CERFnet P. O. Box 85608 San Diego, CA 92186-9784 Email: help@cerf.net Phone: (619) 455-3900, (800) 876-2373 Fax (619) 455-3990 Sales Manager: NO LONGER THERE: John Lancia lanciaj@cerf.net (Carlos Robles) carlos@cerf.net POPs (T1) Oakland 510-987 Pacbell Palo Alto 415-852 Pacbell (56K) San Jose From tomj Thu Jul 8 14:15:38 1993 Received: by fido.wps.com (5.67/9999.1) id AA05545; Thu, 8 Jul 93 14:15:33 -0700 From: tomj (Tom Jennings) To: pushp@CERF.NET (Mohta Pushpendra) Date: Thu, 8 Jul 1993 14:15:13 -0700 (PDT) Subject: CERF connection for resellers Message-Id: <9307082115.AA05537@wps.com> X-L2L: wps.com Status: O I am the manager of The Little Garden, a San Francisco Bay Area Internetwork. Randy Bush (psg.com) suggested I contact you. We are looking for a new network connection. We have approximately 30 sites, ranging from 20 or so 14.4K's to 56K and T1. We want to grow our network, and have a number of 56K, 256K microwave, and T1 customers awaiting us finding a new supplier. We have a slightly less likely prospect of connecting a sister network, in Portland Oregon, with T1 lines we will supply, if the economics works out. We have spent the last four months researching our options on suppliers, obtained a very large amount of practical insight into regional telephone companies' long-lines, and the practical economics of reselling IP. Which brings me to my point in contacting you. I contacted CERF a few months ago, and again just last week. I was told about CERF's reseller agreement for a T1 connection, $2000/mo, plus an additional $1000/yr per 56K. This is fine on the face of it, but I had further reservations and questions, and I have not been able to get them answered. One, questions arise re: accounting for traffic/load. Counting 56K connections seems too simplisitic for our case; many examples follow: We have a POP in Palo Alto, fed by 56K, which in turn feeds 10 small nets fed at 14.4K. Is this one 56K, or 10 14.4K's? If we fed another site with 56K from that existing 56K, would that count as two 56Ks, or one? How do we count 256K microwave (or other) links? And most importantly, we have an immediate need to supply a T1 customer, who will in turn provide IP to ISDN customers. How would this be counted? Two, and I think more important. There is obviously a big difference between a "typical" customer (is there such a thing :-) with a T1 connection, and a reseller like us, who through aggregation will end up using a sizable proportion of the available bandwidth as we grow in size over time. We are quite willing to pay for bandwidth we use. One possible method of determining load is some averaged measure of traffic. An off-the-top-of-my-head method might be, 5-minute peaks measured once per hour, averaged over 48 hours. I mention this only to illuminate my idea, rather than as an actual proposal. GIVEN an ability to measure load in an agreeable manner, we propose a cost structure such as: <128K $2000/mo <512K $3000 <1024K $4000 voice +1-415-552-8156 PS: Not to rule out "head counting" entirely; we may be able to work out a method that will allow us to keep costs reasonable for sites we add, and allow CERF to allocate resources for us as we need. We could for example round things to the nearest 56K for < T1 sites, and in cases like our Palo Alto POP mentioned above, decide to count that as one 56K "head", or 10 14.4K "heads" at a lower annual rate each. The T1 connection to the site desiring to resell to ISDN customers remains problematic but probably resolvable. I look forward to working with you on this. -- Tom Jennings -- tomj@wps.com -- World Power Systems -- San Francisco, Calif. From tomj Thu Jul 8 15:35:19 1993 Received: by fido.wps.com (5.67/9999.1) id AA05815; Thu, 8 Jul 93 15:35:16 -0700 From: tomj (Tom Jennings) To: rgnet Date: Thu, 8 Jul 1993 15:35:10 -0700 (PDT) Subject: CERF -- finally Message-Id: <9307082235.AA05810@wps.com> X-L2L: wps.com Status: O Well I'm glad I decided to blow off Carlos. He had it all wrong. They charge $1000/yr per added NETWORK. Mohta Pushpendra called me back about 10 minutes after I s)ent the message. I went over all the basics, etc. I told him how we'd seen the linght re: economics. They still want to charge per network number, but I mentioned we might be interested in the following compromises: -- We put our 14.4K sites as we do now into one subnet. Mostly, they don't want *any* support calls from these. I assured him we handle them completely. Alternet has never received a single call from us on these. He mentioned if CERF had a lot of little networks that remained unreachable, the'd get annoying calls. >From our POV, putting the 14.4K sites under Cygnus' subnet solves that. -- We might want to organize or just declare that certain POPs of ours (Palo Alto-like, say Santa Cruz) are a "network number" and as long as they don't have to support the small fry, they might be happy with this. -- They would still have a problem with a reseller like us wanting to pass on infinite ability to resell. I told him, well, we could probably compromise on this -- while we do have explicit reseller resellers (the T1/ISDN guys are an excellent example), in fact it's a moot point on the 14.4K members, to which he agreed (because it's so bandwidth limited in the first place). For others, we could make explicit agreements. In microcosm, this is *exactly* what we worked out as unlivable for Alternet, and in turn, RGnet. So I don't don't think it's such an awful thing. -- Mohta said there is a small possibility they may be interested in the PDX/SF long line. I reiterated it is merely an option to keep mind, rather than any necessity; that the very existence of it depends on the economics. Other than that, CERF does not care about it. -- CERF is quite literally in the middle of completely reworking their "associate member" pricing. They all it that rather than "resellers" pricing, because of a side effect of CIX policy which states that only peers can talk to CIX, or something. Mohta said it's in the agreement; I said I hadn't read it in detail. I guess I have to. Yuck. They want to be competitive in exactly this area we're talking about. We spent a long time (20 minutes) simply ggoing over all those fuzzy areas of intent, how we view things, what we want, etc. I told him how we had "seen the light" of the economies of scale from the generic "their" side. I think he appreciated this somewhat, thoguh I detected a note of scepticism. He asked about NOC stuff, I told him that our internal backbone is UPS'ed (OK, so it will be soon) and that the low-speed people are frequently not, but they are isolated from the backbone and their agreements and expectations are quite different. This sufficed. Mohta is leaving for A'dam Saturday, and in two weeks they will be able to talk to us in detail about prices and arrangements. -- Last and hardly least -- come april, if it costs $2000/yr to publish a network number to a NAP, then we'll have to pay more too. This was a specific example he gave me. It may be completely fabricated, and then it may not be. You tell me. But it reiterates to us that the playing field will be profoundly different come April. If we can juggle our initial network as presented to CERF -- Cygnus with all the 14.4's, and an additional 3 - 4 56K pops -- come Apr94 of the $1K becomes $3k, our total monthly might remain reasonable. * * ** * I took a different tack, somewhat. I told him we are possibly willing to compromise on a number of fronts. It is obvious to me that if we deal with CERF, we will have a more intimate relationship with them than we set out to. I don't think we're gonna get a "here's a pipe, send us $X/mo and never talk". I don't see networking like we're talking about at that stage yet. Come April, the world will change. Better to be in tight with a supplier that we can work with. CERF will have to compromise here too, and trust us on some of this POP vs. customer stuff. No one has any hard solutions far as I can see. ANothre way to look at this: we'll be helping CERF define how they handle others like us. A smaller step than we wanted, but we did decide to back down in scale a bit... -- Tom Jennings -- tomj@wps.com -- World Power Systems -- San Francisco, Calif.