Community discussions

MUM Europe 2020
 
changeip
Forum Guru
Forum Guru
Topic Author
Posts: 3804
Joined: Fri May 28, 2004 5:22 pm

performance - benchmarking

Mon May 29, 2006 2:04 am

I am starting this thread to share some experiences with bandwidth testing on 2.9.24. It will become more complete the more testing I document.

Using bandwidth test between 2 mikrotik routers, both running on 3.0ghz Pentium D (631). These are dual core chips but is only using single core with Mikrotik. Network cards are Intel PCIe 82573 onboard. Intel MT quad port tests will be next. Cross over cable between the two routers.

NOTE: This first testing is between 2 routers directly, not in the recommended bandwidth testing configuration. Two end bandwidth testing devices with a middle router will be used going forward. This means 2 machines generating traffic 'thru' the router being tested. Again, the results below are end to end bandwidth tests with only 2 routers, 1 generating packets, the other receiving them. Still pretty amazing.

128 byte packet size:
610000 p/s, 712mbps

1500 byte packet size:
80,000 p/s, 975mbps

This configuration ran solid overnight 2 days in a row. Router generating traffic was at 100% cpu, the receiving router was at 22% cpu average.

It was actually possible to use 2gbps when sending both ways and larger packet sizes. (http://h1x.com/mt/gigabit/gigabit.html)

http://h1x.com/mt/gigabit/gigabit2.html

More to come - I'm just posting to brag about 610,000p/s in the first test : ) I assume if the proper bandwidth testing was setup (3 devices) I will see even better. I will test a routerboard 532 as well for comparison.

Sam
 
jarosoup
Long time Member
Long time Member
Posts: 600
Joined: Sun Aug 22, 2004 9:02 am

Mon May 29, 2006 5:06 am

Good god that's fast! 8)

Are you noticing any variation in what the bandwidth tester is giving you for speed reports? I'm having a hard time trusting the external bandwidth tester as if I use it on 2 PCs across a MT router, it's not always the same. For example, I have been testing queues on a bridged router, and sometimes the bandwidth is only 13Mb, sometimes it's closer to 30Mb, without changing out anything or touching the config on the tester or the router. Running the "onboard" testers seem to be more accurate.

Anyway, just curious if you've noticed this at such high speeds...very nice starting point though.
 
changeip
Forum Guru
Forum Guru
Topic Author
Posts: 3804
Joined: Fri May 28, 2004 5:22 pm

Mon May 29, 2006 7:32 pm

I'm having a hard time trusting the external bandwidth tester as if I use it on 2 PCs across a MT router, it's not always the same.
Yes, the Windows version seems to stink at generating packets.

I'm going to be using the F.I.R.E. cd to generate ddos type traffic, spoofed IPs, etc to perform some of this testing as well... I'll post the results. I need to load up the router with 100,000+ routes to see how it handles routing with a large table.

If only freebsd liked flash disks... sheesh!

Sam
 
User avatar
jager
Trainer
Trainer
Posts: 296
Joined: Mon Oct 31, 2005 2:44 am
Location: Sierra Leone
Contact:

Tue May 30, 2006 2:48 am

Great test, and great results!
 
Aldo
just joined
Posts: 1
Joined: Thu Jun 01, 2006 12:39 pm

Performance

Thu Jun 01, 2006 1:46 pm

Hi All,

I read all these high performace benchamrks but I am concerned that with as little as 100 Mbps passing through a mikrotik the CPU is already at 100%.

I am using an Intel Xeon 3.2 Ghz, 1 gig Ram on an Intel motherboard with Intel GigEthernet cards.

I am also using 2.9.24 version of mikrotik.

Should this be the case ?

I am only using Mikrotik as a VLAN switch and intend to do some bandwidth limiting on the VLAN.

Thanks
 
dwa
just joined
Posts: 18
Joined: Fri May 28, 2004 12:18 pm

Re: performance - benchmarking

Mon Jun 05, 2006 8:48 pm

Using bandwidth test between 2 mikrotik routers, both running on 3.0ghz Pentium D (631). These are dual core chips but is only using single core with Mikrotik.
Intel 631 Pentium 4 3.0GHz is Single Core CPU with Hyper-Threading.
 
changeip
Forum Guru
Forum Guru
Topic Author
Posts: 3804
Joined: Fri May 28, 2004 5:22 pm

Mon Jun 05, 2006 10:11 pm

You are correct.. i originally ordered the 631 and then swapped it for the 930 chip... Figured it would come in handy once SMP was supported by MIkrotik (if ever).

http://www.newegg.com/product/product.a ... 6819116238

Sam
 
User avatar
normis
MikroTik Support
MikroTik Support
Posts: 24387
Joined: Fri May 28, 2004 11:04 am
Location: Riga, Latvia

Tue Jun 06, 2006 8:34 am

"if ever" = next majour release
 
dwa
just joined
Posts: 18
Joined: Fri May 28, 2004 12:18 pm

Tue Jun 06, 2006 3:55 pm

You are correct.. i originally ordered the 631 and then swapped it for the 930 chip... Figured it would come in handy once SMP was supported by MIkrotik (if ever).
To what then the CPU occured testing? With 631 or 930?

Who is online

Users browsing this forum: Bing [Bot], Cvan, lsettembri and 109 guests