"Linda J. Crosby" <lcrosby@ST-LOUIS-EMH2.ARMY.MIL> (05/29/91)
I have some questions about Frame Relay....but first let me give you
a quick sketch of our setup:
Setup:
--------- Ethernet 0
| |----------- [mix of Unix, MVS, & PC systems; running
| | Ethernet 1 TCP/IP]
--------| |----------- [Unix systems; running TCP/IP]
Serial 0| CISCO | Serial 1
[56KB to| |----------- [X.25 link MVS/comten-comten/MVS
DDN] | | via DDN]
| | Serial 2
(X.25 STD)| | - - - - - - [ currently vacant ]
---------
T1 link
============================= [ between SIMA East & SIMA West ]
Access/Communicate with: primarily within MILNET domain (.mil), but not
exclusively.
Questions:
1. a. What would Frame Relay do for us?
b. How would it improve our communications with the rest of the
world?
c. Who else is using Frame Relay now?
2. What additional circuits/equipment would we need to implement
Frame Relay?
3. What would the cost/s of circuit/equipment be?
Linda J. Crosby
Technical Liaison
USAMC SIMA
<LCROSBY@ST-LOUIS-EMH2.ARMY.MIL>robelr@mythos.ucs.indiana.edu (Allen Robel) (05/30/91)
> Questions: > 1. a. What would Frame Relay do for us? > b. How would it improve our communications with the rest of the > world? > c. Who else is using Frame Relay now? > > 2. What additional circuits/equipment would we need to implement > Frame Relay? > > 3. What would the cost/s of circuit/equipment be? That's enough to fill several articles. I can recommend a few that answer most of these questions. Also, someone on BIG-LAN provided a writeup that I'll send you directly (if there's enough interest, I'll post it too.) regards, allen ------------------------------------------------------------- TeleNotes. The Frame Relay Solution. US Sprint. Volume 1 Number 2. Concise and amazingly unbiased booklet that discusses benefits, limitations, etc. available at no charge from: TeleNotes Editor, Public Affairs Department Sprint International 12490 Sunrise Valley Drive Reston, VA 22096 Frame Relay Protocols, Standards and Controversies. Charles M. Corbalis. Business Communications Review. March 1991. pp 70-75. The cisco/DEC/NTI/Stratacom Frame Relay Specification. Edward R. Kozel. conneXions (a couple months ago, sorry I don't have the exact date). Frame Relay Networks: Not as Simple as They Seem. Holget Opderbeck. Data Communications, December 1990. (deals with congestion control issues). Some standards documents I'm aware of follow: 1. ANSI T1.606: Frame Relaying Bearer Service--Architectural Framework and Service Description, American National Standards Institute, Inc., 1990. 2. ANSI T1S1/90-175 (Addendum to T1.606) Frame Relaying Bearer Service--Architectural Framework and Service Description, American National Standards Institute, Inc., July 1990. 3. ANSI T1S1/980-214 (T1.6ca): DSSI--Core Aspects of Frame Protocol for Use with Frame Relay Bearer Service. American National Standards Institute, Inc., July 1990. 4. ANSI T1S1/90-213 (T1.6fr): DSSI--Signaling Specification for Frame Relay Bearer Service, American National Standards Institute, Inc., July 1990. "Frame Relay Specification With Extensions", Revision 1.0, Document Number 001-208966, Digital Equipment Corporation, Northern Telecom, Inc., StrataCom, Inc., September 1990 CCITT Recomendation I.122, Framework for Providing Additional Packet Mode Bearer Services, Blue Book, ITU, Geneva 1988
beach@ddnuvax.af.mil (beach) (05/30/91)
Linda, The major prerequisite is to have a network which uses frame relay as an access protocol. Frame relay is more flexible than simply using T-1 pipes to interconnect routers. The reason is that frame relay gives you a mechanism (albeit limited) to individually address multiple end systems on the network while only requiring a single port on the router, i.e. one connection to the network. If you simply use big pipes, then you have to get a separate port for each destination you want to talk to. That gives you a pretty limited fan-out per router, and would increase tandem traffic on intermediate routers. It would also complicate the basic design of connectivity as you increase the number of routers. Frame relay lets you preestablish (key word) what are essentially permanent virtual circuit from one end point to some number > 1 other end points. Which PVC a "frame" uses is determined by an address mechanism using a field in the frame header called the DLCI. Normally a particular DLCI refers to a particular PVC in the network. This isn;t always the case. At least one vendor will let you configure a DLCI to represent an end system, instead of a particular circuit. You can also use a mechanism to discover the IP address of what's on the other end of a DLCI dynamically, as well as find out what DLCIs are available to use, again dynamically. /darrel --------- Darrel Beach beach@server.af.mil snail-mail: SAIC Montgomery Network Systems Engineer 205-279-4075 SSC/SSMT USAF DDN Program Office AV: 446-4075 Gunter AFB, AL 36114
