2

I have grow to prefer x86 based router OS's like Vyatta and pfSense over their competitors Cisco and Juniper (I never really used Juniper, but still.). However, they feel "fake" to me, like "Frankenstein" routers. I think my greatest worry is that I am missing out on something by not using the main contenders.

Are there any benchmarks out there that compare the main metrics (throuput, etc.) of x86 router operating systems to their proprietary counterparts?

4 Answers 4

3

Packets-per-second is the metric that you want to look at, but be very careful to make sure that you're comparing apples to apples.

A router will have a higher PPS number when dealing with 64 byte packets than when it's moving 1500 byte frames, and the size of the routing table can be a big issue if the routing isn't being done in hardware.

That's one of the main reasons to go with a hardware router, by the way - they use ASICs that are specifically built for routing, which generally lets them achieve much higher throughput than x86 hardware doing routing in software.

For a real-world anecdote, a Vyatta 3520 (a rebranded Dell) is specced at 3 million packets per second of routing with 64-byte frames; I've seen one top out and start losing frames with the interfaces nowhere near saturated at just over 300,000 PPS when it was dealing with real-world frame sizes (averaging 1400 bytes) and lookups against a full global BGP routing table.

2
  • Unless my math skills are wholly inadequate isn't 300,000PPS * 1400 bytes per packet equal to a 3Gbps network connection at full saturation? I don't where any of my clients (or myself) would ever run into that kind of bandwidth... Even my internal networks are limited to 1Gbps NICs/switches...
    – Soviero
    Nov 22, 2011 at 23:22
  • @Kevin Your math is right. Just giving you an example of why you need to take a close look at the fine print of the specs to get an accurate comparison between devices, since the real-world performance in that case was only about 10% of the performance in ideal conditions. Nov 22, 2011 at 23:34
1

Cisco has a router performance sheet in pdf that uses 64 byte packets. You could probably use this to compare your model/brand versus a Cisco.

http://www.cisco.com/web/partners/downloads/765/tools/quickreference/routerperformance.pdf

0

there isnt any type of ROUTER benchmark utility available. there are many opensource routers/firewalls available, but you can benchmark one of their part, not WHOLE of them with any utility, as it doesnt makes sense.

e.g, By setting up PfSense firewall/router, you can add big RAM, good I/O controllers, high bandwidth/high throughput NIC and high speed processors and DISKS and then benchmark certain services like TCP throughput, DNS response time, squid proxy server benchmark, etc etc. the biggest weakness of propriety routers is that they cannot be customized. but in case of opensource routers, you can add Tens of GBs of RAMs, Terabytes of Disks space for logs, and you can scale it to any possible way. another issue with propriety routers is that the number of NIC available to them is limited, which is not an issue with opensource. as you are not bound to fixed hardware in them.

In my opinion, i would prefer PfSense, as i am using it on my network and using it to serve more than 10 services and about 8 different sites are connected through it. Everything is working perfectly fine without any type of delays or issues.

0

Popular saying:

nobody ever got fired for buying $LARGE_VENDOR.

The issue about x86 routers is not performance but support. A Cisco router may cost an arm, but it comes with the assistance of TAC in case of something goes wrong.

If you want to compare Cisco routers with others, here are the official specs.

1
  • except both pfSense and Vyatta have commercial support. I'm not familiar with Vyatta's offering but I presume they have good people. Cisco TAC is hit and miss, some great people, some so so at best. With pfSense, you're getting one of the most knowledgeable people in the world with the product with extensive support experience. Nov 23, 2011 at 1:02

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .