nanog mailing list archives
Re: [ppml] too many variables
From: "vijay gill" <vgill () vijaygill com>
Date: Fri, 10 Aug 2007 11:41:46 -0700
On 8/10/07, Ted Mittelstaedt <tedm () ipinc net> wrote:
-----Original Message----- From: ppml-bounces () arin net [mailto:ppml-bounces () arin net]On Behalf Ofvijay gillSent: Friday, August 10, 2007 11:08 AM To: John Paul Morrison Cc: ppml () arin net; nanog () nanog org Subject: Re: [ppml] too many variablesI guess people are still spectacularly missing the real point. The pointisn't that the latest generationhardware cpu du jour you can pick up from the local hardware store isdoubling processing power every n months.The point is that getting them qualified, tested, verified, and thendeployed is a non trivial task. This is nonsense. The hardware cpu de jour that you pick up from the local chop shop is 1-2 years BEHIND what the high end fileserver vendors are using. Companies like HP go through a qualification, testing and verification process for their high end gear that is no less rigorous than what Cisco uses.
I knew reading nanog was a bad idea. However, now that I am well and truly in the weeds, might as well go forth. The phrase that I think I am looking for is.... it's coming to me...almost there.... ah yes amortization specifically amortization...units sold... cost basis... something something. /vijay The big difference is that the PC vendors get the processors from Intel and
AMD when they are in beta, and do their design and development while Intel and AMD are doing their own CPU design and development. So when Intel is done and ready to release, there is little work for the PC vendors left to do to ship complete product. The router vendors are approaching this like Ford and Chevy build car computers. They can get old Pentium 3 700Mhz chips for a few bucks a processor so that is what they are using. They can make an extra $90 in profit selling a $5000 router CPU card that has a $10 processor in it than a $5000 router CPU card that has a $100 processor in it. And from a marketing perspective if the router uses some exotic RISC chip that nobody has ever heard of, (because it's 15 year old obsolete technology) that somewhat insulates them from unflattering comparisons like what people are making here. This kind of attitude is symptomatic of the embedded systems industry. Price the stuff out first THEN develop for it. This is why for example you don't have an Ethernet jack in your automobile that you can plug a laptop in and get a complete fault code analysis for a vehicle failure from an embedded webserver in the engine computer. The embedded systems people insist on reinventing the wheel every time they design something and do their best to ignore what goes on in the PC world. Go ahead and make your arguments about deployment, but it is the router vendors who are foot dragging here. Ted
Current thread:
- Re: [ppml] too many variables, (continued)
- Re: [ppml] too many variables Scott Whyte (Aug 13)
- Re: [ppml] too many variables Leo Bicknell (Aug 14)
- Re: [ppml] too many variables Adrian Chadd (Aug 14)
- Re: [ppml] too many variables sthaug (Aug 14)
- Re: [ppml] too many variables Bruce M Simpson (Aug 24)
- Re: [ppml] too many variables Paul Vixie (Aug 10)
- Re: [ppml] too many variables vijay gill (Aug 10)
- Re: [ppml] too many variables Paul Vixie (Aug 10)
- Re: [ppml] too many variables Steven M. Bellovin (Aug 10)
- Message not available
- Re: [ppml] too many variables Joel Jaeggli (Aug 13)
- Message not available
- Re: [ppml] too many variables vijay gill (Aug 10)
- Message not available
- RE: [ppml] too many variables michael.dillon (Aug 12)