- Test - lists.nanog.org

RE: Cisco Vip Cards.
by Steve Martin 11 Dec '97

11 Dec '97

Be careful. Cisco recognizes that VIP2-40's with OC3-PA's and any other PA co-habitating the VIP card will cause problems (packet loss under peak conditions). This is due to the huge overhead the SARing takes on the VIP's CPU. Cisco will not support multiple PA's within a VIP2-40. This is the primary reason they came out with the VIP2-50. Cheers, Steve -----Original Message----- From: Bradley Reynolds [SMTP:brad@b63695.student.cwru.edu] Sent: Monday, December 08, 1997 6:59 PM To: rad(a)mail.usld.net Cc: nanog(a)merit.edu Subject: Re: Cisco Vip Cards. > I just have a quick qustion. > I'm in the process of determining weather or not to purchase several > VIP cards and modules for a few 7513's I have at remote locations > accross the United states. I spoke with some friends I have at cisco in > referance to known related problems with the VIP cards (according to > cisco there were close to non) but I am interested in some input from > people who are already using VIP cards on 7513's. > > Thanks, > Rob > > We have VIP2-40s running all over our network with various port adapters (OC3, 8port serial etc) and have had no problem besides flakiness when we first implemented them. Everything is working well now. brad reynolds ber(a)cwru.edu brad(a)iagnet.net

5 4

handling large numbers of EBGP peers
by Rob Liebschutz 11 Dec '97

11 Dec '97

I've got a router on Mae West with an open peering policy that has close to 70 peers. I'm currently running gated on a P133 box and have just increased the size of data structures as the number of peers grew. We are looking at going to a Cisco 7200 (with NPE200) or 7500 (with RSP4), but I haven't been able to determine weather these routers will handle this many peers and still have room for expansion. I've heard that Cisco IOS has a software limit, even if the RSPs can handle it, but I haven't been able to tell what it was. Are there routers from other vendors out there handling large numbers of peers? Our bandwidth is too low to want to consider putting a second router on the NAP. I've considered providing ebgp multihop peering to other routers behind this one, but I haven't heard of others doing this sort of thing on the NAP's before. Of course, there are the route servers, but I'm more interested in offering direct peering to anyone that wants it. I only have full routing tables coming in at 2 places in my network and might anticipate having 8 - 10 IBGP peer sessions in addition to external sessions. Rob

5 4

RE: Cisco Vip Cards.
by Steve Martin 11 Dec '97

11 Dec '97

Agreed. I put a known high hitter in one slot and a low use segment in another. -----Original Message----- From: Randy Bush [SMTP:randy@psg.com] Sent: Wednesday, December 10, 1997 10:48 PM To: Steve Martin Cc: nanog(a)merit.edu Subject: RE: Cisco Vip Cards. > I use dual FE and FDDI-PA's all the time in my VIPs. :) They will not operate reasonably under full load. You are over the bandwidth budget of the PCI. randy

1 0

Re: Cisco Vip Cards.
by Michael Gibson 11 Dec '97

11 Dec '97

We have multiple VIP2-40, but the only problems were running PA-2H with other cards. Until CISCO pulled them from market. Also, the PA-OC3 only be the only card in a VIP2-40. Mike Gibson -----Original Message----- From: Randy Bush <randy(a)psg.com> To: Steve Martin <steve(a)conxion.net> Cc: north american nominated origin of gossip <nanog(a)merit.edu> Date: Wednesday, December 10, 1997 3:45 PM Subject: RE: Cisco Vip Cards. >> Cisco will not support multiple PA's within a VIP2-40. > >I do not believe this statement to be true. > >randy >

1 0

RE: Cisco Vip Cards.
by Randy Bush 11 Dec '97

11 Dec '97

> I use dual FE and FDDI-PA's all the time in my VIPs. :) They will not operate reasonably under full load. You are over the bandwidth budget of the PCI. randy

1 0

[summary] charge for bgp feeds?
by Hui-Hui Hu 11 Dec '97

11 Dec '97

thanks for all the responses! it was very helpful, and I hope you guys will keep me posted if anything drastic happens or to correct any inaccuracies. I know that bgp feeds often take a lot of man-hours to deal with incompetent customers, but hope that more nsps might consider waiving the fee for those who know what they are doing. ++cluefactor; disclaimer: often charging for bgp feeds is handled on a case by case basis and sometimes "no" only reflects an individual experience with the provider. e-mail addresses not cited because of spam concerns -h MCI no UUNET they didn't used to. maybe now they do? Sprint yes? BBN $1000 setup Genuity not anymore (at least they didn't charge me :) CERFnet no AT&T no Netcom $750 setup, $300/month SAVVIS no Priori no Verio NE no? CRL no DataXchange no iSTAR CAD$200 setup, CAD$50/month UUNET UK yes Internex $500/mo Best $500 setup (still true?) ENI yes, but often waived for big customers with clues IAGnet no

