[<prev] [next>] [day] [month] [year] [list]
Message-ID: <xvuuf248uo940ii7w0kd1egd.1328717345088@email.android.com>
Date: Wed, 08 Feb 2012 10:10:36 -0600
From: "Gregory G. Carter" <gcarter@...gi.com>
To: "John A. Sullivan III" <jsullivan@...nsourcedevel.com>,
netdev@...r.kernel.org
Subject: Re: Proper value for bandwidth parameters
Set nagios up and do a transfer over the line without TC controls. Then use the figures from the line as a base for setting policy using TC.
This stuff is an art, not an idealized set of edges in a textbook on network theory. So always see what your real capacity is before you set policy.
Gc
"John A. Sullivan III" <jsullivan@...nsourcedevel.com> wrote:
>Hello, all. From a system administrator's perspective, do I need to
>reduce the values I feed to bandwidth defining parameters in tc such as
>the ul m2 values in HFSC?
>
>For example, if I have a 1.544 mbits circuit and I set HFSC ul m2 to
>1544kbits (or any other bandwidth parameter in different algorithms),
>when it calculates its timer for dequeuing the packet (let's assume a
>full sized 1500 byte packet) is it performing the calculation on 1500
>bytes or 1500 bytes plus overhead such as the Ethernet header, CRC,
>preamble, and IFG?
>
>I am assuming it is based upon the 1500 byte packet in the above
>example. In which case, I think I need to reduce my bandwidth value fed
>to tc to account for the overhead lest I exceed the real bandwidth. For
>example, the full sized Ethernet packet will be somewhere around 1538
>bytes so 1500/1538 * 1.544 = 1505kbits. Is that correct? Is there a
>guideline for a multiplier to use? I suppose if we are flooding a link,
>it is most likely with full sized packets. Thanks - John
>
>--
>To unsubscribe from this list: send the line "unsubscribe netdev" in
>the body of a message to majordomo@...r.kernel.org
>More majordomo info at http://vger.kernel.org/majordomo-info.html
>
Powered by blists - more mailing lists