Cisco Networkers 1998

home *** CD-ROM | disk | FTP | other *** search

/ Cisco Networkers 1998 / Cisco Networkers 1998.iso / pc / talks / nwk01 / nwk01.eml < prev next >

Wrap

Text File | 1998-04-08 | 47.7 KB | 681 lines

<HTML> <HEAD> </HEAD> <BODY>   Hello and welcome to Networkers. My name is Chris Beveridge. I'm a Corporate Consulting Engineer with Cisco Systems. Today I'll be covering the basics of telephony networking, Part one of two parts. On today's agenda we'll be covering the basics of analog telephony, digital telephony, and a concept we like to call consolidated transport networking. To start off with, I'd like to identify the key components of transport networking for telephony, the telephone set itself, a key system which basically optimizes the use of many telephone lines. Mechanical to electronic are the options within a key system, and two to ten sets is a typical size for a key system to support. A PBX is similar to a key system but it's much larger and has more advanced calling features such as caller ID and call forwarding, call conferencing, etc., and typically supports ten to hundreds, to even thousands of telephone sets. Additionally, in the central office they have what is called a CO switch or central office switch which is a much larger and more capable wide area voice switch. Starting off with analog telephony, you may have heard a term called POTS, P-O-T-S. It's an acronym for Plain Old Telephone Service, and that's used in - it's been used for about 100 years actually for the standard transport of any two telephone sets, it's the switching system between them, and the cabling. Now you notice on this graphic you see the words tip, ring, and sleeve. Those indicate the parts of the connector, the quarter inch plug, that support the twisted pair - that's the tip and ring - and the sleeve which supports the ground lead. I'd like to now go through basic call progression between two telephone sets and a telephone switch in a network. Here we have two local loops which is what is referred to when you look at the connection between the telephone switch and any end device which would be, in this case, the telephone set itself and another telephone set on the other side. So you have two local loops and a switch. The switch provides negative 48 DC voltage. In this condition, we have a DC which is an open circuit. So when the telephone is on-hook, it's actually not a closed but it is an open circuit, and no current flow on either local loop. As we progress through this, one person may take the telephone on the left-hand side off-hook, closing the circuit. This causes the telephone switch to provide DC current and dial tone to that telephone set. As we continue through our call progress, the user on the left-hand side will dial digits or dial - pulse dial digits and provide pulses or tones into the telephone switch. Notice there's still DC current flowing in the local loop. As the telephone switch receives the dial digits or tones, it does what's called an address-to-port translation. In other words, what it's trying to do is establish that the number that's being sent to it is a legitimate number for one of its other ports. As it finds the other port that is a legitimately dialed port, then it rings the local loop on the right-hand side. Now, that ringing may continue and if no one picks up, then obviously the person on the left-hand side will hang up. On the other side of this connection, when the user pulls the phone off the hook at the other end, closing the circuit, voice energy and DC current is flowing between the two phones - circuit switched through the telephone switch. And that's the fundamentals of a voice conversation end-to-end through any kind of telephone switch. Now let's look a little bit at signaling within the analog telephony realm. There are supervisory issues, addressing issues, and call progress issues that we'll cover. In terms of off-hook signaling, there are really two types in analog telephony, a thing called loop start which is almost all telephone sets, and ground start which is almost all PBXs. And there's a concept called seizure where you seize the trunk or you seize the local loop with a loop start routine. In a ground start routine, that seizure is detected when you take one wire and ground it, and it's called ground start. So you can seize in one direction or another. In other words, two PBXs or two telephone switches can seize the other independently. With loop start, the local loop from the telephone switch to the telephone is closed when you pull the receiver off of the hook. In ground start, the trunk between the switches is tied to the ground lead. Details on loop start at the top of this diagram shows the circuit open, battery voltage provided to the PBX or central office switch, the loop itself which is local or station loop, and the station itself which is the telephone set. As you pull the phone off the hook, DC current flows back through the telephone set from the switch. And in the event of a ringing request from the telephone switch to the telephone, that is AC voltage provided from the switch to the telephone set. There's a common term used in the industry called E&M signaling and I'm not entirely sure this is true, but I believe E&M stands for ear and mouth which is kind of ridiculous in a technical sense, but it works. Everyone understands E&M in telephony world and there are different flavors of E&M: two-wire, four-wire and there are different types of four-wire - there's five different types. PBXs and switches use E&M signaling to separate signaling leads for each direction. The E-Lead is the inbound direction to a switch, the M-Lead is the outbound direction to a - from a switch and this allows the independent signaling. And you can see from this diagram here, the on-hook state, if the E-Lead is open and the M-Lead is grounded; and the off-hook state if the E-Lead is grounded and the M-Lead is tied to battery voltage or negative 48 volt DC. In terms of signaling and addressing, the three different types of hand or telephone sets that you may encounter include dial pulse, DTMF or dual tone multifrequency, and ISDN or out-of-band message-based telephone sets. In the analog transmission in-band signaling is used. In other words, part of the audio frequency is used for signaling. These include the digits 0 through 9, * and #, for a total of 12 digits. Pulse dialing is represented here by a concept called make and break. Make is circuit closed, break is circuit open. When you go off-hook, if you go through the diagram here, there's a period of time between when the phone goes off-hook and when a user dials digits. When a pulse phone is dialed, the pulse period is typically 100 milliseconds. This diagram represents the dialed number three. And there is a concept or a standard actually in the United States and it's a 60 milliseconds break, 40 milliseconds make, or the 60/40 break/make shown here. Inter-digit time is important as well. You have to be able to discern between this dialed three and whatever the next digit dialed would be and those inter-digit timers, typically 700 milliseconds or longer. In tone dialing or DTMF, dual tone multifrequency, there are combinations of frequencies, audio frequency tones, that are used for each individual digit dialed. For example, the number 5 here would be, if you look on the left, it's 770 hertz and that in combination with 1336 hertz gives you the number 5. So the telephone key pad can be represented using dual tone as shown in the DTMF example here. Again, the timing between digits, 60 milliseconds break, 40 milliseconds make. There are also some standard network call progress tones that you may be familiar with. Dial tone is 350 hertz plus 440 hertz, that's a continuous tone. Busy would be 480 plus 620 hertz, and there's on-time and off-time, so that's where you get your busy tone, that's a standard. There's a normal ringback. In other words, as you're talking with someone, or you're dialing someone and they - you hear their phone ringing, that's actually called ringback. That's a 440 plus 480 tone. That's on for two seconds, off for four seconds, etc. So you can see the different examples of the progress tones here on the chart. In terms of voice channel bandwidth requirements, the lion's share of the voice frequency range is from about 200 cycles to about 4 kilohertz, so about 4 kilohertz in total. This includes voice signal itself, the voice energy, as well as tone dialing and systems control signals. A local access network from a typical central office will include anywhere from 40,000 - 50,000 lines and then there's what we call a feeder route. Each serving area in a feeder route has many twisted pairs in large cables transported back into the central office. Starting to become more standard, there are fiber optic transmission systems that go between these feeder routes and the central office. A typical large cable bundle will hold 900 pairs of telephones at local loops. In the most basic form, a switching system can be looked at as a patch cord or a patch panel. Now whether this is accomplished with electronic switching, digital switching, mechanical switching, somebody taking a cable and plugging into a jack, the switching concept is still the same. Basically a manual control or switch cord board shown here is the most basic form of a telephone switch. There's also the ability here to manually ring a telephone. If we look at the PSTN or Public Switch Telephone Network hierarchy in the United States, we have different classes of central offices. There are the highest class of central office which is called a Regional Center or Class 1, switches between large regions, such as the San Francisco Bay Area, New York City area, Seattle, etc. Then it goes down to the next level, which is a Sectional Center, that would be parts of that region. So for example, in New York City you might have the different boroughs each be a Sectional Center. These classes of central office further subdivide all the way down to a Class 5 and a remote switch unit, as you can see here on the list. The types of voice circuits that are supported between serving areas is basically three different types: a thing called off-premise extension, a thing called foreign exchange or FX, and automatic ring down, ARD. You might think about the automatic ring down as like the bad phone. Echo in voice networks is