October 1996 - Test - lists.nanog.org

Re: Cisco's AIP vs HSSI
by Kent W. England 15 Oct '96

15 Oct '96

At 09:46 AM 15-10-96 -0400, Padma Krishnaswamy wrote: >One other point about the AIPs relative to the HSSI/DSU ; you may not be >interested in this feature , but the AIPs implement traffic shaping- ie You can >set Peak and Average rates and burst size; you could use this to subrate your >traffic -should you be in a situation where this is desirable. > > While this is true, I would caution the novice that TCP sources don't really benefit from rate shaping and the sort of burst capability you get with VBR. TCP appears to work with the explicit rate control of ABR, but there are some concerns in the research community that it might not work because both TCP and ABR are working the congestion management problem simultaneously and there may be pathological or even common situations where these congestion management schemes conflict or degrade each other. Time will tell. To date, we know that TCP works well with UBR and large buffers. Large is defined by TCP and the path length and works out roughly to twice the bandwidth delay product on the longest path. --Kent

1 0

Re: Cisco's AIP vs HSSI
by salo＠msc.edu 15 Oct '96

15 Oct '96

> Date: Tue, 15 Oct 1996 09:46:23 -0400 > From: kri(a)bellcore.com (Padma Krishnaswamy) > To: nanog(a)merit.edu, salo(a)msc.edu (Tim Salo) > Subject: Re: Cisco's AIP vs HSSI > > One other point about the AIPs relative to the HSSI/DSU ; you may not be > interested in this feature , but the AIPs implement traffic shaping- ie You > can set Peak and Average rates and burst size; you could use this to > subrate your traffic -should you be in a situation where this is desirable. That is my understanding. I think that this is part of a some more general observations: o The AIP card is a more intelligent card and can perform more functions today and is much more likely to perform more functions in the future. o I suspect that few, if any, new ATM features will be added to the HSSI card. Again, your router vendor should be telling you all of this. > I havent looked around to check out any new ADSU type products since mid 95, > but it used to be that you could not do this with an ADSU. This is also my understanding. I thought I saw some work being done on an enhanced DXI specification, (the interface between the router and the ATM DSU), but I think that, for the most part, not much work is being done with ATM DSU solutions today. ATM DSUs were largely developed as a quick-to-market solution for adding ATM functionality to routers, without the need for new router boards. Their time has passed. -tjs

4 5

Re: cable modems [Re: Intra/Inter - was Inet-II]
by Milo Medin 15 Oct '96

15 Oct '96

Really Ken, just because you consult to PacBell, you shouldn't be so obvious in your posturing. :-) Cable modems by themselves don't don't the performance problem, in fact, that's what @Home has been talking about for sometime now, and why we believe you have to build an intelligent network architecture, not just dumb pipes. Some folks do believe that just fixing the loop will fix system performance, and they are clearly in the wrong. However, I will venture to say that unless you do something like what we're doing, it doesn't matter if the loop is ADSL or Cable or laser, it's not going to end up running very fast. This is not just a loop problem. As for the hype machine, KPCB has a pretty good track record (Sun Micro, Intuit, Netscape, etc...), and I am very pleased in my interactions with those guys. They have an excellent reputation in the VC community. As for @home and it's financials, I deal with lots of analysts, and they seem to really like us. Our service is up in Fremont in commercial service, and we're turning on a bunch more cities (we have 3 other regions in test as I type). (Not really hype-like to actually field real service ;-)) Just look at some of the industry analyst reports for cable folks, they have been very favorable... Of course, the RBOC's could have fielded drastically cheaper fast services ages ago, but that would have canabalized all those nice leased line products... :-) Thanks, Milo

1 0

Re: Digest Possible??
by William B. Norton 15 Oct '96

15 Oct '96

It is not currently in place, but it is on the queue. Bill At 04:58 PM 10/14/96 -0500, Andrew Smith wrote: > >This question has probably been asked many times over, but I haven't >been able to find it or an answer. > >Is there, and if there isn't, is getting it setup possible, >a nanog-digest mailing list. > >--------------------------------------------------------------------------- >Andrew Smith ** awsmith(a)neosoft.com ** Network Engineer ** 1-888-NEOSOFT > ** "Opportunities multiply as they are seized" - Sun Tzu ** > ** http://www.neosoft.com/neosoft/staff/andrew ** >--------------------------------------------------------------------------- > >

