|
Conferences and Events | Online Resources | Programs | Security | Services | Shared Network | Technical Support | Training Home » Technical Support » Research and Innovation » Voice over IP » Voice over IP |
|||||||||||||||||||||||||||||||||||||||||||||||||||||
Voice over IPPresented at the MOREnet Spring Technical Conference, March 1-2, 2001 PrefaceThe purpose of this white paper is to provide an overview of Voice over IP (VoIP) technologies. We begin by defining VoIP, discussing why it is important in today's network infrastructure. Next, we discuss the architecture of VoIP, including an overview of basic voice concepts, and conclude with a discussion of Quality of Service issues pertinent to VoIP. To test VoIP, MOREnet, cooperating with Moberly Area Community College, initiated a pilot project with the goal of testing voice over IP as a means of "toll bypass." Our analysis of this project is included in this white paper as well. One final consideration to bear in mind: this paper presents VoIP technology from the perspective of Cisco System's solutions. We focused on Cisco's VoIP products for three main reasons: (1) Cisco is currently the market leader in VoIP solutions, (2) to reduce interoperability issues during testing, and (3) because of the familiarity with Cisco products held by the researchers. Defining VoIPThe term Voice over IP has come to mean a number of things and encompasses a broad range of technologies. In general, VoIP is a technology that digitizes and compresses voice conversations, then stuffs them into IP packets for transport over a public or private IP data network. Using this method, voice traffic can originate and be delivered to any point on an IP network that has a valid IP address. Note that fax transmissions cannot be compressed; although a fax can be sent over VoIP as packets, it must be sent as an uncompressed data stream. Contrast to Voice over Frame, ATM, etc.How might we contrast VoIP to other voice over packet delivery technologies such as Voice over Frame (VoFR) or Voice over ATM (VoATM)? VoIP is layer 3; VoFR/ATM are layer 2. VoIP verses Vo'X' can be compared to the queen in the game of chess verses a pawn, rook, or bishop. The queen can move anywhere on the board, while the others are restricted to a determined path. As an example: You have a friend in another state to whom you would like to make a voice call (toll-free, without using the traditional Public Switched Telephone Network). Your friend has a VoIP-capable device, as do you. Connect a phone into a port on each device and program the devices with a dialing plan that says "when I call this number, send it to this IP address." If you both have a connection to the Internet, the call goes through. The process is no different than accessing a web page. With the other Vo'X' technologies, in order to make this simple call, you would need to get at least one and often several carriers involved to build a point-to-point virtual circuit between the two endpoints before a call could occur. FlavorsToll Bypass - Sometimes referred to as "hop on, hop off" or "phone-to-phone" calling is where you use the data network to bypass toll charges between locations. This is probably the easiest form of VoIP to implement and provides the quickest return on investment. It is the "get your feet wet" of the voice over data network technologies. Public carriers such as Level3 are using this technology to provide extremely low-cost long distance. IP Telephones - Smart phones that use the Ethernet, rather than a separate network for voice. IP phones are like any other device on your network requiring an IP address. Commonly obtained from a DHCP server, IP phones can have public or private addresses, or can be statically addressed (by hand!). IP phones can be powered by an AC adapter or from switches or patch panels made for this purpose. The IP phone sets come in several styles ranging from a few function keys to built-in web browsers. A call manager or a PC-based Private Branch Exchange (PBX) is required with these phones to do call setup, billing, etc. Soft Phones - Software-based IP "phones" that run on a PC. This is referred to as "PC-to-phone" or "web-phone" calling. Examples are Dialpad.com or Net2Phone.com. A sound card with a microphone is all that is required; for best audio, a headset with a microphone made for IP telephony works best. Why is this important? Why do this?Advantages of packet switching over circuit switchingThe Public Switched Telephone Network (PSTN) is a circuit-switched network. When a telephone call is initiated, a circuit is established between the calling party and the called party that reserves a path, bandwidth, and processing time for the call. Today's circuit-switched networks are Time Division Multiplex (TDM) based networks. TDM sends data at a constant rate, even when there is no data to send. In this case, TDM time slots are filled with "filler" data to keep the data stream constant. The path is reserved and 64k of bandwidth is consumed in both