gfw@ihburn.att.com (06/05/91)
Linda, Another reason to use frame relay is to minimize serialization delays. Say for example that you were able to afford only a 56kbps link to a few different places, and also that your traffic is a mix of character (telnet/rlogin) and file transfer (ftp, rcp). The small telnet packets get stuck behind the large file transfer packets (which are being dribbled out at 56kbps). A 1500 byte packet would take 1500*8/56kpbs or 200ms to serialize out the interface. Studies have shown that users are willing to tolerate character echo times of slightly over 200ms if the variance is relatively small. Now considering that character echos get stuck behind large file xfer packets somewhat randomly and that when they do they get slowed down significantly, you can see that not only will the echo times be larger than the 200ms benchmark, but the variance will also be large. This drives users crazy. Frame relay helps a bit here. If you can afford a T1 for local loop (access to the frame relay network from your site), then the serialization times can be made much less, since the frame relay network should be running at T1 speeds. What you have to do then is, given the cost of private line 56kbps service and the costs of a T1 local loop with whatever framerelay charges you're likely to incur, make a decision based on the costs and service levels you can afford. What you're really doing is trading off the costs of a T1 over the full length path with the packet charges of frame relay service to the same destinations (given that the increase in local loop charges, DS0 to DS1, are affordable because of increased performance). This all depends heavily on how much frame relay costs. So this benefit, plus the use of a single outbound interface to interconnect multiple routers (as suggested by Beach) are the major pluses. Greg Wetzel +1 708 979 4782 G_F_Wetzel@att.com AT&T Bell Laboratories (IH 1B-213) 2000 N. Naperville Road Naperville, IL 60566
forster@cisco.com (Jim Forster) (06/06/91)
>> This all depends heavily on how much frame relay costs. So this benefit, >> plus the use of a single outbound interface to interconnect multiple routers >> (as suggested by Beach) are the major pluses. Greg, Linda, and others, Frame Relay has a lot of benefits, especially for non-IP/CLNP public networks, but we not should overstate them. In particular, if you take a given FR net design, and transform it into a router design, simply by replacing all the FR switches with routers, you will end up with a Router net that has all of the same bandwidth/delay properties as the FR net. The advantages that I seen in FR are that it allows completely flexible addresses, because it does level-2 switching. Therefore the backbone equipement can support lots of organizations each of which has the same DECNet Area 1, or Novell IPX Net 1, etc. Each such organization won't be able to communicate with the other organizations due to the addressing problems, but thats usually also consistent with security goals. Protocols with universal address administration, such as TCP/IP & CLNP, don't have this problem. -- Jim
eckert@immd4.informatik.uni-erlangen.de (Toerless Eckert) (06/07/91)
From article <35724@boulder.Colorado.EDU>, by forster@cisco.com (Jim Forster): > >>> This all depends heavily on how much frame relay costs. So this benefit, >>> plus the use of a single outbound interface to interconnect multiple routers >>> (as suggested by Beach) are the major pluses. > > Greg, Linda, and others, > [...] > The advantages that I seen in FR are that it allows completely flexible > addresses, because it does level-2 switching. The other advantage (or disadvantage) is that the frame relay network will most probably be run by a commercial provider for the whole frame relay net, whereas in the case of leased lines you'll have to control all active parts in your net. So it's mostly a different way of splitting the administrative responsibilities. It is not very different from an X.25 net in this respect - you can build up you're own X.25 net by only leased lines, but you really don't want to, you'll only use a PDN if you can't afford the leased lines, or the traffic characteristic does not make leased lines necessary - I think this will compare very well with frame relay, because if you build up your own frame relay net, it will be more expensive due to the frame relay switches that you have to buy (or cisco implements frame relay switching - how about it ?) P.S.: (X.25 has useless overhead for the LAN interconnect purpose, that's why they invented frame relay) --- Toerless.Eckert@informatik.uni-erlangen.de /C=de/A=dbp/P=uni-erlangen/OU=informatik/S=Eckert/G=Toerless bandwidth - the final frontier.
