Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 11 Feb 2006 17:48:04 +0000 (GMT)
From:      Robert Watson <rwatson@FreeBSD.org>
To:        dima <_pppp@mail.ru>
Cc:        Marcos Bedinelli <bedinelli@madhaus.cns.utoronto.ca>, freebsd-net@freebsd.org
Subject:   Re: Network performance in a dual CPU system
Message-ID:  <20060211174648.X90460@fledge.watson.org>
In-Reply-To: <E1F7fzS-000GUW-00._pppp-mail-ru@f16.mail.ru>
References:  <E1F7fzS-000GUW-00._pppp-mail-ru@f16.mail.ru>

next in thread | previous in thread | raw e-mail | index | archive | help

On Sat, 11 Feb 2006, dima wrote:

>> The system is mainly being used as a dedicated router. It runs OSPF, BGP 
>> and IPFW (around 150 rules). OSPF and BGP are managed by Quagga. The box 
>> has 2 gigabit interfaces that handle on average 200Mbp/s - 50K packets/s 
>> (inbound and outbound combined), each one of them.
>
> The second CPU wouldn't help you for sure. There's only one [swi1: net] kernel thread which deals with all the kernel traffic. The option of per-CPU [swi: net] threads was discussed on freebsd-arch@ several months ago, but it wouldn't be implemented soon. So, the only hardware option is installing the fastest CPU possible.
> There are several software (FreeBSD specific) options though:

If you set net.isr.direct=1, the netisr workload is moved from the netisr 
thread to the thread performing the dispatch -- typically, the ithread.  If 
you have multiple interfaces and they are assigned different ithreads, then 
the work can occur in parallel.  However, there are some other properties of 
this setting that are important, so it affects different workloads in 
different ways.

Robert N M Watson

> 1. You should surely try polling(4). 50kpps mean 50000 interrupts and the same amount of context switches, which are quite expensive.
> 2. FastForwarding. It's the most suitable for you. As I know, Quagga inserts its dynamic routes to the system routing table. And FastForwarding is aware of routing table and firewall rules. And the most exciting: you can switch it on/off without reboot:
> # sysctl net.inet.ip.fastforwarding=1
> The only limitation is it applies to IPv4 unicast traffic only. There's no documentation on this feature as i know (am I wrong, or should I report this as a documentation bug?) but you can look at the comments in the beginning of /sys/netinet/ip_fastfwd.c
> The authors reported up to 1Mpps (see page 10 at http://people.freebsd.org/~andre/FreeBSD-5.3-Networking.pdf)
>
>>
>>
>> Some of you have asked for the following information:
>>
>>
>> - As I indicated before, polling is currently disabled.
>>
>>
>> - Hyperthreading (HTT) is disabled.
>>
>>
>> mull [~]$vmstat -i
>> interrupt                          total       rate
>> irq1: atkbd0                        3466          0
>> irq6: fdc0                            10          0
>> irq13: npx0                            1          0
>> irq14: ata0                           47          0
>> irq21: fxp1                     20462527          8
>> irq28: bge0                   3511765157       1444
>> irq29: bge1                   3633124373       1494
>> irq30: aac0                      1842472          0
>> cpu0: timer                    566751007        233
>> Total                         7733949060       3181
>>
>>
>> mull [~]$netstat -m
>> 644/646/1290 mbufs in use (current/cache/total)
>> 643/407/1050/17088 mbuf clusters in use (current/cache/total/max)
>> 0/5/4528 sfbufs in use (current/peak/max)
>> 1447K/975K/2422K bytes allocated to network (current/cache/total)
>> 0 requests for sfbufs denied
>> 0 requests for sfbufs delayed
>> 0 requests for I/O initiated by sendfile
>> 0 calls to protocol drain routines
>>
>>
>>
>> Thank you,
>>
>> --
>> Marcos
>
> _______________________________________________
> freebsd-net@freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-net
> To unsubscribe, send any mail to "freebsd-net-unsubscribe@freebsd.org"
>



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060211174648.X90460>