finn@eleazar.dartmouth.edu (Andy Behrens) (07/21/88)
We will soon have a T1 circuit between two of our sites, but expect to
be using less than half of its bandwidth. We would like to bridge
together two Ethernet networks at these sites.
All the bridges that I have seen either use an entire T1 circuit, or
use only 56Kb of it. Does anyone know of one which can be configured
for data rates between these limits?
Andy Behrens
andyb@burcoat.uucp
{astrovax,linus,harvard,decvax}!dartvax!burcoat!andyb
andyb%burcoat@dartmouth.EDU
andyb%burcoat@dartcms1.BITNET
(603) 448-5000jlw@lznv.ATT.COM (J.L.WOOD) (07/21/88)
In article <9392@dartvax.Dartmouth.EDU>, finn@eleazar.dartmouth.edu.UUCP writes: > We will soon have a T1 circuit between two of our sites, but expect to > be using less than half of its bandwidth. We would like to bridge > together two Ethernet networks at these sites. > > All the bridges that I have seen either use an entire T1 circuit, or > use only 56Kb of it. Does anyone know of one which can be configured > for data rates between these limits? > I would recommend that you investigate using ATT's ISN for this purpose. One if ISN's features is a MAC-layer bridge called Ethernet Bridge. This can be combined with an ISN packet controller at your base location and a remote concentrator at your remote location. In addition you need to add a device called the Channel Division Multiplexor. This device will split up the T1 channel into N voice channels where 1 <= N <= 20 and the remainder is a single bit synchronous data channel with a V.35 I/F. This is ideal for the wire trunk card on the ISN which adapts to the external bit rate and can vary from 9.6Kbps to 2.048Mbps depending on traffic needs. As an added bonus you also get STARLAN, synchronous, asynchronous, Host-Host (UNIX SV or DEC VMS), and 3270 switching capability at both locations with shared bandwidth with the Ethernet Bridge. Contact your local ATT Business Office. Joe Wood lznv!jlw
wunder@hp-sde.SDE.HP.COM (Walter Underwood) (07/22/88)
All the bridges that I have seen either use an entire T1 circuit, or use only 56Kb of it. Does anyone know of one which can be configured for data rates between these limits? T1 equipment almost always uses an external clock. Good T1 equipment still works when you change the clock rate, so actually running at .75122 Mbits shouldn't be a problem. 2:1 muxes are pretty common for 56Kbit lines, and probably exist for T1. Some of the muxes are also compression boxes, so you might look around for compression equipment that takes two or more channels in and puts out a T1. Some compression boxes do flow control by varying the clock. If they want the source to slow down, they slow it down. It is a good idea to test equipment to see if it can handle that sort of thing. At HP, I believe that we use cisco boxes through Datamiser (Datamizer?) compression boxes that use the clock for flow control. wunder
peter@ernie.NECAM.COM (Peter DiPrete) (07/22/88)
In article <9392@dartvax.Dartmouth.EDU>, finn@eleazar.dartmouth.edu (Andy Behrens) writes: > > All the bridges that I have seen either use an entire T1 circuit, or > use only 56Kb of it. Does anyone know of one which can be configured > for data rates between these limits? This question requires a little knowledge about how 56 Kb circuits and T1 circuits (actually called facilities) are related. A T1 (sometimes called DS1) facility is composed of 24 multiplexed 56 Kb (called T0 or DS0) circuits. If your application will use less than half of the T1 facility, you might consider a T1 mux and use as many T0 circuits out of the 24 that you feel you require for the particular application. Of course, the additional hardware costs money, and there may be better ways to use the bandwidth. Bridge communications, now part of 3Com, has a wide variety of products that will allow you to bring 56 Kb INTO the LAN you are shipping via the T1 line and share it that way. And since the Bridge boxes are effectively "packet switches," this might make the MOST effective use of the lan. It would also let you attach PC's and other equipment to the lan and share them long distance. We are looking at Bridge IB/3's to connect three LANs, one in Japan, one in Virginia, and one in San Jose! I'll let you all know how it works out. Peter Di Prete NEC America ...!uunet!altnet!ernie!peter ...!sun!imagen!ernie!peter ...!sun!snail!maui!ernie!peter
donegan@stanton.TCC.COM (Steven P. Donegan) (07/22/88)
