vaf@Valinor.Stanford.EDU (Vince Fuller) (01/31/91)
Does anyone have any good estimates for what the maximum number of serial lines, both T-1 and 56KB, a single CSC-3 AGS or MGS system (with one Ethernet) can be expected to handle? At present, we limit our routers to 12 56KB ports (on 3 SCI's) or 4 T-1 ports (on 2 MCI's) and don't believe we are suffering any bottlenecks. If we assume that we want to allow each T-1 the potential to run at full-speed, we're probably limited to 6 ports (6x1.5MB = 9MB) by the speed of the Ethernet - is this considered excessive? In the case of 56KB ports, I expect that we're more limited by physical space for appliques and cable routing than anything else - is that also a reasonable assumption? Thanks, --Vince
aggarwal@nisc.jvnc.net (Vikas Aggarwal) (01/31/91)
>> at full-speed, we're probably limited to 6 ports (6x1.5MB = 9MB) by the speed >> of the Ethernet - is this considered excessive? In the case of 56KB ports, I Just a reminder that each T-1 is full duplex and the potential flow at a T-1 interface can be 1.5x2 Mb/s at full capacity. However, if you are considering the ethernet only, you pretty much still get 6 T-1's - we took the Multibus speed into consideration for average sized packets and pretty much got the same value. In fact, I use a figure of 6 T1's and no ethernet, or 5 T-1's and one ethernet (if you construct a model and calc the worst case, you would end up with what I just mentioned). -vikas vikas@jvnc.net (609) 258-2403 Network Engineering -2400 JvNCnet, Princeton, NJ -----------------------------------------------------------------------------
ssw@cica.cica.indiana.edu (Steve Wallace) (01/31/91)
In <31893@boulder.Colorado.EDU> vaf@Valinor.Stanford.EDU (Vince Fuller) writes: >Does anyone have any good estimates for what the maximum number of serial >lines, both T-1 and 56KB, a single CSC-3 AGS or MGS system (with one Ethernet) >can be expected to handle? At present, we limit our routers to 12 56KB ports >(on 3 SCI's) or 4 T-1 ports (on 2 MCI's) and don't believe we are suffering any >bottlenecks. If we assume that we want to allow each T-1 the potential to run >at full-speed, we're probably limited to 6 ports (6x1.5MB = 9MB) by the speed [stuff deleted] Remember, a T1's 1.5Mbs is full duplex, ethernet is not. There's also no contention for the media. I'd guess that 3 T1s = 1 ethernet. Steven Wallace ssw@silver.ucs.indiana.edu
bri@jake.tmc.edu (Brian Holmes) (01/31/91)
In article <31893@boulder.Colorado.EDU> vaf@Valinor.Stanford.EDU (Vince Fuller) writes: >If we assume that we want to allow each T-1 the potential to run >at full-speed, we're probably limited to 6 ports (6x1.5MB = 9MB) by the speed >of the Ethernet - is this considered excessive? Since the back-plane of an AGS+ is 500MBps, 6 T1 ports is probably a pretty light load. In the latest price list I have they have a layout with an AGS having 32 v.35 interfaces. -- PHONE: (313) 577-3750 FAX=577-5626 Wayne State University BITNET: bholmes@waynest1.bitnet 5925 Woodward INTERNET: bri@jake.cc.wayne.edu Detroit, MI 48202 U.S.A
tjh+@andrew.cmu.edu (Tom Holodnik) (02/02/91)
> Excerpts from netnews.comp.dcom.sys.cisco: 31-Jan-91 Re: Recommended > maximum num.. Brian Holmes@jake.tmc.ed (673) > In article <31893@boulder.Colorado.EDU> vaf@Valinor.Stanford.EDU (Vince > Fuller) writes: > >If we assume that we want to allow each T-1 the potential to run > >at full-speed, we're probably limited to 6 ports (6x1.5MB = 9MB) by the > speed > >of the Ethernet - is this considered excessive? > Since the back-plane of an AGS+ is 500MBps, 6 T1 ports is probably > a pretty light load. In the latest price list I have they have a > layout with an AGS having 32 v.35 interfaces. Do you mean "an AGS+ having 32 v.35 interfaces?" I haven't seen the latest hardware offerings from cisco. Are they shipping a T1 card for the high speed (CBus) backplane? Seems to me that they should. Utilizing the CBus backplane, rather than the Multibus, would give one chassis greater capacity to handle T1 lines. -tom
bri@jake.tmc.edu (Brian Holmes) (02/02/91)
In article <8beOPyG00WCo8ZMgZb@andrew.cmu.edu> tjh+@andrew.cmu.edu (Tom Holodnik) writes: > >Do you mean "an AGS+ having 32 v.35 interfaces?" > > I haven't seen the latest hardware offerings from cisco. Are they >shipping a T1 card for the high speed (CBus) backplane? Seems to me that >they should. Utilizing the CBus backplane, rather than the Multibus, >would give one chassis greater capacity to handle T1 lines. > >-tom Actually they have a picture with 36 v.35 interfaces. I'm not sure about the CBus T1 card or not. -- PHONE: (313) 577-3750 FAX=577-5626 Wayne State University BITNET: bholmes@waynest1.bitnet 5925 Woodward INTERNET: bri@jake.cc.wayne.edu Detroit, MI 48202 U.S.A
rob@idacom.uucp (Rob Chapman) (02/04/91)
> at full-speed, we're probably limited to 6 ports (6x1.5MB = 9MB) by the speed > of the Ethernet - is this considered excessive? As a general rule of thumb, I usually guage an Ethernet to run at a maximum of 2/3's of its rated bandwidth, due to collisions and overhead. Also its traffic flow is non-deterministic indicating some elasticity is required to maintain this rate. Rob