries. The price of these calls can have a high impact on the IT budget (see chart 1).
Destination Country Cost Per Minute
For most large companies, US calls should not cost more than three cents a minute. The cost savings for international calls, on the other hand, by using VoIP is obvious after considering the volume of calls that occur.
-Packet Magazine V.12, N.2, page 63
Utilization ? Getting your Money’s Worth
It’s a fact that data communications is bursty. Meaning, data transfer peaks for a moment and then is stagnant. Consider when you are browsing on the Internet. Data transfer is high as the page downloads. Once loaded, you sit and read. The connection is idle and bandwidth is not being used. For a business, this unused bandwidth is wasteful because it could be used for other traffic that may need it.
Utilization is formally defined as ?The percent of total available capacity in use.? Capacity being the total ?data carrying capability of a circuit or network in bits per second.? The cost associated with high-speed circuits is too great to allow them to go unused. Optimum network utilization occurs for Ethernet under 37%. After this point the network is too saturated with communications and token passing methods out perform CSMA/CD (Carrier Sensed Multiple Access with Collision Detection). For token passing methods utilization can approach upwards of 70%. WAN links, such as those used for VoIP, should be operating at about 70% utilization before considering an upgrade.
Mixed Media Requirements
In a nutshell: DATA ? accurate not timely
VOICE ? timely not accurate
The level of service that users expect, when making a phone call, is extremely high. It has been found that if users experience as little as 500ms round trip delay, they consider it a problem. Consider the graph below from the Voice and Data Handbook, 1999. Problems are few until delay goes beyond 300ms and becomes a concern at about 500-600ms.
kilometers per millisecond. Also part of propagation time is the delay caused by putting data onto the media. It’s dependent upon the data volume and the speed of the line but consider for example to put a 1,024 byte packet onto a T1 line it would take about 5 milliseconds.
The delay we do have controller over is rightfully called, variable. The type of routers chosen, even the protocols used, the speed of the media; all of these are variable delays. It is important to consider the all aspects of your network design. Increasing performance by reducing delay can be much more cost effective than simply adding more circuits.
Packet loss is exactly how it sounds; packets are dropped for one reason or another. It’s not so important for Voice over IP because of built in codecs to compensate for up to 10% loss. Fax, however, is inherently more sensitive to both delay and packet loss. The quality of links should be evaluated along with source and destination hardware. If media is not of sufficient speed or reliability consider upgrading to ensure throughput.
Jitter isn’t much of an issue with data communications but for voice you could get the wrong idea if someone said, ?gone the people have,? when they meant, ?have the people gone?? Jitter occurs when packets arrive at the destination out of order. Packets can be numbered, taken into a buffer and released at the same time. This obviously contributes to delay but if jitter is occurring sufficient bandwidth must exist.
Voice over ATM
?There are basically two models for integrating voice and data ? transport and translate? Transport is the transparent support of voice over the existing network. Simulation of tie lines over ATM using circuit emulation is a good example. Translate is the translation of traditional voice functions by the data infrastructure. An example is the interpretation of voice signaling and the creation of switched virtual circuits (SVCs) within ATM? ? Internetworking Technologies Handbook, Second Edition.
Whenever you speak of the network technologies involved with transporting simultaneous voice and data, you must choose from the few that are scalable and fast enough to handle the workload. ATM is one technology that definitely does just that. Beginning at the way the header is made up; ATM is arguably the best choice for transporting voice. The header, itself contains a pointer, which allows a digital signal level 0 (DS0) structure to be maintained. DS0 are the lines that today transport voice. They are multiplexed together to get larger and larger number of signals through.
Signaling with VoATM is compared in the pictures below. VoATM has the ability to either transport voice signals transparently through the network or to interpret and move the signals at ATM speeds. The second is more advantageous because of the use of SVCs or Switched Virtual Circuits. These are circuits, which do not have a physical end-to-end connection between users established. Rather, signals are passed through the network along a logical path that works exactly the same as if a sold connection was there. Allowing VoATM signaling translation is better for three reasons:
? SVCs are more efficient users of bandwidth than PVCs.
? QoS for connections do not need to be constant, as with PVCs.
? The ability to switch calls within the network can lead to the elimination of the tandem private branch exchange (PBX).
The addressing used for VoATM is 20 bytes in length and supports both public and private addressing schemes. Routing is handled by Private Network to Network Interface (PNNI) protocol. Newton’s Telecom dictionary describes PNNI as an extremely scalable, full function, dynamic, multi-vendor protocol. The way it works is a virtual circuit (VC) request a connection with a certain QoS through the ATM network. The source ATM switch goes out and finds the best route matching the QoS requested. Each switch along the path is checked to determine if it has the appropriate resources necessary. When the connection is established, voice traffic flows between end stations as if a leased line existed between the two.
VoATM has many built in feature for controlling delay and delay variance. The VCs can request specific bit rates with bandwidth and delay guarantees. There are also VC queues allowing each traffic stream to be treated uniquely. The use of small, fixed-size cells reduces queuing delay and delay variation due to variable-size packets.
Voice over Frame Relay
Frame Relay is one of the most widely implemented WAN technologies. Its inexpensive, yet reliable track record has made it very popular. The signaling used for Frame Relay has been historically proprietary. This has inhibited it’s progress into the voice market. The Frame Relay forum however has developed a set of standards known collectively as FRF.11 for VoFR (Voice over Frame Relay).
Static tables handle addressing for VoFR?certain dialed digits choose which PVC to use. Voice is routed depending on the protocol chosen for establishing PVCs. Depending on the protocol such things as bandwidth limits, hops, delay or some combination can determine route, although most concentrate mainly on bandwidth utilization.
When it comes to preventing delay Frame Relay falls a bit short. The frame size is variable. This means that delay variance is also variable. Different size frames pass through networking devices at different speeds. The smaller the frame the faster the passing but it’s an inefficient use of bandwidth because of the extra information associated with each frame. Longer frames take considerably longer but because more information is encapsulated within each frame it’s a better use of bandwidth. Up until now, the solution to this problem has been proprietary. However, the Frame Relay Forum is defining what is known as FRF.12, which will create an industry standard to solve the small frame size problem.
Voice over IP
What’s so different about Voice over IP rather than VoATM or VoFR? VoIP is capable of converging voice and data at the application layer, rather than manipulating lower layers. This has the most appeal to people interested in cable, DSL and wireless networks because it allows service providers to bundle their services.
To make this a bit clearer, the protocols running over the network itself control all the functionality instead of the network itself. Regardless of the technology running under it VoIP provides a solution for everyone. In order to do this VoIP must provide a solution for signaling, routing and addressing.
Signaling for VoIP has three distinct areas: PBX to router, router to router and router to PBX. The corporate intranet, to the PBX, looks like a trunk line. Signals are sent from the PBX through the corporate intranet to seize a trunk using any of the common signaling methods. FXS or E&M signaling is used for fax and in the future common channel signaling (CCS) or Q.SIG will become available as a digital signaling method. The PBX then forwards the dialed digits to the router in the same way they would be sent to a Telco switch.
Within the router, the digits are mapped to an IP address and using Q.931 call setup establishes a request to the remote address. Meanwhile, the control channel is used to set up the Real-Time Control Protocol (RTCP) audio streams and Resource Reservation Protocol (RSVP) is used to guarantee QoS.
When the remote router receives the Q.931 call request, it signals a line seizure to the PBX. After the PBX acknowledges, the router forwards the dialed digits to the PBX and signals a call acknowledgment to the originating router.
All the responsibility for session establishment and signaling is with the end stations. To successfully accomplish this, additional enhancements must be made to the signaling stack. H.323 is such an addition and will be discussed in-depth next.
Corporations should already have an IP addressing scheme in place. The voice interfaces will show up as additional nodes, either as an extension of the existing scheme or with new IP addresses. The dial plan mapper performs translation of these addresses. The destination telephone number or some portion is mapped to the destination IP address. When the number is received from the PBX, the router compares the number to those mapped in the routing table. If a match is found, the call is routed to the IP host and is transparent to the user.
VoIP real strength is rooted in IP’s mature and sophisticated routing protocols. By using routing protocols such as Enhanced Interior Gateway Routing Protocol (EIGRP) specific factors including delay are taken into consideration for best route decisions. Other advanced features like policy routing and access-list allow you to create highly secure networks. Increasing innovations, such as tag switching, are also being developed to allow better traffic engineering. This will lead to the ability to shift traffic load based on different variables, such as time of day.