a - is an impairment that needs to be engineered for and understood for proper end-to-end transport. Between the talker and the listener there are many components of delay. But if you delay any kind of audio system and you have any kind of a reflection between the end talker and the listener you'll produce what's called talker echo and listener echo shown here on this diagram. In a normal signal flow the listener generating the signal on the two-wire local loop encounters the central office. And in the central office, the - there's a thing - and there's a devise called a two-wire to four-wire hybrid. Transport between central office switches is almost always done with four-wire interfaces. So you have a transmit direction and a receive direction split out, whereas on the local loop it would be both directions supported on the same twisted pair - two wires. If you have an impedance mismatch in this hybrid, it's a very common cause of echo. And basically what happens as you see here, the receive and transmit signals are superimposed upon each other, so the listener and the talker may hear echo in this situation. Echo is always present in any audio system. The trick is to either increase the echo path loss, in other words, attenuate in dBs, or make the delay path or the delay short enough in milliseconds that it's not noticeable by the user. It's a very subjective issue, but this graphic tends to show the tendency of the echo path delay, if it gets longer, even though it's a very faint signal, will be a problem. If the echo path loss is very large - or very small, then the - even a short delay will be noticeable. So the ways to defeat echo, basically you can increase the loss in the echo path, often done, and basically the - it's a static setting is the problem. The re-tuning of the hybrid, the two- to four-wire hybrid in the central office is a good example of this. Echo suppressing basically renders the two-way conversation to a half-duplex. In other words, just one side or the other is talking or allowed to pass signal energy at any one time. The best way, and the most effective way, is a thing called an echo canceller. Very commonly used in central offices and other end equipment. And basically what an echo canceller does is it looks at the incoming and outgoing signals and does what we call adaptive filtering, or comparison of the two signals. If they are close enough to be identical, then the return path of the voice signal is canceled completely, and it allows for a very clean way to eliminate echo completely. The problem, or the issue that goes along with that, of course, is that human beings don't like total silence. So if you're echo canceling in one direction, you have to introduce small amounts of noise so that end users don't think that the other person is either hung up or been disconnected. So that covers analog telephony. This technology dates back to the 1900s and late 1800s, and essentially a very simple system. The information exchange is based on voltage, current flow, grounding, etc. Now I'd like to start to cover the basic digital telephony concepts. Digital telephony really takes three forms: in the form of purely digital trunking between switches; in a hybrid form with analog local loop, plain old telephone service extensions; and what's called a channel bank what's represented here as CB, and a digital network switching system. And then the end device, or telephone swit - or telephone set, can be a digital device itself, an ISDN phone for example, which would be a digital local loop and a digital network. So how does an analog voice signal become digital? Well this technique is called pulse code modulation, invented by a fellow named Nyquist. And his theorem was that if you sampled a voice signal at twice the frequency that you needed to support, that would be sufficient to render it to a digital format and therefore transport across a digital network. This is done with a device called a codec which includes a thing called a sampler, and that sampling rate is once every 125 microseconds, or 8000 samples per second. Within pulse code modulation, the analog to digital conversion creates what called quantizing noise. As you're measuring, or sampling, the analog waveform and outputting digital bits from it, because this is a approximation of the actual analog waveform - and when it's reconstructed will be approximately what it was when it entered the system - there is a chance for quantizing noise. So the quantizing stage introduces a small amount of noise. The PCM technique counts on the ability of the user to filter out that noise to some acceptable level. Within digital telephony, there are framing formats - different framing formats for T1, which is North America and Japan and parts of Korea and Taiwan, and then E1/J1 which is effectively the rest of the world and other parts of Japan. As you look at the table here, we have a sampling frequency, a channel bit rate, time slots per frame, channels per frame, bits per frame, the framing format or what it's called, the framing indicator, which bit or bits indicate the frame technique, the system bit rate itself, and the signaling techniques. Within the DS1 framing format, we