In article <9392@dartvax.Dartmouth.EDU>, finn@eleazar.dartmouth.edu (Andy Behrens) writes: > > All the bridges that I have seen either use an entire T1 circuit, or > use only 56Kb of it. Does anyone know of one which can be configured > for data rates between these limits? The Bridge Communications IB/3 is capable of operating at a wide variety of speeds. I have used a pair (at 56k) for about a year and they are working flawlessly. The price is steep though, 13-14k ea! You could get better price/performance in the sub T1 range with ACC 4030's using 2 56k links for each pair and use multiple pairs between sites as desired. The 4030's support the spanning-tree protocols to allow multiple parallel links between LAN segments (which could provide path/media diversity for disaster recovery). The 4030 costs around 5k ea. so 224kb worth of internet bridgeing would cost in the range of 20k as compared to 28k for Bridge. -- Steven P. Donegan These opinions are given on MY time, not Sr. Telecommunications Analyst Western Digital's Western Digital Corp. stanton!donegan || donegan@stanton.TCC.COM || donegan%stanton@tcc.com
hedrick@athos.rutgers.edu (Charles Hedrick) (07/23/88)
>All the bridges that I have seen either use an entire T1 circuit, or >use only 56Kb of it. I have never heard of a bridge that actually expects T1-formatted data. In general bridges and routers expect a standard synchronous line. If you have a 56K circuit, you just connect them to the modem, CSU/DSU, or whatever. If you have a T1 circuit, things are more complex. T1 is not just a 1.5 Mbps line. It has a fancy format of its own, which is really designed to multiplex multiple 56K subchannels. To use a whole T1 circuit with a bridge or router, you have to use a box that takes a generic 1.5 Mbps signal and puts it into the funny T1 format. I call such a box a T1-izer. If you want to do several different things with your T1 line, then instead of a simple T1-izer, you use a multiplexer. This can take several generic signals, whose bandwidth totals 1.5 Mbps or less, and multiplexes them onto a T1 line. In general the bridges and routers don't know and don't care exactly how fast a line is. They just lock onto the clock that comes from the T1-izer of multiplexer. You can use them with a T1-izer if you want to use the whole bandwidth of your line, or you can hook them to one channel of a T1 mux if you want to share the line. This certainly applies to Ungermann-Bass and cisco equipment, and I'm reasonably sure it applies to other vendors as well. Cisco's high-speed serial card will work at any speed from 9600 to 3Mbps (in theory, though I don't know of anyone who has the equipment to use it above T1 speeds). The only way to build a bridge that would work with a whole T1 but not with a piece of a T1 would be if the bridge dealt with the T1 formatting itself. I've never heard of such a beast.
David@cup.portal.com (07/24/88)
VitaLink makes MAC-layer bridge products capable of running at speeds between 56k and T1, for instance, 256k, 512k, 772k, etc. The trick is to obtain the 512k circuit. Once you have the appropriate speed circuit, the bridge runs at that rate. The easiest way to do this is to use a T1 CSU device such as the SCITEC "Saturn" product. This mates the 1.544MHz T1 circuit to a selectable lower speed circuit while maintaining things like pulse density according to the rules of T1. The only drawback of this kind of device is that you can only put one channel of data over the T1, so if you are using a 512k circuit for your bridge, you waste the remaining bandwidth (a megabit or so). If you want to derive more than one channel from the T1, look at the various T1 mux vendors such as NET, DCA, GDC, etc. David@cup.portal.com
ron@topaz.rutgers.edu (Ron Natalie) (07/25/88)
I'm not sure what you're saying. Most Ethernet bridges that work over synchronous lines will adapt to what ever speed you give them. They'll run at whatever the clock speed provided them is. -Ron
ron@topaz.rutgers.edu (Ron Natalie) (07/25/88)
> A T1 (sometimes called DS1) facility is composed of 24 multiplexed 56 > Kb (called T0 or DS0) circuits. If your application will use less > than half of the T1 facility, you might consider a T1 mux and use as > many T0 circuits out of the 24 that you feel you require for the > particular application. Bogus. A T1 circuit is a serial line of 1,544,000 bits per second. It is not "composed of" anything. The frequent telephone use is to use a channel bank that decomposes it into 24 56K data channels or even more voice or telegraph channels. T1 Multiplexors are available that will carve up the bandwidth to almost any specification. For example, our AVANTI Ultramux allows me to reallocate how much bandwidth I give to each of my interfaces at whim. Of course, it isn't cheap. If you are not using the T1 for anything else, you do not need to waste money on a Multiplexor. For less money, Digital Link provides a true T1 CSU/DSU, that allows you to plug a Bridge such as the UB DLB or the Bridge Communications (Vegalink box) into the T1 line directly. We use them for Cisco gateways because we find MAC level bridges do not provide sufficient fault isolation in a network as diverse as ours. -Ron
nlp@vu-vlsi.Villanova.EDU (Nick Pine) (07/26/88)
Hi Andy!
