|
Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com |
Subject: Re: pop-before-smtp (was Re: FreeInet and checking mail?)
From: Lawrence Greenfield (leg+
andrew.cmu.edu)Date: Wed Jun 07 2000 - 13:22:32 CDT
- Next message: Administrator: "Re: POP3 server for postfix"
- Previous message: Liviu Daia: "Re: Content filtering 101"
- In reply to: Alexander Nosenko: "Re: pop-before-smtp (was Re: FreeInet and checking mail?)"
- Next in thread: Brad Knowles: "Re: pop-before-smtp (was Re: FreeInet and checking mail?)"
- Reply: Lawrence Greenfield: "Re: pop-before-smtp (was Re: FreeInet and checking mail?)"
- Reply: Brad Knowles: "Re: pop-before-smtp (was Re: FreeInet and checking mail?)"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
[...]
No sane person will ever try to scale IMAP server even to few thousand
(simultaneous) users. It isn't "shot and forget" protocol SMTP is. Question
is, why it have to be done at all?
Wow, how surprising.
Our production IMAP server (which you can connect to using anonymous
IMAP at cyrus.andrew.cmu.edu) served well over 9000 unique logins in
the last week of the spring semester. It peaked at over 5500
simultaneous connections.
The standard IMAP quota for users is 25 megabytes, and is raised upon
request up to (I think) 80 megabytes.
Hardware:
cyrus.andrew.cmu.edu is a 2-processor Ultra 60, with 2 gigabytes of
memory.
Software:
cyrus.andrew.cmu.edu runs Cyrus IMAP. I think Cyrus IMAP performs
pretty well, but it certainly doesn't handle as many simultaneous
connections as many proprietary solutions which employ aggressive
threading techniques.
Yes, IMAP requires more resources than POP. Of course. It provides a
huge amount more features than POP, and these are features that users
want (or will want, once they understand that they're available).
Brad Knowles wrote:
Then uploads to the folders will be disabled, and timeouts will
be set low (I don't give a flying mongolian cluster sexual
intercourse what future RFCs might say).
Ironically, this _hurts_ scalability. Reestablishing connections,
reauthenticating users, and reestablishing protocol context is much
more expensive than a single idle TCP/IP session sending a NOOP every
30 minutes.
Brad Knowles wrote:
Companies that specialize in helping to build highly scalable
mail systems ought to either love him or hate him, depending on
whether you (or they) think it is actually technically possible to
build IMAP-based mail systems that will be physically capable of
scaling to handle millions of users.
In fact, companies currently provide such systems. Some ones of
interest:
www.messagingdirect.com Unix based IMAP servers
www.iplanet.com ditto
www.mirapoint.com IMAP appliances, meant to scale horizontally
www.software.com IMAP backed with large Oracle databases
Larry
- Next message: Administrator: "Re: POP3 server for postfix"
- Previous message: Liviu Daia: "Re: Content filtering 101"
- In reply to: Alexander Nosenko: "Re: pop-before-smtp (was Re: FreeInet and checking mail?)"
- Next in thread: Brad Knowles: "Re: pop-before-smtp (was Re: FreeInet and checking mail?)"
- Reply: Lawrence Greenfield: "Re: pop-before-smtp (was Re: FreeInet and checking mail?)"
- Reply: Brad Knowles: "Re: pop-before-smtp (was Re: FreeInet and checking mail?)"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]