Brian Stempin on 22 May 2008 10:46:39 -0700


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: [PLUG] NIC bonding


One note about 802.3ad:

from http://en.wikipedia.org/wiki/802.3ad
A limitation on link aggregation is that it would like to avoid reordering Ethernet frames. That goal is approximated by sending all frames associated with a particular session across the same link[2]. Depending on the traffic, this may not provide even distribution across the links in the trunk.

I've bumped into this in past implementations before.  The above statement means that an NSF transfer may not (read: probably will not) use all of the available links.  My experience is that the traffic is very rarely distributed evenly across a server's nics because of this.  In order to fully utilize your link, you'd have to use multiple NSF (or whatever protocol you decide on) connections.

On Thu, May 22, 2008 at 11:40 AM, Mag Gam <magawake@gmail.com> wrote:
Hey Eric:

Let me guess you are a centos user too  ;-)

Acutally, I have considered using GFS (the redhat product). What do you think of it? Is it easy to manage? I acutally bought a HP product which is pretty fast. For the RAID, I am a little concerned. I am planning to implment RAID5 for 12 drives of 750GB. The problem is, if there is a bad drive it would take a very long time to place the RAID group as "stable". In other words, it would be in degraged mode eventhough we are in the process of replacing the drive and the array is rebuilding it self. Any thoughts on an optimal solution? I was thinking create 3 seperate RAID 5 volumes on the HW level (if thats possible with my HP product).

Anyway, thanks for your feedback. I was looking for a discussion like this.

TIA


On Thu, May 22, 2008 at 9:18 AM, Erek Dyskant <erek@blumenthals.com> wrote:



On Thu, 2008-05-22 at 09:05 -0400, Mag Gam wrote:
> Thanks for the responses. I am planning to setup a high performance
> network for my school as a summer project. High availability isn't my
> requirement but bandwidth is my primary concern. We have professors
> running peptide synthesis. Each model can easily output files large
> as .5-1TB and they move the files around on the LAN. Seems I will go
> with mode4.
>
> Now I have to decide which filesystem I must use for sharing. I
> suppose NFS.
>
>

Several things to consider:

1. Instead of NFS, you may consider iscsi with GFS.
2. Buy the most expensive drives you can afford.  You might hit the
drive's limitations before you hit the network's, especially if you're
using gigabit.  Also, put some thought into your RAID array or RAID
controller.


___________________________________________________________________________
Philadelphia Linux Users Group         --        http://www.phillylinux.org
Announcements - http://lists.phillylinux.org/mailman/listinfo/plug-announce
General Discussion  --   http://lists.phillylinux.org/mailman/listinfo/plug


___________________________________________________________________________
Philadelphia Linux Users Group         --        http://www.phillylinux.org
Announcements - http://lists.phillylinux.org/mailman/listinfo/plug-announce
General Discussion  --   http://lists.phillylinux.org/mailman/listinfo/plug


___________________________________________________________________________
Philadelphia Linux Users Group         --        http://www.phillylinux.org
Announcements - http://lists.phillylinux.org/mailman/listinfo/plug-announce
General Discussion  --   http://lists.phillylinux.org/mailman/listinfo/plug