Aydin Monitor Systems/Private Networks (215-646-8100) for whom I sometimes
work, also makes T-1 muxen. And of course Pine Associates, Ltd., soon to open
a Grenoble, France office, will design and build almost anything for an
extraordinary price :-).
Nick
PS: I believe the CCITT T-1 frequency tolerance spec is +/- 75 Hz. or 50 ppm.,
and most clock recovery circuits, which use resonant circuitry or phase-
locked loops, would not work if you turned the clock down 50%. The easiest
way to send arbitrary data and still meet the 1's density requirement is
to make every 8th bit a 1, and give up 12% of the bandwidth, or use "B8ZS"
signaling, if available from the carrier.pst@comdesign.uucp (Paul Traina) (07/26/88)
From article <7632@cup.portal.com>, by David@cup.portal.com: > If you want to derive more than one channel from the T1, look at the various > T1 mux vendors such as NET, DCA, GDC, etc. A handy thing about T1 muxes is that you can put anything you want over the circuit. Therfore, for a distributed company, you can nix your phone bills, wipe out all your leased lines et al between sites and just pay for one T1 line. Also, in a business park environment, groups of companies could get together and lease "chunks" of someone else's T1 line -- much cheaper (but much more limiting) than purchasing subrate (<T1) lines. Disclaimer: T1 muxes are what my company does. I don't speak for them, I don't tell them what to do with what I make (in fact, I don't deal with T1 stuff at all). Anything you hear from me is guesswork & unofficial. -- Paul Traina - {uunet|pyramid}!comdesign!pst - comdesign!pst@pyramid.com To believe that what is true for you in your private heart is true for all men, that is genius.
howard@cos.com (Howard C. Berkowitz) (07/29/88)
In article <Jul.25.11.48.50.1988.21899@topaz.rutgers.edu>, ron@topaz.rutgers.edu (Ron Natalie) writes: > > A T1 (sometimes called DS1) facility is composed of 24 multiplexed 56 > > Kb (called T0 or DS0) circuits. If your application will use less > > than half of the T1 facility, you might consider a T1 mux and use as > > many T0 circuits out of the 24 that you feel you require for the > > particular application. Actually, the original derivation was 24 64KBPS channels, into each of which could be inserted one digital voice channel or approximately 56 KBPS of data. T1 systems replaced two N systems, an analog carrier system which carried twelve voice channels. The 1.544 MBPS was derived as follows: A normal telephone voice channel has 4kHz of analog bandwidth. An analog channel can be digitized by sampling it at twice its maximum frequency, thus 8000 BPS. At each sample, the amplitude is converted to a digital value. Originally, this was a 7 bit (i.e., 128 level) code with the 8th bit used for telephone signaling; the quality was sufficiently low that the code was changed to 8 bits and other methods used for signaling. 