|
|
|
|
||||||
![]() |
|
|
LinkBack | Outils de la discussion |
|
|
#1 |
|
Messages: n/a
Hébergeur: |
Hello everyone
I've been requested to estimate the hardware needed for construction an smtp relay server. No need for Antivirus / SpamAssasin checking. The expected traffic is 50Kb messages to ~60,000 recipients. I've tried to google for benchmark / Capacity planing formulas with no results. Any advice would be very welcome and appreciated. Thank you, Maxim. |
|
|
|
#2 |
|
Messages: n/a
Hébergeur: |
hq4ever@gmail.com wrote:
> No need for Antivirus / SpamAssasin checking. > > The expected traffic is 50Kb messages to ~60,000 recipients. You should provide more detail - e.g. are those recipients local or remote, and if remote, do you want to estimate the overall delivery time or the time it takes to queue them? In what time frame are those messages handed over to postfix? Is it one single message to all those recipients? If the recipients are remote, the bottle neck is most likely the internet connection, not some server software. Just my 0,02 EUR ;-) -hannes |
|
|
|
#3 |
|
Messages: n/a
Hébergeur: |
Hannes Erven wrote: > hq4ever@gmail.com wrote: > > > No need for Antivirus / SpamAssasin checking. > > > > The expected traffic is 50Kb messages to ~60,000 recipients. > > You should provide more detail - e.g. are those recipients local or > remote, and if remote, do you want to estimate the overall delivery time > or the time it takes to queue them? Yes, of curse. The recipients are remote. This is separate message delivery to each recipients in a time frame of 6hr. Expected rate is 200msg / Min. I would like to estimate as you say the "overall delivery time". > In what time frame are those messages handed over to postfix? Is it one > single message to all those recipients? > The sending application sends the messages at a constant rate of ~100msg/Sec and then stops until the next day transaction. > If the recipients are remote, the bottle neck is most likely the > internet connection, not some server software. > Network bandwidth should be sufficient as the server sites in the ISP's server farm. Thank you. > > Just my 0,02 EUR ;-) If I only had those 0.02 EUR each time some one offered... ![]() > > -hannes -- Maxim Vexler. |
|
|
|
#4 |
|
Messages: n/a
Hébergeur: |
hq4ever@gmail.com wrote:
> Hello everyone > > I've been requested to estimate the hardware needed for construction an > smtp relay server. > No need for Antivirus / SpamAssasin checking. > > The expected traffic is 50Kb messages to ~60,000 recipients. > > I've tried to google for benchmark / Capacity planing formulas with no > results. > > Any advice would be very welcome and appreciated. > > > Thank you, > Maxim. If you are relaying email, a critical factor is the ability of the remote sites to receive email. It they're slow or on a slow network, that impacts your ability to deliver. You could estimate a time it takes to delivery an average message, multiply that by the total number of messages, and divide it into the number of concurrent Postfix processes, you could get a ballpark delivery time for all the messages. Don't forget the time it takes for the relay to accept the incoming messages. For example, assumptions: 50,000 messages Average receive time: 2 seconds Average send time: 10 seconds Number of Postfix processes: 300 50,000 x 2 sec. = 100,000 sec. for receipt 50,000 x 10 sec. = 500,000 sec. for delivery ------- 600,000 sec. total 600,000 sec. / 300 process = 2,000 sec 2,000 sec / (60x60) = 5.6 hours This is blue sky stuff for 1 modern x86 system. I'd want a minimum of 2 Postfix servers. The hardware type depends on your computing budget and business enviroment: high end corporate data center, versus a couple of desktops stuffed in a closet. Costly caching SCSI disk arrays are faster. -- Greg |
|
|
|
#5 |
|
Messages: n/a
Hébergeur: |
> I've been requested to estimate the hardware needed for construction an
> smtp relay server. > No need for Antivirus / SpamAssasin checking. > > The expected traffic is 50Kb messages to ~60,000 recipients. > I've tried to google for benchmark / Capacity planing formulas with no > results. > Any advice would be very welcome and appreciated. Maxim, Designing a system like that requires several sets of data or workload profiles to get a grip on your needs. Also, redundancy, failover, and reliability can play a critical role in your design. Several companies, including ours, does this type of work all the time. Hope that s, -Phil Carinhas pac(at)fortuitous(dot)com -- Fortuitous Technologies Inc - http://fortuitous.com Performance Engineering, Capacity Planning & Design |
|
![]() |
| Outils de la discussion | |
|
|