[AG-TECH] AG & MTU Size

[Bill Nickless]

Jokes aside--

TCP can work with MTUs greater than Ethernet (>1500) because the IP routers 
tell TCP when the MTU they want to use can't be forwarded all the way along 
the path from source to destination.  The IP routers send ICMP messages to 
support this TCP Path Discovery (see RFC 1191).

Media streams, and especially IP Multicast media streams, aren't set up to 
perform path MTU discovery.  I don't even think it's well defined in the 
multicast case.

There's a theoretical guaranteed maximum and a practical maximum MTU for 
media streams.  Theoretically, you're guaranteed 576 bytes MTU because of 
the minimum IP MTU in RFC 791:

     Total Length is the length of the datagram, measured in octets,
     including internet header and data.  This field allows the length of
     a datagram to be up to 65,535 octets.  Such long datagrams are
     impractical for most hosts and networks.  All hosts must be prepared
     to accept datagrams of up to 576 octets (whether they arrive whole
     or in fragments).  It is recommended that hosts only send datagrams
     larger than 576 octets if they have assurance that the destination
     is prepared to accept the larger datagrams.

     The number 576 is selected to allow a reasonable sized data block to
     be transmitted in addition to the required header information.  For
     example, this size allows a data block of 512 octets plus 64 header
     octets to fit in a datagram.  The maximal internet header is 60
     octets, and a typical internet header is 20 octets, allowing a
     margin for headers of higher level protocols.

In practice, just about everyone runs on an Ethernet, which gives you 1500 
byte MTUs.

One of the primary reasons to move TCP to an MTU of 9000 bytes is that the 
recovery speed of TCP when it encounters congestion uses the Maximum 
Segment Size (MSS) rather than bytes.  So a TCP using jumbo frames will 
recover from congestion more quickly than a TCP using 1500 byte frames.

AG-style UDP-based media streams don't play fairly in the presence of 
congestion, and their MTU doesn't have an impact on recovery from 
congestion.  Based on that argument, I wouldn't recommend moving AG to a 
larger MTU.


At 11:58 AM 5/14/2003 -0400, Mike Weaver - MICS/USDOE wrote:
>Does anyone have any information or has anyone done any research on the
>effect of MTU size on the AG?  The Joint Engineering Team (JET) of the Large
>Scale Networking (LSN) group of the Interagency Working Group on IT R&D
>(IWG/ITRD) out of the National Coordination Office for Information
>Technology Research and Development (NITRD) reporting to the Whitehouse
>Office of Science and Technology Policy and the National Science and
>Technology Council - whew...that's a mouthful! - is suggesting an MTU size
>of 9K+.
>
>Does anyone know the effects of trying to tune Windows/Linux's TCP/IP stacks
>to work with 9000-byte MTUs?  How about the video and audio tools
>(vic/rat) - any issues that anyone might know about?
>
>Thanks,
>
>Mike
>
>--
>Mike Weaver
>Sr. Network Administrator
>SC-31/Germantown Building
>US Department of Energy
>1000 Independence Avenue, SW
>Washington, D.C. 20585-1290
>Voice: 301-903-0072
>Fax: 301-528-2701
>Email: weaver at er.doe.gov

===
Bill Nickless    http://www.mcs.anl.gov/people/nickless      +1 630 252 7390
PGP:0E 0F 16 80 C5 B1 69 52 E1 44 1A A5 0E 1B 74 F7     nickless at mcs.anl.gov