8000 * 8 = 64000; 64000 * 24 = 1544000 Again for purists, the encoding process is done by a D-type channel bank, the OUTPUT of which is a DS1 signal. > > Bogus. A T1 circuit is a serial line of 1,544,000 bits per second. > It is not "composed of" anything. The frequent telephone use is to > use a channel bank that decomposes it into 24 56K data channels or > even more voice or telegraph channels. It is quite true that a T1 circuit provides 1,544,000 bits per second, but this can be misleading because all those bits are not normally available. There are several constraints on bit patterns which can run through the line. The original constraints are due to the way that clocking information is carried in T1 signals. DS1 (really more proper than T1; T1 is specifically a Bell System code for DS1 signals carried on repeatered twisted pair) is an _isochronous_ signal, different from synchronous (i.e., separate clock) or asynchronous (i.e., carrying in-band explicit timing bits). Isochronous signals carry in-band implicit timing information. The reason for the clock constraints are that a certain average number of pulses per unit time are needed to assure that clock can be recovered with the signal. A common analogy is that the average pulse density keeps a "flywheel" spinning in the receiver, and this flywheel (actually a tank circuit or phase lock loop) generates local timing. The original T1 system used repeaters every 6000 wire feet on 24 AWG or larger cables; these repeaters replaced the loading coils previously used every 6000 feet, so no new splices were needed on existing cables. These repeaters needed the following rules to be followed to assure adequate zero-crossing signals: 1. At least 3 one bits in every 24 time slots 2. No more than 15 consecutive zeroes 3. One bits sent with "alternate mark inversion:" zero bits are of zero voltage, but ones are sent with alternating polarity above or below zero. Two consecutive bits of the same nonzero polarity are a "bipolar violation," usually considered an error but in fact used for local loop control in DDS customer-premises-to-central-office links. Additional constraints have been added for network management and signal quality monitoring. I don't remember all of the details, but, from memory, the Extended Superframe Format calls for something like every 2047th bit to be a one. Someone with the ESF references handy can correct this. The point of all this is that a user who expects to pump 1.544 MBPS of live data into a DS1 facility will be sorely disappointed, because, unless significant signal scrambling is done, the transmission network will stuff bits into the data in order to guarantee the network stays synchronized. In applications where 56KBPS data is being sent, it's quite easy to guarantee appropriate signals by forcing every 8th bit to be a one, shifting the speed from 56 to 64 KBPS. Where higher speeds are needed, more processing is required. There are rumors, for example, that NSA crypto gear operating into DS1 facilities have their encryption algorithms designed to meet DS1 pulse density criteria. Normally, however, there are no more than 1.344 to 1.536 MBPS available for transparent data use. Incidentally, the higher speeds of the DS hierarchy have even more constraints. DS2 signals are not an integral multiple of the DS1 rate, nor is DS3, because bit stuffing is used for speed matching and synchronization. -- howard@cos.com OR {uunet, decuac, sun!sundc, hadron, hqda-ai}!cos!howard (703) 883-2812 [W] (703) 998-5017 [H] DISCLAIMER: Opinions expressed are not necessarily those of the Corporation for Open Systems, its members, or any standards body.