have the 193rd bit of each frame is used for frame synchronization. So this - this bit must fall into an ideal position in time. We will cover network synchronization in a few minutes. With - there are two types for DS1, two types of framing, D4 framing and ESF framing. D4 is - refers to a D4 channel bank, it was a Western Electric product from years ago and it's still used as the nomenclature. It's actually more accurately called Super Framing in standards. Extended Super Framing uses 24 frames with framing CRC and a facilities data link channel. The ESF framing pattern is 001011. There's also channel associated signaling and common channel signaling options within ESF and DS1. Within the Extended Super Frame format, in each frame as is shown on the left-hand side, one through 24, there are S-bits, there are each channel time slot, or robbed bits, and then there are signaling bit use on the right-hand side. As you look at the S-bits, you have Fe, or the framing alignment bits, F-sub-e is what it's called here, and that pattern in frame 4, 8, 12, 16, 20, and 24 is sequentially, 001011. Within the S-bits there are also facilities data link bits and B-sub-c bits which are a CRC6 error counting mechanism. Then the traffic is transported in bits one through seven of each channel, or each frame, and signaling bits are robbed from bit eight in channel associated signaling. Now those bit designations are shown on the right here. They can either be transparent, shown with the asterisk. They can use all A bits. You can have a combination of ABAB bits, or a combination called the ABCD bits. If a signaling scheme requires all four bits, then you would use the ABCD scheme. Within the E1 frame format, which is used in Europe and most of the rest of the world, rather than robbing bits we use time slot 16 for A, B, C, and D signaling bits, this is channel associated signaling once again. Also in time slot 16 is what we call the multiframe alignment signal. In E1, a multiframe is 16 frames of 32 time slots each. These are aligned with time slot zero, which is the synchronization time slot. There is a fixed-bit pattern in time slot zero for synchronization, so unlike T1, which uses a single bit, time slot zero in E1 is used entirely for frame synchronization. In terms of digital signaling schemes for Extended Super Frame, we have audio address signaling which is the DTMF tones we spoke about earlier, and then we have supervisory on-hook, off-hook, and address signaling which are the robbed bits that we looked at earlier. So on-hook/off-hook indications for the telephones, and dial pulse information, basically bit twiddling. In E1, common channel signaling is an option in time slot 16, which means basically it's a open data channel. You can use it for almost anything you'd like. There are many proprietary versions of common channel signaling, or CCS. The most common variety is ISDN, which is a standard digital signaling scheme. Now to keep frame synchronization, it's very important to have bit synchronization, time slot synchronization, and frame alignment. These are all critical components. And synchronization sources are called primary reference sources, or other reference sources that are below the primary reference source. You may have heard these referred to as stratum sources, or primary bits clocks, b-i-t-s clocks, etc. And really, the idea is to make sure that all the devices within the transport system, in the digital realm, are able to keep synchronization on the frame level, on the time slot level, and on the bit level. So if we look at an ESF multiframe, the entire multiframe takes three milliseconds to pass. So that's 24 frames. If we look at the individual frame broken out here, frame number 12, each frame itself takes 24 time slots in 193 bits or 125 microseconds. And then if you pulled out channel 12, or time slot 12 of the frame, you would see that each bit requires 648 nanoseconds. All of these have to have their ideal positions in time preserved, and this is done usually with atomic clocks. The representation here shows the different stratum levels that are referenced in the U.S. hierarchy. The little map there shows Missouri. The reason for that is that historically, AT&T had atomic clocks hidden in caves in Hillsborough, Missouri. And since then there are many other alternate carriers - long distance providers that have different locations, but historically the primary reference source or the master clock was located in Missouri for AT&T's wide area network. As you can see there are very stringent requirements on this, and the timing is taken from the primary reference source and fed into the toll office switches which are then distributed per LATA, L-A-T-A, is Local Area Transport Access, down into end offices, devices called DCS, or Digital Cross-connect Switches, and from there down into stratum four levels which is typically fed a PBX or any other kind of voice switch. So digital telephony in summary is basically analog emulation and pair gain concepts. It's the backbone of the largest telephony network in the world, largest interoperable network in the world, and the signaling information is effectively based on approximately 30-year-old