I have a question, possibly not directly related to James, that perhaps the wise people on this list may be able to answer. I am finding that smtp traffic forms a large (60%-70%) part of the bandwidth used by a server that runs a number of services, including http and smtp.
In trying to find out more about this issue, I am using tcpdump and related tools to measure bytes coming into and out of the network interface, that is, bytes on the wire. I have also created a mailet, placed at the top of the root processor, that logs details of every mail coming in, including sender, date, remote host, and size. The latter is calculated by writing the MimeMessage out to a byte array (using MimeMessage.writeTo()) and measuring the byte array length.
I find that there is sometimes a large difference between tcpdump measured bandwidth (incoming only) and the total from my mailet over the same period of time. I am noticing regular differences in size of the order of 100%. For example, between 00h00 and 01h00 last night tcpdump measured 1.5Mb incoming smtp bytes and my mailet totaled 160kb worth of email. Since bandwidth is precious (and expensive) commodity, this is of some concern.
I would guess that one or more of the following is true:
1) Certain emails are being rejected after being recieved by James but before they are placed in the root processor.
2) SMTP traffic is *very* inefficient in certain circumstances.
3) My measuring techniques are broken.
Can anyone shed anymore light on this?
Thanks, Craig
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