1 0

Re: Cisco's AIP vs HSSI
by kri＠bellcore.com 15 Oct '96

15 Oct '96

One other point about the AIPs relative to the HSSI/DSU ; you may not be interested in this feature , but the AIPs implement traffic shaping- ie You can set Peak and Average rates and burst size; you could use this to subrate your traffic -should you be in a situation where this is desirable. I havent looked around to check out any new ADSU type products since mid 95, but it used to be that you could not do this with an ADSU. Padma Krishnaswamy

1 0

Re: BBN outage
by Rob Liebschutz 15 Oct '96

15 Oct '96

> Most of the SF Bay Area portion of BBN Planet's network has been offline > since late last night. I've finally heard, via a friend who is a major > customer of theirs, that they are having power problems in Palo Alto. > > Now, I haven't heard a peep about this on any of the mailing lists I'm on... > and since knowing about BIG outages like this would make it easier for > me to answer my customer's questions... am I missing something? > > Is there a mailing list that I should be on? Or is this just a private > BBN issue that I'd need to call their ops center people about (people who > I'm sure are very very busy without small network operators calling them) > > -matthew kaufman > matthew(a)scruz.net > > I just spoke to their NOC and was told that a power switch that is supposed to be able to switch them between 3 different power utilities failed at 12:30 this morning (friday). They bought and installed 2 large diesel generators today and are hoping to be back up using the generators within 30 minutes. Rob

6 11

Re: BBN outage
by Dorian R. Kim 15 Oct '96

15 Oct '96

On Mon, 14 Oct 1996, Patrick J. Chicas wrote: > On Mon, 14 Oct 1996, Rob Gutierrez wrote: > > > Telco wise, BBN does have a nice dirverse system. Their MCI T-3 is on > > a microwave that goes to Palo Alto, where it drops off to MCI's bay area > > fiber ring between San Jose to San Francisco. > > Their HUB is connected by one microwaved DS-3? This is not good. Granted, This DS3 is one of 4 private interconnect BBN has to MCI. I don't think there's anything wrong with it's current provisioning given that their other backbone DS3 have diverse routing.. Please correct me if I've missed something. -dorian rk, or 2) Was not available ... See my previous comments about rusty generators. > Telco wise, BBN does have a nice dirverse system. Their MCI T-3 is on > a microwave that goes to Palo Alto, where it drops off to MCI's bay area > fiber ring between San Jose to San Francisco. Their HUB is connected by one microwaved DS-3? This is not good. Granted, most DS-3 microwave radios have hot standby radios, muldems ect. on each end but, they are always ripe for outage due to interference or path fade. I can see the microwave as a segment of a SONET OC-3 link, only if the other side is fiber. > Some of their backbone T-3's > come into a Pac Bell OC-48, which has an entirely seperate protection > path into Stanford itself (Stanford alltogether has two Pac Bell OC-48's, > with a third one planned!). Another fiber path belongs to MFS for more > dirversity. So on that level, they have their act together. This is good.. How many of BBN's peer connection run over the SONET ring? Regards Patrick J. Chicas Email: pjc(a)unix.off-road.com URL: http://www.Off-Road.com -------------------------------- The Off-Road Center of The 'Net!