concepts. Digital telephony is - within the analog telephony emulation realm, is voice encoding, we saw PCM and Nyquist, limited signaling capabilities, typically robbed-bit signaling or channel associated common channel signaling, and loop consolidation. In this next section, I'd like to call consolidated transport networking, we're going to look at many aspects of the different types of transport for digital and analog telephony, including remote branch office access, gateways between different signaling systems, network trunking, and gateways between cell relay, Frame Relay, packet circuit emulation services, and DS0. Also we'll talk a little bit about call processing. Within consolidated transport networking, or CTN, there's really two issues, or two large areas to cover: PBX trunking and branch or remote office accesses. Within PBX trunking, you have PBX trunk pathing, if you're building a enterprise network and you have requirements between the PBXs to provide intelligent voice switching within the network; and then on the branch and remote office access side, virtual switching access, tie line, and off-premises extension transport and alternate packet-type routes. Within PBX trunk pathing, typically what happens is tie lines, leased lines, or some kind of transport using permanent virtual circuits or PVCs, are established between switches and between routers, and they're interworked with the other circuit emulated services with a device called an interworking function, or IWF. So PBXs can directly trunk between themselves in an enterprise environment, or trunk out to the PSTN, or Public Switched Telephone Network. Critical to this, though, when you choose an ATM network, or a frame or packet-based switching network, is adaptive clocking; the ability to modify the delivery or change the synchronization scheme within the network to accommodate for variability in voice delay. So in this diagram, we see from PBX1 inbound cells or from the IWF or interworking function switch one on the left, into the ATM network. Within the ATM network there may be fixed delays such as transport serialization, etc., switch times, etc., and there may also be variability of delay in terms of queuing, and queuing is really the critical piece of it here. So to understand how to deal with the adaptive clocking issue, you need to effectively recover from the variability of delay, although you may have a total amount of delay that is controllable with echo cancellation. So what we use is a reassembly FIFO, first-in/first-out, queue, in the interworking function switch shown here, IWF2, so that the PBX2 receives a constant rate of PCM samples that it can then deliver to the end user. There's another concept within network synchronization called synchronous residual time stamp, or SRTS. What this does is places a time stamp within the ATM cell that is tied to primary reference source one shown in this diagram. So the transmit clock, the receive clock, and primary reference source one are tied together in locked synchronization. There may be primary reference source two provided to the IWF, and this concept allows what we call "plesiochronous," or almost synchronous timing within a network. Alternatively, you can tie everything to the same reference source. In this case we show a primary reference source two, the ATM network IWF switches one and two shown here, and the PBXs, are all tied to the same clock source. There are many methods you can use to do this including loop clocking the PBXs at each end to the ATM switches, or providing a direct external clock source to them. Within PBX trunk pathing, network synchronization can be accomplished by referencing trunk paths rather than direct trunk sources, shown here in this diagram. So for primary reference source two provided directly to the PBX1 shown here, that may be provided through the PSTN to that PBX on what we call an off-premises extension. The CES, circuit emulation service, interworking interface to the switch provides a reference clock from PRS2 to all of the other CES IWFs in this diagram. In that way you can preserve a very high level of synchronization, but not necessarily to the primary reference source or stratum one clocking. It may, however, be sufficient to operate the network. When designing your PBX trunking scheme, there's a few things to consider. The fact that there are N squared connections - every time you add a connection it adds a complexity and a dual - a full-duplex concept in a VC within the network. Also tandem hopping, or hopping between three PBXs, a end device requesting a connection through a second PBX to a third PBX is called a tandem hop. And a tandem hop can cause problems with synchronization for your PBX trunk pathing. Alternatively, you can go into a dedicated point-to-point circuit although that's not as efficient bandwidth-wise, and may be more expensive to implement. And the signaling network if it's in-band would be the same as the trunk pathing network. But if it's out of band, or if it's common channel signaling, then it would be considered to be the point-to-point signaling between switches. Also platform availability. If you have a PBX or some kind of switching device in