directions until the call is completed. While this provides an inefficient use of resources, it nevertheless accounts for the non-bursty, low latency, and reliable nature of today's PSTN network. The packet-switched network stuffs data into individual packets to be routed across the network. A packet-switched network can be highly efficient and flexible. Packets are not sent until they are filled and bandwidth is only consumed when packets are sent. Voice over IP, utilizing compression techniques, silence detection, and Quality of Service (QoS) features, attempts to share the advantages of both. Cost SavingsYou have your LAN wiring in place. There is no longer a need to maintain two separate networks, two separate sets of wire, or separate staff--one to do the phone system and another to do the LAN administration. It's commonly called the integration or convergence of voice, data, and video on one network--and there are cost savings associated with this approach. Toll-bypass allows for savings. For instance, one school district in Kansas City could call another district in Columbia, toll-free as it rides the data network between the two. Additionally, a school in Columbia could call a school supply store in Kansas City, making a local call as it jumps off the network in Kansas City. There are also savings related to reducing the number of lines or not having to purchase additional lines from the local telephone company (Telco), as more of the voice traffic can be shifted to the data network. Other cost savings can include: the bulk purchase of CO (Central Office, i.e., the phone company) lines, volume discounts on long distance (because it can be aggregated), a single call processor can service multiple locations (rather than one per location), and/or a single voicemail system, etc. on top of the benefits to having a single system to administer, maintain, and upgrade. ArchitectureParts and PiecesThe architecture of an IP telephony network and the associated parts and pieces required, depends greatly on what you are trying to accomplish. For "toll bypass," the outlay of equipment is relatively minor. All that is required is an IP to PSTN gateway device. This functionality can now be built into a router or can be purchased as an add-on to a PBX. In either case, you already have a phone system in place comprised of telephone sets, wiring, PBX or key system, Telco access lines, etc. The VoIP gateway acts as an interface, routing calls on and off the IP network. Those seeking to replace all the redundancy and features found in today's office telephony environments with LAN based equipment, require at least two other devices, in addition to the gateway. The first of these is a specialized server that handles call setup, billing, call forwarding, conference calling, etc. In other words, the functionality found in most PBX's today. One common implementation of this, called a "soft PBX," is a PC running Windows NT, on top of which runs call management software. The second component is the telephone itself. These come in a variety of forms ranging from a normal looking handset that connects to a sound card in your PC, to stand alone units that are impossible to tell by appearance from a standard office phone (these devices are referred to as IP phones). As discussed earlier, they connect to the LAN via an Ethernet connection and require an IP address, as any other node on an IP based LAN. They are usually remotely programmable and require a source of power, either through an attached AC/DC adapter or from a hub to which they are attached. The latter type of IP phone (which is powered via a hub), is being supported in increasing number by makers of IP telephony equipment and distributes 48-V power over the unused fourth pair of wiring in a typical Ethernet wiring plan. E.164 NumberingAs an IP address identifies a particular host in the data world, the PSTN uses a specific numbering scheme, which complies with the International Telecommunication Union Telecom Directorate (ITU-T) E.164 standard, to identify a particular voice terminal device able to originate or receive voice/fax calls. In North America, the North American Numbering Plan (NANP) is used, which consists of an area code, an office code, and a station code. Area codes are assigned geographically, office codes are assigned to specific switches, and station codes identify a specific port on that switch. The format in North America is 1Nxx-Nxx-xxxx, where N is a digit between 2 and 9 and x is a digit between 0 and 9. Internationally, each country is assigned a one- to three-digit country code; the country's dialing plan follows the country code. In Cisco's voice implementations, numbering schemes are configured using the destination-pattern command. CODECSWhile a Cisco VoIP device can accept a digital call directly, a coder-decoder (CODEC) must convert analog calls before they can be transported over a packet-switched network. It is the responsibility of the Digital Signal Processor (DSP) to perform the conversion using a variety of digital encoding schemes. The two most popular encoding techniques are Pulse Code Modulation (PCM) and Adaptive Differential Pulse Code Modulation (ADPCM). In both cases, the analog sound is converted to digital by sampling the analog sound 8000 times per second and converting each sample into a numeric code. CODECs also exploit the redundant characteristics in analog sound, to conserve bandwidth through a variety of compression techniques. Coding techniques are standardized by the ITU-T in its G-series recommendations. The most popular coding standards for telephony and voice packet are:
