[TriLUG] Tuning WAN links

Shawn Hood shawnlhood at gmail.com
Fri Nov 2 11:02:07 EDT 2007


My apologies for lack of clarity.

There are LANs on each end of this link.  15-30 servers in each LAN.  The
bulk of the traffic between these machines occurs within their respective
LANs.  Only a portion of the data travels over the WAN link.  The traffic
over the WAN link will be coming from multiple servers.  That said, tuning
each machine for optimal utilization of the WAN link isn't necessarily best
for traffic destined to the local LAN.  I don't want to adversely impact
performance of the local LAN just so I can have optimal WAN performance.
That said, I would prefer to tune the machines to perform well on the local
LAN, and have another machine or router on each end tuned for optimal
throughput for that link.  I just don't know if this is a feasible approach.

Will using machines that serve as gateways for the WAN link (and are also
tuned for this link) translate into better performance overall?

Shawn



On Nov 2, 2007 12:42 AM, Jeremy Portzer <jeremyp at pobox.com> wrote:

> Shawn Hood wrote:
> > Thanks for the response, guys.
> >
> > I'm mostly satisfied with the local network performance on both ends.
>  What
> > could be done to increase the performance without modifying settings on
> each
> > box involved.  Could a box at each end be used to shape the traffic in a
> way
> > that would optimize this link, or do we really need to try to tune each
> box?
> >
>
> Well your question was a bit academic without explaining the end goal...
> is transferring data between these two "boxes" the primary purpose of
> this WAN link?  Or are you just using those boxes in order to test the
> link for use by other purposes?  As Aaron alluded to (I'm sure you read
> his whole message, right?), typically links of this size would be used
> for lots of "smaller" connections (many computers talking to many other
> computers) rather than transferring of large amounts of data between
> only two hosts.
>
> Either way, I don't see what kind of "tuning" you're after if you don't
> want to modify the settings on the boxes involved.  (Isn't "tuning"
> primarily "modifying settings"?)   If there only two of them and this
> link is dedicated to them, why is this a problem?  And as Aaron pointed
> out, if the issue is network buffer sizes, you can't get around that by
> adding more hosts.
>
> If there are really more than two boxes, you haven't really accurately
> described the scenario - and there is no fundamental reason to assume
> that the default settings of the routers and switches aren't adequate to
> serve a "normal" load of many connections.
>
> I really like Aaron's suggestion of "lab testing" things with a third
> host to simulate the WAN link by adding latency artificially.  Though
> personally I would completely avoid any *BSD if at all possible.  *grin*
>
> --Jeremy
> --
> TriLUG mailing list        : http://www.trilug.org/mailman/listinfo/trilug
> TriLUG Organizational FAQ  : http://trilug.org/faq/
> TriLUG Member Services FAQ : http://members.trilug.org/services_faq/
>



-- 
Shawn Hood
(910) 670-1819 Mobile



More information about the TriLUG mailing list