the network that may be vulnerable to downtime or power outages or isolation due to somebody digging through your trunk groups, the platform availability will determine what kind of reroute options you have and whether those reroute options are acceptable for the network performance that you're looking for. One alternative to PBX trunking is intelligent voice switching. In other words, providing dynamic switched virtual circuits between voice switches. PBX signaling methods that are used for this are DPNSS, ETSI-QSIG, and ISO-QSIG. These are acronyms that will be covered in great detail in Part 2 of this Basics of Telephony presentation. Using intelligent voice network switching allows efficient trunk groups, so you don't have to establish your network in a hard and fixed manner with VCs that aren't always in use. This adds to the efficiency of the WAN utilization as well. PBX to network signaling, tandem switching replacement, dynamic setup of VCs, and also the addition of methods to reduce bandwidth consumption, such as compression which will be covered in Part 2 in great detail, and voice activity detection. The concept that two ends of a voice conversation at any one time, one person is speaking, one person is listening. So the speaker's voice energy should be transported across the network; however, the listener's voice energy should be cut off and provided to - the bandwidth provided to other users within the network. For consolidated transport for branch and remote office access, there's been an explosion of data networking within branch offices, and voice and data traffic patterns are getting to be more and more similar - high utilizations and very different from the public switched network traffic behavior and patterns. There have also been some technology advancements that allow for increased compression and more efficient utilization of bandwidth. So consolidated transport can take many forms. A virtual switch interface from a PBX into the access WAN, may include a Layer 2 switch, it may include a router, it may include an ATM switch, and the access WAN itself may be this physical transport between these devices. So here you see ISDN phones, DTMF phones, and data devices, all supported within the small office, the home office, and the branch, the remote access. The virtual switch access is effectively just a PBX providing trunk groups or off-premise extensions to individual phones or to other small key systems or PBXs shown here. For supportive branch remote office access tie line and off-premises extension transport, shown here we have an access WAN with PBXs and small key systems or smaller PBXs at remote sites. So at each remote site users can call other extensions at that remote site, never actually encountering the PBX or the voice switch, and therefore not using access WAN bandwidth. It's more efficient. The other consideration with consolidated transport is the ability to do alternate routing. If you are placing all of your eggs in one basket in terms of voice transport and data transport, you need to be careful that you have alternatives when that one method is down. If there are any serial path devices that can fail, you need to be able to dial off-net, for example, which is the reference to the switching over a public switched network cloud, switching over a backbone network, in other words an internal alternate path, or switching through your access WAN through your virtual switch interface. All of this, of course, gets down to billing and cost of the network. An example shown here - and this is just a very broad and general example of how consolidated transport networking bandwidth can be used more efficiently and then lower your cost. Here we see on the left the billing cycle cost in dollars, the fixed cost would be tie trunk, or point-to-point meshed circuits between voice switches. Then you have direct dial, DDD in the green there, and that rises quickly, and then the "A" plan which would be a negotiated plan with the carrier; and then there is the consolidated transport networking which is - provided the intelligent switching support is within the network and you have the ability to consolidate bandwidth with compression and voice activity detection, lowers your overall cost and fixes that cost so it's more predictable and easier to budget for and bill for. Also included are things like fax services, for example. Through the Internet, through the PSTN, through access WAN, through your own private network, fax back, fax services, click to fax are examples of fax services that need to be supported and can be supported with the proper intelligent network trunking. So that covers consolidated transport networking. So that concludes our talk on basic analog telephony, basic digital telephony, and consolidated transport networking. Please attend Part 2, which by the way does not appear on the CD. It is a live presentation at Networkers. Part 2 will revisit all of these subjects and go into more detail, drill down a little bit more, and also cover voice coding techniques and compression techniques, voice transport and delay, digital voice signaling techniques, and network planning and design issues. Thank you. </BODY> </HTML>