Traditionally, IP traffic has been handled on a ?best effort? mechanism. Traffic was first come, first serve but voice is not tolerant to retransmission and delay. Also the variable packet size problem is an issue. Once again using RSVP to initially find a route through network and then using RFC 1717 to break up the large packet to a standard, smaller size was the solution. Weighed fair queuing was also used to put different traffic types into specific QoS queues and thus reducing queuing delay.
The ITU created the H.323 standard to enable mixed-media communications over packet based networks that do not provide QoS. The standard is said to be an umbrella encompassing various associated standards (See chart 2). Although H.323 provides support for audio, video, data and multipoint conferencing, only the audio support is mandatory.
H.320 H.321 H.322 H.323 H.324
Purpose Narrowband ISDN Broadband ISDN, LAN, ATM Guaranteed bandwidth packet networks No guaranteed bandwidth packet networks and Ethernet Analog PSTN telephone system
Audio G.711, 722, 728 G.711, 722, 728 G.711, 722, 728 G.711, 722, 723, 728, 729 G.723
Video H.261, 263 H.261, 263 H.261, 263 H.261, 263 H.261, 263
Multipoint H.231, 243 H.231, 243 H.242, 243 H.323
Control H.320, 242 H.242 H.231, 243 H.245 H.245
Interface I.400 AAL I.400, TCP/IP UDP/IP, TCP/IP V.34
-The Irwin Handbook of Telecommunications, 4th edition.
H.323 power comes from its multitude of other standards. Many applications are possible by using this architecture including: Internet telephony, desktop videoconferencing, LAN telephony, conference calling and mixed media conferences such as voice, video and whiteboard.
Interoperability is a key feature in today’s networks. H.323 uses industry open standards which when followed by vendors allows other products to work together. A general H.323 architecture is shown in figures 1 & 2 below. The TCP/IP network uses TCP (reliable connection-oriented protocol) for call setup and UDP (fast, connection-less protocol) for voice packets. A signaling channel known as the RAS channel is used for communications between devices. Real-Time Transport (RTP) is used to sequence packets, compensating for UDP’s lack of this capability. Real-Time Control Protocol (RTCP) monitors QoS.
-Figure 1-Radcom VoIP Technology Protocol Reference poster.
? Gatekeeper ? Manages a zone (collection of H.323 devices).
o Required Functionality ? Address translation, admissions and bandwidth control.
o Optional Functionality ? Call authorization, bandwidth management, supplementary services, directory services, call management services.
? Gateway ? Provides interoperability between different networks, converts signaling and media e.g. IP/PSTN gateway
? H.323 Terminal ? Endpoint on a LAN. Supports real-time, 2-way communications with another H.323 entity. Must support voice (audio codecs) and signaling (Q.931, H.245, RAS). Optionally supports video and data e.g. PC phone or videophone, Ethernet phone.
? MCU ? Supports conferences between 3 or more endpoints. Contains multipoint controller (MC) for signaling. May contain multi-point processors (MP) for media stream processing. Can be stand-alone (i.e. PC) or integrated into a gateway, gatekeeper or terminal.
Types of VoIP Implementations
VoIP through a Router
? If a PBX already exists, it makes maximum use of existing resources
? The service is completely transparent to users
? The connection can be completed over any available packet network.
? Blockage of voice calls should be rare since the PBX can complete the call over the PSTN.
This configuration allows you to connect devices directly to the network. Analog telephones can be connected using an Ethernet adapter through a PC. The PC gives you a lot of versatility because it can substitute for the telephone’s button interface. Calls within the zone are controlled by the VoIP gateway rather than having a PBX onsite. This implementation is inexpensive and great for branch offices.
Also known as the un-PBX. This implementation has PBX hardware and software function loaded on a PC running something like Windows NT or Unix. The various cards can be loaded into the PC and generate call-processing programs. Obviously, though, the fault-tolerance of an un-PBX compared to a real PBX is no contest. PBXs are very specialized and refined systems that are far more robust than any PC.
VoIP through a Gateway
This implementation is very similar to VoIP through a router, however, instead of using a router to route the calls; the functionality is part of the PBX. This can be a function of one of the cards in the PBX or simply a stand-alone device connected to the PBX. According to the Irwin Handbook of Telecommunications, ?Some manufacturers such as Lucent and Nortel provide IP trunk cards, but others do not, in which case the PBX would connect to either the router or the gateway through standard T1/E1 or analog tie trunk cards.