In Cisco's voice implementations, compression schemes are configured using the codec command. It is imperative to remember that a CODEC of one type can only communicate with a CODEC of the same type. Therefore, G.726 ports will only communicate with other G.726 ports, and G.711 ports will only communicate with other G.711 ports. Connecting to the PSTNFor anything other than connecting one branch office phone to another, you will need to connect to the Public Switched Telephone Network- traditionally AT&T or "Ma Bell." This requires some knowledge of the types of connections available. Analog vs. DigitalThe telephone network has historically been an analog network. The human voice is analog. When you speak into a receiver, the sound waves travel through the air and cause a diaphragm in the receiver to vibrate. This vibration is modulated and sent down the line as an analog signal. Because of resistance in the wire, eventually the signal will weaken (this phenomenon is known as attenuation).If the signal must travel a great distance, attenuation can be a problem. For a signal to be recognized at the other end, it must be periodically boosted. Devices called loading coils boost the signal, but they also boost any line noise picked up along the way. This is where digital telephony is an improvement. By converting an analog sound wave to digital (through a process called sampling), the digits you dial or the sound of your voice can travel over the wires as a series of 1s or 0s. Although digital signals succumb to the same degradation and line noise as analog signals over distances, devices called repeaters are able to recreate the 1s and 0s perfectly. The result is the signal arrives exactly as it was sent. Access to today's telephone network consists of a mixture of analog and digital. The VoIP device can interface with either. A VoIP device configured with an analog voice module (AVM) can interface analog lines from a PBX (Private Branch Exchange, a telephone switch), or an analog fax machine or telephone handset. Likewise, a VoIP device configured with a digital voice module (DVM) can interface digital lines from a PBX, a digital channel bank, or a digital handset. Each VoIP application must be examined to determine its analog or digital requirements. The number of lines needed often determines which type of connection is ordered. An analog line can only carry one voice call at a time. A digital line, such as a digital T1, can carry up to 24 voice conversations at a time by dividing each voice call into a "time slot" referred to as multiplexing. Signaling Types: FXS, FXO, E&MAnalog and digital lines connect to telephony equipment on ports, as PCs connect to the ports on a hub. As LAN equipment talks over media in a variety of formats, such as Ethernet or Token-Ring, likewise, there are signaling standards for access to telephony equipment. There are three primary signaling types—FXO, FXS, and E&M. Foreign Exchange Station (FXS) ports provide that interface to a station device. A station may be a telephone, fax machine, modem, or any analog end-user device. An FXS interface provides battery and dial tone services to the device. The Foreign Exchange Office (FXO) interface points to an office. The office may be the central office (CO), a PBX FXS card, or the FXS card on another voice device. The FXO device receives battery and dial tone services from an FXS card in response to the FXO interface going off-hook. The Ear & Mouth (E&M) interface connects trunk lines, usually called tie lines. E&M is a signaling technique used mainly between PBXs or other network-to-network telephony switches to avoid problems such as glare (when both ends "pick up" at the same time). E&M interfaces use 4 or 6 wires, as opposed to the 2 wires that FXS/FXO interfaces use. The additional wires are the E and M leads, and refer to the direction of the signaling (an E lead at one end is connected to the M lead at the other, so when the M(outh) lead signals a call, the E(ar) lead sees the signal and can "seize" the trunk so the call can be processed). Dial Peers and Dialing PlansDial peers are the points to which or from which voice calls originate or terminate. Dial peer definitions describe the characteristics associated with each point of a "call-leg." A call-leg is a discrete segment, which makes up an end-to-end call. There are two types of dial peers relating to VoIP. POTS (Plain Old Telephone Service) dial peers are dial peers that point to a station (a handset, fax machine, etc.