Re: Jumbo MTU sizes

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

 




> This is a good example of needing to know the real problem before
> fine-tuning an assumed solution.

> TCP is a windowed protocol, thus increasing the MTU size will not help.

Don't think of increasing the MTU in such a limited fashion.  Although in a lot of cases, changing the MTU would be a throughput benefit.  You can optimize an MTU for a certain error and dropping rates.  I think its naive to say that increasing (or decreasing) the MTU wil have no benefit what so every.  Besides the 1500 byte MTU is only optimized for 10-bit ethernet.  In any case, I'm not limiting the discussion to a simple MTU change.  

Assume for example that you have two modem connections (for each direction of transfer) and that in each direction the MTU is relatively large.  Thus your argument for dropped or prioritized frames is moot.  If you imploy synchronization and a sonet-type scheme, you can almost eliminate the initial-per-packet transmission latency.  Besides, its obvious that the MTU can be optimized for certain session types (e.g. interactive or download) especially on a pt-to-pt link.  The real implementation problems surface when trying to implement the remote proxy on the other side of the link since an MTU of 1500 is assumed for ppp0 running from an 802.3 based network.

I'd like to point out that a lot of design choices where upper layers depend on some condition from lower protocol layers sometimes have adverse effects for scalability and functional improvement in the future.  In this example, since the ppp device is implemented as an ethernet device, it assumes a certain MTU.  Also, on a TCP/IP network data flow can be uniquely identified by the sequence number and IP.  Since this identification on the transport side depends on the IP address, this has very annoying consequences for mobile networks.  The TCP sessions don't migrate well with changing IP addresses.

Regards,
Mazhar
-
: send the line "unsubscribe linux-net" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[Index of Archives]     [Netdev]     [Ethernet Bridging]     [Linux 802.1Q VLAN]     [Linux Wireless]     [Kernel Newbies]     [Security]     [Linux for Hams]     [Netfilter]     [Git]     [Bugtraq]     [Yosemite News and Information]     [MIPS Linux]     [ARM Linux]     [Linux RAID]     [Linux PCI]     [Linux Admin]     [Samba]

  Powered by Linux