Our SMTP frontend died (twice) around 8am on Friday morning, so some ofthe stats for this are partial stats and some of them are missing abouttwo hours of data. That said, this week we:

• got 10,583 messages from 249 different IP addresses.
• handled 17,948 sessions from 1,258 different IP addresses.
• received 257,246 connections from over 50,000 different IPaddresses.
• hit a

I've been experimenting with AoE lately, and as a resultI've been looking at the protocol more than I did in my earlierexposure . Unfortunately, the more I look at the AoEprotocol, the more uncomfortable I get.

The AoE protocol is quite simple; requests and replies are simpleEthernet frames, and a request's result must fit in a single replypacket. This means that the maximum read and write sizes per request arebounded by the size of the Ethernet frame, and thus on

I've been considering writing a fully asynchronous ATA over Ethernettarget driver (I'm not entirely happy with the current one and itsperformance). If Linux asynchronous IO worked on sockets (specificallyraw network sockets), there is a nice simple design where you just setup a pool of buffers and then cycle each one through a little statemachine (network in to disk IO to network out).

(This works especially great for AOE because the protocol is both basedon raw packets and completely unordered,