Date: Tue, 6 Jan 1998 20:24:14 -0600 From: "Tyson Boellstorff" <tboellst@willinet.net> To: "MegaFred" <mfred@zen.triax.com>, "dennis" <dennis@etinc.com> Cc: "Luis E. Muņoz" <lem@cantv.net>, "freebsd mailing list" <freebsd-isp@freebsd.org> Subject: Re: ISP Conversion Message-ID: <00b401bd1b13$59fb1e40$02010c0a@moe.splarg.org>
next in thread | raw e-mail | index | archive | help
>A 25XX series with really heavy loads (particularly with 2 ethernets) is a >much >less capable box than a unix router, so what fuels that choice? Is it the >company >you are going with?...they have lots of different products. The company you're working with fuels that choice. If I walk into an outfit and see 25 Cisco routers in operation, I am not likely to start a spiel about how much better PC's are (unless asked). Training & support costs at that point dictate that I just add another Cisco, and go away, mumbling. If they tell me that they want to have me design their net, now that's a different matter entirely. I get very frustrated when I talk to people who throw everything out the window just because the current consultant hated his predecessor's products, and manipulated an easily swayed customer into wasting cash by the barrel. I certainly do say that you should use what a significant percentage of other users are using. That's the dog tail theory: If you're a member of a tiny tail, you have no chance to wag the support dog. If you manipulate your hardware into what the support people are familiar with, odds are that you will not experience some of the weird problems I have inherited at my current job. Just like with Unix, if you use standard building blocks, you find you have no limits (except those imposed by poor network planning). The question then is how standard can you be and stilll get performance. Also, bear in mind how large the hardware vendor is in relation to their chosen market. The FX chipset is not better than the HX even though the FX sells better. See what percentage of servers are HX machines, and the proportions change dramatically. There is a human factor we all deal with. If you acknowledge it, you will find that some choices are made for you. If you don't, your users will be frustrated. I have never had a problem telling my customers my product's limits. They seem to find it refreshing. In addition, they are much more likely to buy from me, because they get candid information from me. If I were to swear that every product I sold was the best in the land, I would not be believed, and I would create an adversarial relationship. > >We're arguing about what's appropriate. PC routers are just are reliable as >standalones when used only as a router, the question with a unix box is >how much other server stuff (if any) do you put on it. > Well... The problem is that you have this attractive box just sitting there with 3% utilization. Something gets "temporarily" added to it, and it stays there forever. The best way around that is having a plan and sticking with it. I have two RS6000's at work, but I will not load perl on either, for that reason alone. I will get another box, because the beauty of PC's is that they are cheap and extremely reliable if you configure them correctly. My boss is cool on this, because we have already inherited many problems that "happened" when we grew too large to conveniently extricate ourselves from poorly placed applications. I say to advise them that while they _can_ add apps to their PC router, they should not if they ever plan on growing for the above reasons.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?00b401bd1b13$59fb1e40$02010c0a>