Hi Matthew,
Currently I am testing 82574 and 82571/2 chips with same traffic in
a forwarding path using,
netperf client ---> machine with 82571/2/4 chips --> netperf server
with below commands,
./netperf -t UDP_STREAM -l 3600 -H <target_ip> -- -m 64
./netperf -T 2 -t TCP_STREAM -l 3600 -H <target_ip>
In both the cases I see major drop in CPU utilization in top for
82571/2 chips but not for 82574 chips. As mentioned I do see drop in
number of interrupts in vmstat.
Rgds,
Nishit Shah.
On 6/14/2012 6:04 AM, Vick, Matthew wrote:
>> -----Original Message-----
>> From: Nishit Shah [mailto:[email protected]]
>> Sent: Wednesday, June 13, 2012 5:42 AM
>> To: Vick, Matthew
>> Cc: [email protected]
>> Subject: Re: [E1000-devel] problem with simplified balancing on 82574
>> chips
>>
>>
>> Hi Matthew,
>>
>> Bingo !!!!. It has solved first issue.
>> Thanks for doing it.
>>
>> I will open an issue on SourceForge page for the same with all the
>> details.
>>
>> One thing I want to put here is with this change I don't see a
>> major drop in CPU utilization in case of 82574 chip. (It is same even
>> with setting constant values like 3000 or 8000) Whereas I can see good
>> amount of drop in 82571/572/573 chips.
>> I do see less number of interrupts in vmstat but CPU utilization
>> is not dropping the way it is dropping for other chips.
>>
>> Rgds,
>> Nishit Shah.
> Great--glad to hear this worked for you! There are a few explanations for the
> behavior you're seeing: for one, 82574 behaves differently than 82571, 82572,
> and 82573, as it by default will use a different interrupt scheme. Also,
> depending on your traffic, that algorithm may not be the best choice. What
> are you comparing your CPU utilization to?
>
> Cheers,
> Matthew
------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
E1000-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/e1000-devel
To learn more about Intel® Ethernet, visit
http://communities.intel.com/community/wired
