Re: The Gorgon's Knot. Was: Re: Verio Peering Question
| When Sprint was filtering there was a demonstrable need based on the 64meg | limit that | mainstream routers had for memory at the time. I do not see that there is | any such physical | limitation today and I guarentee that the router vendors (all two of them) | have learned the lesson | of not including enough address lines on the equipment to allow for easy | memory upgrades. So we should throw away all the 7200s and similar routers today because they are in the way of growing numbers of long prefixes, replacing them with new routers manufactured since the time of the above-mentioned lesson? And when shall we throw away the 12000s and similar routers (or components thereof) because they are underpowered in the face of routing-table growth, compared to well-established alternatives? Incidentally, the lesson learned was that sheer storage AMOUNT is only a (perhaps small) part of the problem, compared to the processing necessary to use that storage in support of dynamic routing (in terms of CPU and in terms of accesses to that memory). Sean.
On Fri, 28 Sep 2001, Sean M. Doran wrote:
So we should throw away all the 7200s and similar routers today because they are in the way of growing numbers of long prefixes, replacing them with new routers manufactured since the time of the above-mentioned lesson? And when shall we throw away the 12000s and similar routers (or components thereof) because they are underpowered in the face of routing-table growth, compared to well-established alternatives?
Still using an 8086 for a desktop? Obsolences does happen....
At 14:12 -0700 28-09-2001, Patrick Greenwell wrote:
On Fri, 28 Sep 2001, Sean M. Doran wrote:
So we should throw away all the 7200s and similar routers today because they are in the way of growing numbers of long prefixes, replacing them with new routers manufactured since the time of the above-mentioned lesson? And when shall we throw away the 12000s and similar routers (or components thereof) because they are underpowered in the face of routing-table growth, compared to well-established alternatives?
Still using an 8086 for a desktop?
Obsolences does happen....
If it happens faster then you amortize the cost of capital then your investors will go elsewhere. This is not 1999. Money is no longer cheap. Business plans need to be profitable. Spending millions of dollars on a new set of core routers on less than a 3 year cycle can lead to low stock prices, de-listing and possible bankruptcy.. Economic reality does happen .... -- Joseph T. Klein +1 414 915 7489 Senior Network Engineer jtk@titania.net Adelphia Business Solutions joseph.klein@adelphiacom.net "... the true value of the Internet is its connectedness ..." -- John W. Stewart III
Sean, Not at all. Modern 7200-series routers, with newer NPE's and more memory can easily handle full tables today, and into the future. Therefore, we don't need to through away 7200s. However, should we all be held hostage to those unwilling to upgrade their existing routers, and perhaps eventually, upgrade to new routers? Routers are, basically, specialized computing devices, with fairly short lives, compared to things like household appliances, arc welders, or phone booths. This is reflected in their shorter depreciation schedules. As the upgrades that extend the life of the routers, in dealing with larger routing tables, are also the cheapest - i.e. RAM, controlling routing table size to prevent vast expenditures of money to replace existing routers simply doesn't hold water. The Internet did not collapse on the day that 2501s became incapable of handling a full view of the routing table. There was little gnashing of teeth or rending of garments when it happened. That is a lesson well remembered. - Daniel Golding
-----Original Message----- From: owner-nanog@merit.edu [mailto:owner-nanog@merit.edu]On Behalf Of Sean M. Doran Sent: Friday, September 28, 2001 5:09 PM To: bortzmeyer@gitoyen.net; jtk@titania.net; vbono@vinny.org Cc: nanog@merit.edu Subject: Re: The Gorgon's Knot. Was: Re: Verio Peering Question
| When Sprint was filtering there was a demonstrable need based on the 64meg | limit that | mainstream routers had for memory at the time. I do not see that there is | any such physical | limitation today and I guarentee that the router vendors (all two of them) | have learned the lesson | of not including enough address lines on the equipment to allow for easy | memory upgrades.
So we should throw away all the 7200s and similar routers today because they are in the way of growing numbers of long prefixes, replacing them with new routers manufactured since the time of the above-mentioned lesson? And when shall we throw away the 12000s and similar routers (or components thereof) because they are underpowered in the face of routing-table growth, compared to well-established alternatives?
Incidentally, the lesson learned was that sheer storage AMOUNT is only a (perhaps small) part of the problem, compared to the processing necessary to use that storage in support of dynamic routing (in terms of CPU and in terms of accesses to that memory).
Sean.
So we should throw away all the 7200s and similar routers today because they are in the way of growing numbers of long prefixes, replacing them with new routers manufactured since the time of the above-mentioned lesson?
No. Not at all. But nor should we cry wolf and defend our corporate policies with phoney hysteria. If you are filtering to keep your internal routing tables clean fine, to get the most out of your older or less than optimum equipment, fine too, but don't say the "the internet is in danger of imminent collapse if we don't do this" either.
And when shall we throw away the 12000s and similar routers (or components thereof) because they are underpowered in the face of routing-table growth, compared to well-established alternatives?
As soon as it becomes pratical. But again, don't defend the (legitimate) position of wanting to get the most mileage out of your installed equipment base or *not* wanting to spend millions of dollars on forklift expansion by saying that its for the good of the internet. Although this is certainly the first time I've seen ivory tower idealism (aka nice clean routing tables with short allocations) match up to the goals of a real world corporation (aka lets encourage customers to buy our service and at the same time prolong the life of our core routers by a year or so).
Incidentally, the lesson learned was that sheer storage AMOUNT is only a (perhaps small) part of the problem, compared to the processing necessary to use that storage in support of dynamic routing (in terms of CPU and in terms of accesses to that memory).
Historically router CPU technology has lagged behind the server industry, at least in raw processing capability. Well, there was the fine machine made by Bay, but that proved beyond a shadow of a doubt that faster with more memory does not a stable router make. But that's a different thread. -vb
participants (5)
-
Daniel Golding -
Joseph T. Klein -
Patrick Greenwell -
smd＠clock.org -
Vincent J. Bono