Re: Port numbers and IPv6

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

 



In message <20050715205554.15355872E3@xxxxxxxxxxxxxxxxxxx>, Noel Chiappa writes
:
>    > From: Ned Freed <ned.freed@xxxxxxxxxxx>
>
>Let me make sure I understand you here:
>
>    > IMAP4 has the characteristic that you often have a huge number of
>    > incoming connections, only a few of which are active at any given time.
>    > Designing servers to accomodate huge numbers of connections is a bit
>    > tricky, but workable: ...
>    > The 65536 limit, OTOH, has to be dealt with by using multiple server IP
>    > addresses, which in turn usually require multiple interfaces ...
>    > ... that doesn't mean nobody is hitting the 65536 limit imposed by
>    > source port numbers. They are, it causes problems
>
>You're saying that there are servers which have close to (or more) than 65K
>connections to a *single client IP address* (i.e. it may be a NAT, with a
>number of hosts behind it)? (If a server is talking to a number of different
>client IP addresses, it can have up to 65K connections to *each*.)
>

Ned isn't the first person I've heard this observation from.  Yes, 
there are some really large-scale systems that run into this limit.

Sure, there are work-arounds, such as assigning multiple IP addresses 
to the server and using DNS-based load balancing.  That doesn't change 
the fact that the basic design has run afoul of an address space limit.

Circa 1974, in a computer architecture class, I heard Fred Brooks point
out that *every* successful computer design eventually ran out of address 
space.  The same applies to network protocols.


		--Steven M. Bellovin, http://www.cs.columbia.edu/~smb



_______________________________________________

Ietf@xxxxxxxx
https://www1.ietf.org/mailman/listinfo/ietf

[Index of Archives]     [IETF Annoucements]     [IETF]     [IP Storage]     [Yosemite News]     [Linux SCTP]     [Linux Newbies]     [Fedora Users]