1 0

Errors from a 7513..
by Adrian Chadd 11 Dec '97

11 Dec '97

%SYS-2-GETBUF: Bad getbuffer, bytes= 40470 -Process= "<interrupt level>", ipl= 1 -Traceback= 60159874 601C5FEC 601C8E18 601CB450 60196D00 %SYS-5-CONFIG_I: Configured from console by vty1 (203.55.84.2) %SYS-2-GETBUF: Bad getbuffer, bytes= 38198 -Process= "<interrupt level>", ipl= 1 -Traceback= 60159874 601C5FEC 601C8E18 601CB450 60196D00 %SYS-2-GETBUF: Bad getbuffer, bytes= 62691 -Process= "<interrupt level>", ipl= 1 -Traceback= 60159874 601C5FEC 601C8E18 601CB450 60196D00 %SYS-2-GETBUF: Bad getbuffer, bytes= 38118 -Process= "<interrupt level>", ipl= 1 -Traceback= 60159874 601C5FEC 601C8E18 601CB450 60196D00 %LINEPROTO-5-UPDOWN: Line protocol on Interface Serial12/6, changed state to down %LINEPROTO-5-UPDOWN: Line protocol on Interface Serial12/6, changed state to up %SYS-2-GETBUF: Bad getbuffer, bytes= 60021 -Process= "<interrupt level>", ipl= 1 -Traceback= 60159874 601C5FEC 601C8E18 601CB450 60196D00 %SYS-2-GETBUF: Bad getbuffer, bytes= 56788 -Process= "<interrupt level>", ipl= 1 -Traceback= 60159874 601C5FEC 601C8E18 601CB450 60196D00 I have been getting pages of these lately. Its a 7513 w/ dual RSP2 modules and 64mb RAM. Its running IOS 11.2(8) P w/ 1 SSIP controller (8 Serial). 1 VIP2 controller (1 FastEthernet)(4 Serial). 1 FastEthernet/IEEE 802.3 interface(s) 12 Serial network interface(s) Has anyone on the list come across this before? Adrian

4 3

RE: Cisco Vip Cards.
by Steve Martin 11 Dec '97

11 Dec '97

Correct. :) -----Original Message----- From: Tom Eastgard [SMTP:eastgard@nwnet.net] Sent: Wednesday, December 10, 1997 2:12 PM To: Randy Bush Cc: Steve Martin; north american nominated origin of gossip Subject: RE: Cisco Vip Cards. what they won't support, I believe, is anything in the second adapter slot after the first PA-(the 0c-3 atm thingie) is installed....limitation has to do with the PA- (oc-3 atm), only. TE On Wed, 10 Dec 1997, Randy Bush wrote: > > Cisco will not support multiple PA's within a VIP2-40. > > I do not believe this statement to be true. > > randy >

1 0

Re: PPP Question
by Chris MacFarlane 10 Dec '97

10 Dec '97

More to the point is that the frame size is negotiated while handshaking. Once two devices agree on the size it is fixed for the duration of the session. Chris MacFarlane Manager, Data Operations ACC Telenterprises Canada 4162132023 cmacfarlane(a)acc.ca cjm(a)ican.net -----Original Message----- From: George Janosik <gjanosik(a)torrentnet.com> To: nanog(a)merit.edu <nanog(a)merit.edu> Date: Wednesday, December 10, 1997 4:09 PM Subject: PPP Question >A quick protocol question. > >Is a PPP frame size static or elastic? RFC 1661 states: > > Information Field > > The Information field is zero or more octets. The Information > field contains the datagram for the protocol specified in the > Protocol field. > > The maximum length for the Information field, including Padding, > but not including the Protocol field, is termed the Maximum > Receive Unit (MRU), which defaults to 1500 octets. By > negotiation, consenting PPP implementations may use other values > for the MRU. > > Padding > > On transmission, the Information field MAY be padded with an > arbitrary number of octets up to the MRU. It is the > responsibility of each protocol to distinguish padding octets from > real information. > >This is somewhat ambiguous. If I read it straight it sounds like PPP frames >are >static and that padding is used to maintain the MRU between endpoints. > >But, it also says the the Information field "MAY" be padded up to the MRU. > >I have hard time believing that a PPP frame would be static. >Very truly yours, > >George J Janosik >Senior Systems Engineer >Torrent Networking Technologies >gjanosik(a)torrentnet.com > >412-851-1103 > >

1 0

RE: Cisco Vip Cards.
by Steve Martin 10 Dec '97

10 Dec '97

Hi Randy, We have seen situations under heavy load where a VIP2-40 with an OC3 running at 80% and a Fast Ethernet PA running at 50% have resulted in severe packet loss and latency. After addressing this with our Cisco reps (including the Product Manager and Product Engineer for the 7513/VIP family) they acknowledged the problem and stated: "We do not recommend putting any other PA into a VIP2-40 if an OC3-PA is already installed. We have seen adverse performance, in the way of latency and packet loss, resulting from such a configuration. Customer's choosing to do so, do so at their own risk". Whether this is a company wide, advertised position at Cisco is unknown. During a recent, major launch, our Fast-E PA providing uplink connectivity to our download servers was located on the same VIP as one of our major NAP OC3 PA's. At 80% utilization on the Fast E and 50% on the OC3 we started seeing packet loss and increased latency. When the OC3 gets real busy, the SARing work consumes most of the VIP's resources. Additional taxation from other media interfaces pushes the VIP over the edge and it starts dropping packets and delaying throughput. It was such a problem for us that Cisco offered special pricing bundle's for us until the VIP2-50 could come out and handle the aggreagate loads better. Regards, Steve -----Original Message----- From: Randy Bush [SMTP:randy@psg.com] Sent: Wednesday, December 10, 1997 12:37 PM To: Steve Martin Cc: north american nominated origin of gossip Subject: RE: Cisco Vip Cards. > Cisco will not support multiple PA's within a VIP2-40. I do not believe this statement to be true. randy

2 1