—any FXS device), a PBX or CO (or FXO device), or any other traditional telephone network device. VoIP dial peers are dial peers over which calls are routed, in this case, an IP network. There are four call legs over which a call is made: the call leg to which the originating POTS device is attached, the IP networks attached to the originating and terminating POTS devices, and the call leg to which the terminating POTS device is attached. Dial peers are defined on Cisco voice devices with the dial-peer voice {pots/voip} command. The dialing plan describes how to get to each dial peer. Their associate phone numbers identify POTS dial peers in the dialing plan. When digits are dialed, a VoIP device collects the digits and attempts to match a dial string pattern, defined in the dialing plan, in order to determine where to route the call. If no match is found, a fast busy signal is usually returned in the handset. The dialing plan is configured using the destination-pattern command. The dialing plan is equivalent to the routing table for a data network. Steps in a VoIP CallWhen a user picks up the handset, an off-hook condition is signaled to the POTS interface. The FXS voice port issues a dial tone to the handset and waits for the user to dial a telephone number. The user dials the number and the Dual To Multi Frequency (DTMF) digits are collected. After enough digits are accumulated to match a configured destination pattern, one of several things may happen. If the digits match on a POTS dial peer associated with another FXS voice port, a ring is generated to the handset. Once the station picks up, the call setup is completed, and an audio path is established until one of the two stations goes on-hook. If the digits match on a POTS dial peer associated with an FXO port, the matched digits are stripped by default, and the remaining digits are passed along to a connected PBX or any other traditional telephone switch that is responsible for completing the call. If the digits match on a VoIP dial peer, a true VoIP call scenario exists and the following steps are initiated:
QoS IssuesGiven that voice traffic is time sensitive, each device and link in the VoIP network must be examined as a part of a managed IP network to insure QoS (Quality of Service) objectives are met. Engineering the network to support VoIP involves a series of QoS and traffic shaping features that must be implemented. The movement of voice packets must be done expeditiously, with low delay, low packet loss, and a high degree of predictability. Voice quality is shaped by several technologies; the two receiving the most attention are echo cancellation and packet prioritization (implemented to reduce network delay and jitter). EchoEcho occurs when you hear your own voice in the telephone receiver while you are talking. If timed properly, echo is reassuring to the speaker. However, if echo exceeds 25ms, it is distracting and disrupts the conversation. Echo, caused by an impedance mismatch between the 4-wire network switch and the 2-wire local loop, is unavoidable but can be controlled. It is controlled by the use of echo cancellers--special code that listens for the echo signal and subtracts it from the listener's audio signal. In a Cisco VoIP device, echo cancellers are built into the CODECs and are operated on each DSP. Echo cancellation is enabled by default using the echo-cancel enable command. Network Delay and JitterVoice traffic is real-time traffic; if there is too long a delay in voice packet delivery, speech will be unrecognizable. An acceptable delay is less than 200 milliseconds. Though it is outside the scope of this document to discuss the merits and pitfalls of each QoS mechanism, suffice it to say, delay is inherent in voice-networking and is caused primarily by two know factors: propagation delay and handling (sometimes referred to as serialization) delay. Propagation delay is caused by the finite amount of time a signal (either electrical or optical) takes to physically move down the cable and through the switches, even when it is traveling at close to the speed of light. It is, for all practical purposes, the fixed portion of network delay. Handling delay can vary and is caused by the devices that handle voice information. Handling delay can have a significant impact on voice quality in a packetized network. In VoIP, the DSP generates a handling delay of 20ms. The DSP generates a frame every 10ms. Two of these frames are then placed in one voice packet. While compression techniques are used to conserve bandwidth, they also generate handling delay as the time it takes to compress and decompress voice data. The table below (from a Cisco Systems, Inc. white paper titled "About Voice, Video, and Home Applications, http://www.cisco.com/univercd/cc/td/doc/product/software/ios120/12cgcr/voice_c/xtocid826320) shows the amount of delay induced by the most popular CODEC techniques.
