At what point, user and mail throughput is it recomended to add another mail.box? And how does it work?
Subject: Mail.Box
I don’t know that I’ve ever seen any recommendations on thresholds, but perhaps others will remember. If you’re seeing that mail is sitting in mail.box for a few minutes, yet the Router task is running and successfully routing mail, then I’d consider adding a second mail.box. This is done in the server’s Configuration doc - Router/SMTP tab - “number of mailboxes” field. If you set that to 2 and restart your server, you’ll then have a mail1.box and a mail2.box which both function as mail routers.
In case it helps, my personal rule of thumb is to not have more mail.boxes than I have CPUs in the server. When I first came to this company, they had two CPUs in a server but 5 mail.boxes, since the documentation didn’t give thresholds - just that you can have between 1-10, so they thought 5 was a good median number. As a result, the mail.boxes were always queued up because the server was always thrashing trying to look at all 5 mail routers. Bumped that down to 2 and mail started routing within seconds.
Subject: RE: Mail.Box
Some clarifications.Adding multiple mail.box’s doesn’t imply more ‘Routers’. There is only one Rotuer, and it has to look in all the mail.box’s, which is why things got better once you reduced the number.
The purpose of having multiple is to avoid write lock contention when you have very many concurrent tasks trying to write to it at the same time. (many local clients, or may SMTP clients). General rule of thumb is that if you think you might have a problem, add 1 more. The more you add teh less effective each addition becomes. So the advantage of going 1 to 2 is greater than 2 to 3 and so on.