2 1

Re: Cisco's AIP vs HSSI
by salo＠msc.edu 15 Oct '96

15 Oct '96

> Date: Fri, 11 Oct 1996 15:09:51 -0400 (EDT) > From: Pritish Shah <pritish(a)iocenter.net> > To: nanog(a)merit.edu > Subject: Cisco's AIP vs HSSI > > Can anyone tell me which is better and why? Is there any resources out > there that will help me in that decision. The short answer: You will undoubtedly be much happier if you use the Cisco AIP card rather than a HSSI card with an external ATM DSU. A longer answer: HSSI ---- The HSSI card is a fairly dumb interface. It is actually an interface between the Cisco router and an external DSU. As I recall, it has a V.35 interface. I think its top speed is a bit more than a DS-3. Its original function was to connect a DS-3 (45 Mbps) DSU to a Cisco router. ---------- | | ------------ | Router |======| DS-3 DSU |----------> DS-3 leased line | | ------------ ---------- The first available method of connecting a Cisco router to an ATM network was to use an ATM DSU connected to a HSSI interface. The ATM DSU appeared to the Cisco router to be pretty much the same as a traditional DSU. All of the ATM functions, largely segmentation and reassembly, were performed in the ATM DSU. ---------- | | ------------ | Router |======| ATM DSU |----------> DS-3 leased line to ATM Network | | ------------ ---------- The HSSI card generally uses "frame relay" or NLPID encapsulation within the AAL frame. The more standard encapsulation today is the LLC/SNAP encapsulation. I believe that Digital Link built the first ATM DSU. This was a fairly early device, so it implemented AAL3/4, (rather than AAL5 which is the predominant AAL today). The Digital Link cannot use AAL5. The Cisco AIP card generally wants to use AAL5. We have been unsuccessful in connecting a Cisco with a HSSI and Digital Link to a Cisco with an AIP card. The details are another long discussion which I won't bother to type at this time. ADC Kentrox also builds an ATM DSU. It supports AAL5. However, the experience at the ATM NAPs indicate that this device has some performance limitations. Among other things, the DXI specification, (which specifies the interface between the router and the ATM DSU), says that only 16 reassembly buffers are required in the ATM DSU, (if memory serves me...). This limited reassembly space can cause problems. On the other hand, the Kentrox ATM DSU supports AAL5. I think it also interoperates with the Cisco AIP card, although I haven't seen it first hand. Other issues with a HSSI/ATM DSU solution include: o The DXI interface supports only PVCs. In a world which is moving towards SVCs, this is a rather serious limitation. o There are a whole host of new ATM features which the HSSI card/ADSU solution won't be able to support. o Check with Cisco about performance limitations in the HSSI card. o I don't believe that the HSSI card will support anything near OC-3c speeds. OC-3c ATM products are readily available in local area networks. Also note that most people run their wide-area ATM connection through a local ATM switch before connecting it to a router. I would be inclined to make the connection between the local ATM switch and the router OC-3c rather than DS-3 (for a bunch of reasons). Cisco AIP --------- I don't believe that the Cisco AIP card is the ultimate solution, but it is far superior to the HSSI interface. Important advantages include: o The AIP card supports SVCs. o The AIP card supports standard encapsulations, AAL5 and LLC/SNAP. (It might also support NLPID and AAL3/4; check with Cisco.) o The AIP supports a variety of interfaces. I think these include multi-mode OC-3c, single-mode OC-3c and DS-3. Verify with Cisco. o I think the performance of the AIP card may be better than the HSSI card, but I don't have any test results in front of me. Summary ------- Read my short answer and skip the rest. Disclaimer: Verify the facts with your vendor. I typed this in on the fly late at night; the general ideas are most likely correct, but some of the details might be a bit off. -tjs

