Server Loads/hardware standards - recommendations

[Michael Kain]

Julian Field wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
>
>
> Scott Silva wrote:
>   
>> Julian Field spake the following on 10/17/2006 12:29 PM:
>>   
>>     
>>> Scott Silva wrote:
>>>     
>>>       
>>>>> Michael Kain spake the following on 10/17/2006 9:52 AM:
>>>>>   
>>>>>         
>>>>>           
>>>>>> Recently, I've gone from handling 40k messages /day to nearly 30k/hour. 
>>>>>> The change has surfaced in the last month or so.
>>>>>>
>>>>>> My current setup:
>>>>>> Dual P3 1.13
>>>>>> 1GB Ram
>>>>>> FC5
>>>>>>
>>>>>> Mail gateway running MS/clam/SA forwards scanned mail to internal mail
>>>>>> server (when there's a problem, users hit send/receive and that doesn't
>>>>>> cause an error..thus avoiding immediate call) I've used Julian's clam/sa
>>>>>> install script (which is awesome), and read posts relating to new
>>>>>> releases before upgrading/such.
>>>>>>
>>>>>> With spamassassin enabled, the batch list grows and grows, was up to 95k
>>>>>> at one point.. disabling SA in MS cleared that out fairly quickly.  I've
>>>>>> wiped the SA/bayes temp files thinking bayes was backing up, however, it
>>>>>> seems that is not helping.
>>>>>>
>>>>>> What I would like an opinion on is this... Am I trying to do too much
>>>>>> with the hardware that I currently have?  Or do I put together a bigger
>>>>>> beefier machine?
>>>>>>
>>>>>> -Mike
>>>>>>     
>>>>>>           
>>>>>>             
>>> You can make a huge difference to the amount of spam you have to process 
>>> with 2 tools:
>>>
>>> 1) milter-gris
>>> 2) milter-null
>>>
>>> Number 1 implements grey-listing. There are a lot of discussions about 
>>> greylisting on the web, and a lot of people are very wary of it 
>>> initially. I was too. Then I ran a test with a handful of the fussiest 
>>> email users I have (I've got about 2000 users in total). I told them I 
>>> was implementing something new, but refused to tell them what, so they 
>>> would not have any pre-conceptions about it. They *all* loved it, and 
>>> none of them reported any problems at all. So I implemented it across 
>>> all of my users, who are very fussy Computer Science and Electronics 
>>> academics, as well as the students. That was about 6 months ago, since 
>>> when I have had *1* complaint, which I dealt with by adding them to the 
>>> whitelist for it.
>>>
>>> So my conclusion with greylisting is test it with some very fussy users, 
>>> then roll it out to everyone.
>>>
>>> Number 2 implements back-scatter detection. Basically, what this does is 
>>> get rid of all the "This message could not be delivered..." notices that 
>>> weren't generated in response to your own users' mail. It doesn't throw 
>>> away all of them, so that if your users mistype an address, they still 
>>> get the error message from it. But all the delivery failure messages 
>>> generated by forged spam get killed.
>>>
>>> Between these 2, you will remove 80-90% of all the mail coming into your 
>>> site, without losing any genuine real mail at all. This will make your 
>>> hardware go a hell of a lot further, and you will find you don't need to 
>>> spend any money on new hardware at all.
>>>
>>> My MX servers used to just about cope. Then I implemented these 2 
>>> techniques and they now just tick along quite happily, getting very bored.
>>>
>>> Both of the above techniques can be done very easily in sendmail and 
>>> Postfix using the milters which are available from www.snertsoft.com. I 
>>> thoroughly recommend them to everyone.
>>>
>>> Jules
>>>
>>>     
>>>       
>> So the addition of the two milters doesn't add that much load?
>>   
>>     
> No, they don't. And they save way more load than they cause!
>
> Jules
>
> - -- 
> Julian Field MEng CITP
> www.MailScanner.info
> Buy the MailScanner book at www.MailScanner.info/store
>
> Need help customising MailScanner?
> Contact me!
> Need help fixing or optimising your systems?
> Contact me!
> Need help getting you started solving new requirements from your boss?
> Contact me!
>
> PGP footprint: EE81 D763 3DB0 0BFD E1DC 7222 11F6 5947 1415 B654
>
>
>
> -----BEGIN PGP SIGNATURE-----
> Version: PGP Desktop 9.5.0 (Build 1112)
> Comment: (pgp-secured)
> Charset: ISO-8859-1
>
> wj8DBQFFNf0kEfZZRxQVtlQRAv6sAKCwu7kfBHIL7TK/UHTcMG65W+egqACeIXOu
> 0Hoea8EOk74OgOH+J0/iphQ=
> =TPuu
> -----END PGP SIGNATURE-----
>
>   
Outstanding!  I will give these a shot and post the results.

Thank you!