jlw@lznv.ATT.COM (J.L.WOOD) (07/31/88)
Actually there is a nice little trick you can pull to meet the ones density rules for a DS1 circuit. If you are running HDLC or SDLC framing on your data all you need to do is invert the sense of the ones and zeroes and the HDLC bit stuffing takes care of it. This can easily be done if you are using a V.35 (balanced type) interface by frogging the TD and RD pairs at each end of the cable. Joe Wood lznv!jlw
chip@vector.UUCP (Chip Rosenthal) (07/31/88)
In article <4399@cos.com> howard@cos.com (Howard C. Berkowitz) writes: >The original T1 system used repeaters every 6000 wire feet on >24 AWG or larger cables; They still are. That's why the suggestions of running something other than 1.544MHz seem a little silly to me. There are two types of T1 lines: private and public. You can use any kind of framing format you want on private lines, but in either case you must meet the signal requirements of the FCC and your carrier. AT&T requires 1.544MHZ+-50ppm. I believe ANSI (and possibly Belcore) is moving to 30ppm. >These repeaters needed the following >rules to be followed to assure adequate zero-crossing signals: > 1. At least 3 one bits in every 24 time slots > 2. No more than 15 consecutive zeroes > 3. One bits sent with "alternate mark inversion:" Close. The actual one's density requirement is no more than 15 consecutive zeros, and in every window of (N+1)*8 bits there must be N ones, where N is between 2 and 23 (or is that 24?). >Additional constraints have been added for network management >and signal quality monitoring. I don't remember all of the >details, but, from memory, the Extended Superframe Format calls >for something like every 2047th bit to be a one. Someone with >the ESF references handy can correct this. OK...the ESF framing format replaces D4. Both of these formats group 193-bit frames into superframes. ESF has 24 frames per superframe, D4 has 12. Under D4, the 12 framing bits a fixed pattern for alignment to the superframe. Under ESF, the 24 framing bits contain 6 bits for frame alignment, 6 bits as a CRC on the previous superframe, and 12 bits as a facilities data link (FDL). The FDL is what was referred to. It is a 4Kbps communication channel for configuration commands (loopback, yellow alarm, etc.) and status reports (errored seconds, severely errored seconds, etc.). But, the electrical signal requirements are no different between ESF and D4. Unfortunately...I don't think this helps the original poster who wanted to connect to networks together. I don't know offhand if any muxes allow you to allocate several channnels for a data channel. If you are interested in making some phone calls, I could probably pull together a list of mux vendors from my files. As a side issue, this is one of the Great Promises of ISDN. Under 23B+D primary rate, you are supposed to be able to allocate as many 64Kbps B channels as required for your data stream. -- Chip Rosenthal /// chip@vector.UUCP /// Dallas Semiconductor /// 214-450-0400 {uunet!warble,sun!texsun!rpp386,killer}!vector!chip I won't sing for politicians. Ain't singing for Spuds. This note's for you.
howard@cos.com (Howard C. Berkowitz) (08/01/88)
In article <475@vector.UUCP>, chip@vector.UUCP (Chip Rosenthal) writes: > In article <4399@cos.com> howard@cos.com (Howard C. Berkowitz) writes: > >The original T1 system used repeaters every 6000 wire feet on > >24 AWG or larger cables; > > They still are. While most T1 systems do run over 24 AWG or larger, this is not true of new construction in many metropolitan areas such as Washington, DC. C&P Telephone installs many of their new T1 lines on 26 AWG twisted pair with repeaters every 4000 wire feet; this saves copper and space in crowded metropolitan ducts. Note that there are products for sending DS1 signals on media other than twisted pair: fiber, free-space infrared, etc. >That's why the suggestions of running something other > than 1.544MHz seem a little silly to me. There are two types of T1 lines: > private and public. You can use any kind of framing format you want on > private lines, but in either case you must meet the signal requirements > of the FCC and your carrier. I hope I did not interpret anyone's suggestion as running a raw signaling rate other than 1.544 MBPS; not doing so would make it extremely difficult to use commercial test equipment, etc. > Unfortunately...I don't think this helps the original poster who wanted > to connect to networks together. I don't know offhand if any muxes allow > you to allocate several channnels for a data channel. If you are interested > in making some phone calls, I could probably pull together a list of mux > vendors from my files. While it's been several years since I built DS1 mux systems, vendors which allowed you to split a DS1 into several arbitrary streams above 56 KBPS included [NOTE THE DISCLAIMER! THIS DOES _NOT_ IMPLY _ANY_ CORPORATION FOR OPEN SYSTEM EVALUATION!] General Datacomm and Timeplex. As a side issue, this is one of the Great Promises > of ISDN. Under 23B+D primary rate, you are supposed to be able to allocate > as many 64Kbps B channels as required for your data stream. Since the D channel is 16 KBPS and the B is 64, the D channel does make 48 KBPS available for monitoring and bit stuffing for pulse density. -- howard@cos.com OR {uunet, decuac, sun!sundc, hadron, hqda-ai}!cos!howard (703) 883-2812 [W] (703) 998-5017 [H] DISCLAIMER: Opinions expressed are not necessarily those of the Corporation for Open Systems, its members, or any standards body.