Another source of handling delay is the time it takes a Cisco router to determine a VoIP packet's destination and move it to the output queue to be sent out the interface. Output queue delay is a QoS issue and is minimized by a number of QoS techniques (configurable as a part of Cisco IOS) designed to give priority to voice packets over data. Jitter is also a factor that affects delay. Jitter is the variation between the time a voice packet is expected to be received and when it is actually received. Cisco voice devices compensate for jitter by setting up a playout buffer. Play back is controlled by Real Time Protocol (RTP) encapsulation and attempts to match the play back at the destination device to the rate at which it was captured. NOTE: As a precursor to any QoS implementation, the device must be running the appropriate level of IOS software and meet the appropriate memory requirements. Pilot Project — MACCProject ScopeIn the first quarter of 2001, MOREnet, with the cooperation of Moberly Area Community College, initiated a pilot project with the goal of testing voice over IP as a means of "toll bypass". Since the term voice over IP (VoIP) has come to mean any number of things and encompasses a broad range of technologies, we limited the scope of our testing to "toll bypass," sometimes referred to as "hop on, hop off" or "phone-to-phone" calling. ObjectivesThe following objectives were set for the project:
VoIP Network DesignMoberly Area Community College (MACC), consists of 5 campuses--the main campus in Moberly Missouri, and the four satellite campuses located in Kirksville, Hannibal, Mexico and Columbia. These campuses are tied together via T1 Frame-Relay at layer-2 and IP at layer-3, with MOREnet providing the backbone routing. All the campuses except Hannibal were involved in the test. Local Calling PrefixesKirksville665, 627, 785, 626 (Altel), 341 (US Cellular) Columbia
657, 696, 874, 875, 876, 698, Moberly263, 269, 351, 651 (US Cellular) Mexico581, 582, 473 Steps to ImplementationAnalysis and Design PhaseThe proper analysis was done to determine what devices should be purchased and in what configurations, to maximize achieving of the project objectives. At the same time, careful consideration was given to how VoIP devices could be "dropped" into the already existing voice and data networks at each of the four campuses, in a manner that would accomplish the following:
MACC personnel were contacted to provide MOREnet with a list of local prefix codes in each of the four locations. The prefix or local exchange codes are those exchanges which are considered to be a local call for the individual campuses (including cell phone and pager numbers). These prefixes will be used in a later step in building a dial plan for each location. The next step in the analysis phase is to evaluate the existing network to determine where bandwidth and latency issues may be a concern. As a precursor to any QoS implementation, the device must be running the appropriate level of IOS software and must meet the appropriate memory requirements. Equipment PurchasedFor purposes of testing VoIP, a variety of voice access devices, in a variety of configurations, were purchased. Only one vendor, Cisco, was represented in the testing. Cisco's products were selected to reduce interoperability issues and because of the researchers' familiarity with the Cisco IOS. The following equipment was purchased for the project:
Configuring the EquipmentConfiguring voice over IP on any device must begin by insuring that the hardware is installed correctly and the appropriate IOS version is running on the device. Each type of equipment takes a different version of the IOS, downloadable from the Cisco web site. The following Cisco IOS and feature set was installed on each device.
