eli@spdcc.COM (Steve Elias) (08/23/88)
In <1988Aug22.211807.10280@utzoo.uucp> henry@utzoo.uucp (Henry Spencer) writes: >In article <1698@spdcc.COM> eli@spdcc.COM (Steve Elias) writes: >> 802.3 won't provide a deterministic network access time if >> it is under heavy load, as you say. this is a key feature >> of 802.4, i think. not so important for campus type >> networks, but essential for factory & industrial networks. > >But 802.4 cannot provide a (very) deterministic network access time either, >because its token can get lost. I don't know how many times it is necessary >to say this: there is *no* general-purpose local network (that I know of, >anyway) that is free from pathological misbehavior under stress. The kind >of misbehavior, and the kind of stress, differ from one network to another, >but there's no royal road to networking. no royalty intended... i'm not saying that there is any perfect or optimal network for all situations. it seems to me that different networks are going to require different optimizations and different reactions to stress. one idea of 802.4 is to be have a "more" deterministic network access time, apparently. i really don't know if this works in practice, though i suppose it probably does, since there are factory networks using .4. the point being that there is a need for a few different LAN standards. whether having both a token bus and a token ring standard is useless, i'm not qualified to say. but as i learn more about the 802 cornucopia, i'm beginning to see the diffent slants of the two approaches (.4 and .5). perhaps having both specs is just another compromise. did IBM have something to do with this? does anyone have any idea as to why .5 implementations started at 1 or 4 Mbits/sec, and .4 is already chugging along at 10 Mbits/sec? and when is the 16 Mbits/sec token ring chipset supposedly going to be available.
henry@utzoo.uucp (Henry Spencer) (08/26/88)
In article <1715@spdcc.COM> eli@spdcc.COM (Steve Elias) writes: > ... one idea of 802.4 is to be have a > "more" deterministic network access time, apparently. i > really don't know if this works in practice, though i suppose > it probably does, since there are factory networks using .4. Careful: all you can infer from the use of 802.4 is that it works. Not that it works better than 802.3 would. -- Intel CPUs are not defective, | Henry Spencer at U of Toronto Zoology they just act that way. | uunet!attcan!utzoo!henry henry@zoo.toronto.edu
eli@spdcc.COM (Steve Elias) (08/29/88)
In <1988Aug25.170927.4497@utzoo.uucp> henry@utzoo.uucp (Henry Spencer) writes: >In article <1715@spdcc.COM> eli@spdcc.COM (Steve Elias) writes: >> there are factory networks using .4. > >Careful: all you can infer from the use of 802.4 is that it works. Not >that it works better than 802.3 would. i'm surprised that you concede even this point, Henry! media considerations can make 802.4 more attractive in some cases. i think some factories run video information across the same cables which carry the token bus traffic -- to keep an eye out for errant robots & parts. of course, ethernet can run over broadband cable as well -- but the distance limitations persist. any comments about broadband cable plants and standards ? this brings up another big difference between 802.4 and 802.5 -- .5 is baseband, while .4 is broadband...
wallace@cme-durer.ARPA (Evan Wallace) (09/02/88)
In article <1698@spdcc.COM>, eli@spdcc.COM (Steve Elias) writes: > 802.3 won't provide a deterministic network access time if > it is under heavy load, as you say. this is a key feature > of 802.4, i think. not so important for campus type > networks, but essential for factory & industrial networks. Deterministic access time is not necessarily, essential. We had a horizontal milling workstation (or work cell) which operated quite well over a V1 Ethernet. We find that high throughput is an important quality in a factory LAN. We have embraced subnetting as a means to this end, which can be implemented quite cheaply with Ethernet (Ethernet in a box or thin Ethernet). Thus I agree with Steve eliases later comments that bridges and routers will become more important in the future. Particularly as smaller users fight the monolithic LAN concepts that certain large companies have pushed in the past. In reference to 802.3 distance limitations for LANs spanning multiple buildings Steve Elias continues: > a 802.3 to 802.4 bridge could be useful > in this type of situation. nowadays, people seem to be using > remote bridges (nowhere near 10 M bits/sec), or microwave links > and LANBridges (expensive?). theoretically, there is no > distance limitation for 802.4; miles of cable are expensive, though. Another approach to this problem is to use broadband 802.3 to link building sized baseband subnets. Some vendors combine a filtering bridge function with their broadband MAUs. Functionally we have an example of this configuration running at our site. > ... what are the max bit rates of 802.4 and 802.5?? at least > 10 Mbps !? > (i'll check on this, too.) Bit rates for combined data & MAC overhead are: (old)802.4 broadband 5 Mbps 802.4 broadband 10 Mbps 802.4 carrierband 5 Mbps 802.5 (available now) 4 Mbps 802.5 (sometime) 16 Mbps My copy of the 802.5 standard actually lists a 1Mbps rate (amazing). Politics have certainly been involved in the development of the 802 standards, partially motivated by marketing concerns. Some examples of the results of this are an 802.3 standard incompatible with Ethernet V2, and the existance of the 802.5 standard. But politics in standards is not unique to IEEE, it's just one of the things that makes creating good standards so difficult. Evan K. Wallace @ the National Institute for Standards and Technology (formerly the NBS) UUNET uunet!cme-durer!wallace ARPANET wallace@durer.cme.nbs.gov <-yes it's still .nbs