1 0

Re: Cisco's AIP vs HSSI
by salo＠msc.edu 15 Oct '96

15 Oct '96

> Date: Fri, 11 Oct 1996 15:09:51 -0400 (EDT) > From: Pritish Shah <pritish(a)iocenter.net> > To: nanog(a)merit.edu > Subject: Cisco's AIP vs HSSI > > Can anyone tell me which is better and why? Is there any resources out > there that will help me in that decision. The short answer: You will undoubtedly be much happier if you use the Cisco AIP card rather than a HSSI card with an external ATM DSU. A longer answer: HSSI ---- The HSSI card is a fairly dumb interface. It is actually an interface between the Cisco router and an external DSU. As I recall, it has a V.35 interface. I think its top speed is a bit more than a DS-3. Its original function was to connect a DS-3 (45 Mbps) DSU to a Cisco router. ---------- | | ------------ | Router |======| DS-3 DSU |----------> DS-3 leased line | | ------------ ---------- The first available method of connecting a Cisco router to an ATM network was to use an ATM DSU connected to a HSSI interface. The ATM DSU appeared to the Cisco router to be pretty much the same as a traditional DSU. All of the ATM functions, largely segmentation and reassembly, were performed in the ATM DSU. ---------- | | ------------ | Router |======| ATM DSU |----------> DS-3 leased line to ATM Network | | ------------ ---------- The HSSI card generally uses "frame relay" or NLPID encapsulation within the AAL frame. The more standard encapsulation today is the LLC/SNAP encapsulation. I believe that Digital Link built the first ATM DSU. This was a fairly early device, so it implemented AAL3/4, (rather than AAL5 which is the predominant AAL today). The Digital Link cannot use AAL5. The Cisco AIP card generally wants to use AAL5. We have been unsuccessful in connecting a Cisco with a HSSI and Digital Link to a Cisco with an AIP card. The details are another long discussion which I won't bother to type at this time. ADC Kentrox also builds an ATM DSU. It supports AAL5. However, the experience at the ATM NAPs indicate that this device has some performance limitations. Among other things, the DXI specification, (which specifies the interface between the router and the ATM DSU), says that only 16 reassembly buffers are required in the ATM DSU, (if memory serves me...). This limited reassembly space can cause problems. On the other hand, the Kentrox ATM DSU supports AAL5. I think it also interoperates with the Cisco AIP card, although I haven't seen it first hand. Other issues with a HSSI/ATM DSU solution include: o The DXI interface supports only PVCs. In a world which is moving towards SVCs, this is a rather serious limitation. o There are a whole host of new ATM features which the HSSI card/ADSU solution won't be able to support. o Check with Cisco about performance limitations in the HSSI card. o I don't believe that the HSSI card will support anything near OC-3c speeds. OC-3c ATM products are readily available in local area networks. Also note that most people run their wide-area ATM connection through a local ATM switch before connecting it to a router. I would be inclined to make the connection between the local ATM switch and the router OC-3c rather than DS-3 (for a bunch of reasons). Cisco AIP --------- I don't believe that the Cisco AIP card is the ultimate solution, but it is far superior to the HSSI interface. Important advantages include: o The AIP card supports SVCs. o The AIP card supports standard encapsulations, AAL5 and LLC/SNAP. (It might also support NLPID and AAL3/4; check with Cisco.) o The AIP supports a variety of interfaces. I think these include multi-mode OC-3c, single-mode OC-3c and DS-3. Verify with Cisco. o I think the performance of the AIP card may be better than the HSSI card, but I don't have any test results in front of me. Summary ------- Read my short answer and skip the rest. Disclaimer: Verify the facts with your vendor. I typed this in on the fly late at night; the general ideas are most likely correct, but some of the details might be a bit off. -tjs

1 0

Re: Building Evacuation [Re: BBN outage]
by Sean Donelan 14 Oct '96

14 Oct '96

>You have to remember that it's harder to keep people in a given >building than it is to safeguard the equipment there. You always >have to be ready to send your NOC operators packing with a cell >phone and a laptop for up to 48 hours, in case the building has >to be evacuated for one of a dozen reasons. You should have >a backup call center with an ACD set up in some other city or >a contract with someone for a backup service. You realize you have just cursed someone's NOC to this fate. If my next post comes from Monterey, California, you'll know we had to evacuate to our backup NOC. Once upon a time, when Merit ran the Internet, as their last best hope they had a plan for the NOC operators to grab a laptop as they ran from the NOC. One of the harder things to plan is how you are going to get back inside your possibly non-functioning, or damaged network from the outside without leaving too many backdoors for the evil ones. Its a embarassing when your SecureID servers crashes inside a building you can't physically enter, and none of your network equipment will let you log in. Its even more embarassing when a cracker uses your 'secret' backdoor and takes over your network. -- Sean Donelan, Data Research Associates, Inc, St. Louis, MO Affiliation given for identification not representation

1 0