forster@cisco.com (Jim Forster) (06/10/91)
>> The other advantage (or disadvantage) is that the frame relay network >> will most probably be run by a commercial provider for the whole >> frame relay net, whereas in the case of leased lines you'll have >> to control all active parts in your net. So it's mostly a different >> way of splitting the administrative responsibilities. These statements are true, but perhaps aren't complete, as besides the 2 alternatives listed above (Public FR Network and Private leased line & router network), there is a third: Public Router Network. This is what PSI, Alternet, CerfNet, Infonet (Infolan), Finnish PTT, and now ANS are providing. Many PTT's in Europe and Japan are also planning or considering such offerings. In my original posting I was trying to point out that from a purely technical viewpoint, I don't see any Bandwidth/Cost advantages of FR over Router Networks. The addressing advantages are subtle, but sometimes significant. -- Jim
eckert@immd4.informatik.uni-erlangen.de (Toerless Eckert) (06/10/91)
From article <35801@boulder.Colorado.EDU>, by forster@cisco.com (Jim Forster): >>> The other advantage (or disadvantage) is that the frame relay network >>> will most probably be run by a commercial provider for the whole >>> frame relay net, whereas in the case of leased lines you'll have >>> to control all active parts in your net. So it's mostly a different >>> way of splitting the administrative responsibilities. > > In my original posting I was trying to point out that from a purely > technical viewpoint, I don't see any Bandwidth/Cost advantages of FR over > Router Networks. The addressing advantages are subtle, but sometimes > significant. I agree completely. The only real advantage that i see is for non specialised carrier companies, who may have less problems running a frame relay network than a router network - PSI, Alternet and so on are not the typical examples of public data network carriers, and even though there are even european carriers who are already into the router business i do not believe to see a company like german telecom dive into that business. The sad thing is only that there are many arguments that are used in favour of frame relay, that are not advantages of frame relay. For example security is one of the hype arguments for frame relay, as they tell that frame relay can effectively protect different customer groups from each other - well, given the correct software this can be done (much better then) with pure router networks too, but they just don't believe this. On the other hand, the customer has probably to put much more work into configuring his net on a frame relay bearer service compared to a router network that is already based on his protocols, so for a customer going for IP or CLNS it may even be simpler from the point of administration to go for a router network (that's the whole operating theory of public carriers: they only want the money, not the hassle ;-))). Sorry, this whole discussion may be totally misplaced on this group, i'll stop now. --- Toerless.Eckert@informatik.uni-erlangen.de /C=de/A=dbp/P=uni-erlangen/OU=informatik/S=Eckert/G=Toerless bandwidth - the final frontier.
Jim Forster <forster@cisco.com> (06/11/91)
>> The sad thing is only that there are many arguments that are used in favour >> of frame relay, that are not advantages of frame relay. Yes, I agree. >> On the other hand, the customer has probably to put much more >> work into configuring his net on a frame relay bearer service compared >> to a router network that is already based on his protocols, so for a >> customer going for IP or CLNS it may even be simpler from the point >> of administration to go for a router network (that's the whole Yes, I also agree. The difference here, for instance, is that with FR the customer has to manage "N" separate PVC's, while with a Router-Network provider typically they provide an interface directly to the customer Ethernet or Token Ring. >> Sorry, this whole discussion may be totally misplaced on this group, >> i'll stop now. Well, I think we've reached the end (we understand each other's viewpoints and agree!). And while the discussion has not been cisco-specific, I hope it has been interesting and worthwhile to the readers. -- Jim