We were using iSCSI to host Xen VMs. I ran some disk access tests from
within the VMs.
On 9/01/2013, at 9:35 AM, Carlos G Mendioroz <tron_at_huapi.ba.ar> wrote:
> Where was it measured ?
> I would suspect the biggest winners are file servers, followed by large (diskless?) clients, when using not so intelligent NICs, with NFS or iSCSI.
>
> I would not try to look for the advantage at the network side.
> 1500 -> 9000 is a 6x ratio (6 times less interrupts for transfers)
>
> There is a thread touching some of the aspects here:
> http://virtualstorageguy.com/2010/01/06/new-vmware-and-netapp-protocol-performance-report/
>
> -Carlos
>
> Lindsay Hill @ 08/01/2013 15:35 -0300 dixit:
>> I've done it across dedicated iSCSI segments of the network. Fully
>> isolated from the rest of the network, physically and logically.
>> Didn't seem to make much difference.
>>
>> I've seen a few posts over the years that try and measure the
>> performance difference. It seems to be fairly small, but non-zero. It
>> was a hassle trying to implement it though.
>>
>> On 9/01/2013, at 6:22 AM, Brian McGahan <bmcgahan_at_ine.com> wrote:
>>
>>> You normally want to keep them just on the LAN, or at least on a segment of the network where the packets won't get fragmented. If your MTU is 9000 and you hit a layer 3 segment that is 1500 and you need to fragment then you'll lose all your benefit of the jumbo frames in the first place. One of the ways to help fix this is to have the router say "ip tcp adjust-mss" to force the end host to fragment, but for UDP traffic this won't work. So really it depends on your traffic patterns. If a lot of your traffic is east-west inside your LAN then yes jumbo frames will help. If most of your traffic is north-south to the Internet and back then it probably won't help. In general the Internet will not support jumbo frames end-to-end.
>>>
>>> HTH,
>>>
>>> Brian McGahan, CCIE #8593 (R&S/SP/Security)
>>> bmcgahan_at_INE.com
>>>
>>> Internetwork Expert, Inc.
>>> http://www.INE.com
>>>
>>>
>>>
>>> -----Original Message-----
>>> From: nobody_at_groupstudy.com [mailto:nobody_at_groupstudy.com] On Behalf Of marc abel
>>> Sent: Tuesday, January 08, 2013 9:35 AM
>>> To: Cisco certification
>>> Subject: Jumbo Frames
>>>
>>> Is anyone currently using Jumbo frames in their environments? If so what kinds of performance increases (if any) have you seen and what sort of pain points have you experienced.
>>>
>>> Are you running it in a small segment of your network or across the entire environment? Are you running it in your data centers, your campuses, or both?
>>>
>>> I've been doing some reading on the topic and opinions seem all over the place on this. Personally I have set it up in the lab and it did seem to be somewhat beneficial, but then I think about taking it to my environment and i wonder what happens when that flood of 9K packets hits the edge ASA and has to be fragmented and shipped over a VPN. Does the ASA spike it's CPU and roll over? What happens if you are carrying VOIP traffic on the same links?
>>>
>>> All thoughts and opinions appreciated.
>>> --
>>> Marc Abel
>>> CCIE #35470
>>> (Routing and Switching)
>>>
>>>
>>> Blogs and organic groups at http://www.ccie.net
>>>
>>> _______________________________________________________________________
>>> Subscription information may be found at:
>>> http://www.groupstudy.com/list/CCIELab.html
>>>
>>>
>>> Blogs and organic groups at http://www.ccie.net
>>>
>>> _______________________________________________________________________
>>> Subscription information may be found at:
>>> http://www.groupstudy.com/list/CCIELab.html
>>
>>
>> Blogs and organic groups at http://www.ccie.net
>>
>> _______________________________________________________________________
>> Subscription information may be found at:
>> http://www.groupstudy.com/list/CCIELab.html
>
> --
> Carlos G Mendioroz <tron_at_huapi.ba.ar> LW7 EQI Argentina
Blogs and organic groups at http://www.ccie.net
Received on Wed Jan 09 2013 - 10:13:33 ART
This archive was generated by hypermail 2.2.0 : Sun Feb 03 2013 - 16:27:17 ART