Configuring the Voice PortsThe second step is to configure the voice ports. As discussed earlier, voice ports may be digital or analog. Voice port commands define the characteristics associated with a particular voice port signaling type--FXS, FXO, or E&M. Under most circumstances the default voice port values are adequate. Some of the configurable options are:
Configuring Dial Peers—POTS or VoIPEach dial peer defines the characteristics associated with a call leg. An end-to-end call is comprised of four call legs--2 associated with the source device and 2 associated with the destination device. Before dial peers can be configured, a dialing plan must be established to route end-to-end calls. This requires, from the standpoint of the telephony device, the phone numbers or string of digits to be matched on an incoming call and those to be matched on an outgoing call. POTS dial peers enable incoming calls to be received by a particular telephony device. VoIP dial peers enable outgoing calls to be made from a particular telephony device. The programmer must build a list of local calling prefixes, area codes, and station numbers where applicable. Along with the dial string pattern, the voice port to which the dial peer definition applies must be defined. Under most circumstances this is all that is required to establish a connection. However, other configurable options that may apply to the configuration are:
TestingIn the MOREnet R&D Lab, each of the purchased devices was connected together via cabling from the Ethernet port to an Ethernet switch. To each of the devices, where applicable, a handset was connected to an FXS port. A "ping" verified IP connectivity between the devices. Calls were made from device to device to verify the accuracy of the dialing plans. In an alternate setup (to extend testing to the reality of the MACC network design and to test QoS issues), the Cisco 2621 was connected to the Ethernet network over a frame-relay link. This link was establish by connecting two 1720 routers back-to-back, with a frame-relay switch in the middle. Bursts of IP traffic (provided by Chariot software) were sent over the frame-relay link at the same time voice calls were made, in an attempt to test voice quality in a "real-world" scenario. The 2621 was configured with a T1 digital voice module (DVM) and connected to a channel bank for breaking out each of the 24 channels into separate voice ports, configurable as FXS or FXO. The testing phase uncovered the following issues: Incompatibilities between the Cisco MC3810 and VG200When a handset on a 3810 initiates a call to the VG200, no ring is heard in the calling handset. In one instance, when "passing DTMF digits out-of-band" was enabled on the 3810, a connection established between the 2 devices would be dropped after a few seconds. The Cisco 2621 does not support "enhanced dial peer" featuresThese features provide greater functionality in how dial string digits are played out to the voice port. Among the features lacking are prefix insertion and digit strip. Routing on the VG200The VG200 does not support a routing protocol such as RIP, OSPF, or EIGRP. Instead, IP routing is limited to "static" routes, which will not scale well. Scalability IssuesCisco VoIP devices, as tested, will not scale well in large implementations. Each device must be configured with a dial plan supporting all the possible office codes to which the source device might call. A major metropolitan area may have hundreds of prefix codes representing a local call. Each of these codes must be programmed in each device, not located in the local calling area. As the number of VoIP devices deployed in the network multiplies, the administrative task of maintaining the dial plan tables in each device also multiplies. Other than compatibility and scalability issues discovered, the test went well. The quality of the voice calls was good, even under load. ConclusionVoice over IP technology, as tested, performs the advertised functions of transporting voice, with reasonable quality, over an IP network. Field tests with MACC support this finding, as users reported good quality audio during testing after the correct equipment and software was in place. Note that the tests did not generate a large volume of traffic, so quality of service mechanisms were not tested. VoIP still has issues that should be considered during any planning and implementation. Some of the issues will need to be addressed by vendors in future product releases. As noted earlier, interoperability, even among devices from the same vendor, was an issue. During testing, although voice quality was usually good, there were routing and signaling issues that were time consuming to troubleshoot due to a lack of in-depth diagnostic tools. Finally, scalability remains an issue. Until vendors, the VoIP industry, or standards bodies can agree on a system to deal with the phone and VoIP number look-up, large scale and/or interconnected VoIP systems remain difficult to implement at best. VoIP ResourcesURLsCentral office codes Area codes by state Cisco VoIP Technical Tips Cisco Tech Notes - Understanding Dial Peers and Dialing Plans VoIP: An Overview IP Telephony Standards BooksIP Telephony by Walter J. Goralski, Matthew C. Kolon Voice & Data Internetworking by Gil Held, McGraw-Hill, 2000 Example Configuration! |
|||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Copyright © 2002 MOREnet. All rights reserved. Reviewed February 18, 2002.
Contact techsupp@more.net. DMCA and other copyright information. Site Information: Copyright, accessibility, privacy and other information about this site. PageMinder: Receive an e-mail notice when this page updates.
|