Reality is a network should be designed for the worst case you will have to deal with not the average case. This is often a business decision versus an engineering decision. Ie. 95th percentile vs 99th percentile. Or you don't run a controversial web site so you don't need to worry as much about syn attacks. You can't design for average or your network will be overloaded 50% of the time.
Mack McBride Network Architect -----Original Message----- From: cisco-nsp-boun...@puck.nether.net [mailto:cisco-nsp-boun...@puck.nether.net] On Behalf Of bored to death Sent: Thursday, September 02, 2010 12:49 AM To: Seth Mattinen; cisco-nsp@puck.nether.net Subject: Re: [c-nsp] Cisco Routers: Performance benchmark hi, thank you all for your replies. i didn't say ~90Mbit/s for 64Byte frame size is a bad thing (i said the low value is of course because of very small frame-size). i didn't mean they shouldn't give the result of 64Byte frame-size throughput, i meant they should (or it's better to) publish full result based on one universal benchmarking standard (such as RFC 2544). it's true that the average frame size is not 1500Bytes and it's true that you should buy a router based on the topology and other specs of your network. but i think these numbers (specifically the result of RFC 2544 tests) shows the router performance power on different aspects of routing and gives you something to compare with other products (such as vyatta or else) or with other products of the same vendor (cisco 3800 or else). (the main reason of setting RFC 2544 is too give you some performance results to compare between different products on the same branch). for example, if you have a network and you know it may have up to 1 gigabit/s traffic or more or less passing on your router, you should get sure if the router you buy has the power to work under high loads of traffic (instead of just having hope or taking a guess). . i think these numbers (or result numbers from any other universally accepted standard about performance) should be available for network products of any vendors (or at least i think it's better to know these numbers before making decisions). for example, RFC 2544 says you should give benchmark results on traffic with frame-sizes of 64, 128, 256, 512, 1024, 1518 byte. and in theory if we have combination of packets with different frame-sizes, performance is almost equal to the average frame-size of all packets, so if we have test results with these frame-sizes, we can be sure if the router we wanna buy can work under the highest load of the passing traffic on our network or not. thank you. ________________________________ From: Seth Mattinen <se...@rollernet.us> To: cisco-nsp@puck.nether.net Sent: Wed, September 1, 2010 10:51:55 PM Subject: Re: [c-nsp] Cisco Routers: Performance benchmark On 9/1/2010 09:04, Christopher J. Wargaski wrote: > Thanks for posting the URL for the router performance matrix. Anyone > know of a similar matrix for switches (L2 & L3) and firewalls? > Google "cisco switch performance" ~Seth _______________________________________________ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/ _______________________________________________ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/ _______________________________________________ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
