Server hardware specs

I am getting ready to order my hardware for my asterisk deployment.
I am Looking at this setup as the Base Unit for 300 users:

Processor:
Quad Core Xeon 2nd Processor E5335, 2x4MB Cache, 2.0GHz 1333MHz FSB, PE2900 (311-7063)

Memory:
8GB 667MHz (8x1GB), Dual Ranked DIMMs

Hard Drive:
4x 300GB, SAS, 3.5-inch 15K RPM Hard Drive

Hard Drive Controller:
PERC 5/i, Integrated Controller Card

NIC:
Embedded Broadcom NetXtreme II5708 GigabitEthernet NIC

Feature
Integrated SAS/SATA RAID 5 PERC 5/i Integrated

If anybody has any eperience with about 300 users their advice would be appreciated on if this server will be sufficent for my deployment
Thanks
Brian

I really could use advice on if I need to beef up this order

When you ask for hardware advice, please add the intentional usage and witch functions you are to use!

How many users are not that interesting, but concurrent calls and what functions are you planning to use like, meetme, voicemail +++ and are you doing transcoding and is it only SIP +++.

But the short answer is, Yes the machine is over dimensioned.

[quote=“Mats”]When you ask for hardware advice, please add the intentional usage and witch functions you are to use!

How many users are not that interesting, but concurrent calls and what functions are you planning to use like, meetme, voicemail +++ and are you doing transcoding and is it only SIP +++.

But the short answer is, Yes the machine is over dimensioned.[/quote]

Hi Mats,

In your experience how much does the memory as well as the CPU processor consume for each call if all are SIP call transaction while the asterisk is transcoding? What is the percentage difference of it when there is no transcoding so that we could compute the amount of memory as  well as the processor that we will try to purchase.

Kind regards,

Transcoding:
There is a table in asterisk that shows the cost for coding from one format to another, I don’t remember the command right of my head.

And for CPU usage, from a performance test I did a year ago showed that AMD Opteron was superior to Intel Xeon and dual CPU’s didn’t give much better performance in our testbed.

And for memory, we wasn’t able to fill 4G in our testbed and we used meetme and that is one of the more memory hungry apps.

Sorry that I can’t give you any hard facts, then would I break a NDA!

[quote=“Mats”]Transcoding:
There is a table in asterisk that shows the cost for coding from one format to another, I don’t remember the command right of my head.

And for CPU usage, from a performance test I did a year ago showed that AMD Opteron was superior to Intel Xeon and dual CPU’s didn’t give much better performance in our testbed.

And for memory, we wasn’t able to fill 4G in our testbed and we used meetme and that is one of the more memory hungry apps.

Sorry that I can’t give you any hard facts, then would I break a NDA![/quote]

You mean to say (based on your test) that using DUAL core CPU wouldn’t add much help on performance efficiently on handling asterisk calls and features? Is that right? Ma y I could ask what kind of benchmarking method did you conduct on ASTERISK so that I could replicate your findings on my test server if it is fine with you?
By the way may I know what is the load average of the AMD Opteron and XEON.

Kind Regards,

at most we will have 30-40 calls going we only have 2 t-1 pri’s
We will be using the meet me confrence center etc… and the g.729 card will be installed… there will be IAX2 or SIP trunks going to 3 remote offices.

[quote=“Decompression”]at most we will have 30-40 calls going we only have 2 t-1 pri’s
We will be using the meet me confrence center etc… and the g.729 card will be installed… there will be IAX2 or SIP trunks going to 3 remote offices.[/quote]

No Voicemail stored on the asterisk server = nearly no HD is needed, 10G HD is sufficient.

And 8GB RAM is also over dimensioned, ~4GByte is enough.

And 2 CPU is also over dimensioned, 1 is enough.

we will have voice mail on the server.

[quote]You mean to say (based on your test) that using DUAL core CPU wouldn’t add much help on performance efficiently on handling asterisk calls and features? Is that right? Ma y I could ask what kind of benchmarking method did you conduct on ASTERISK so that I could replicate your findings on my test server if it is fine with you?
By the way may I know what is the load average of the AMD Opteron and XEON.

Kind Regards,[/quote]

What I mean is that two CPU’s with dual core doesn’t boost the performance at all in the testbed I used.

And the testbed was sipp bombarding asterisk and as for performance measuring was vmstat, mpstat, top, time, strace and ltrace used if I remember right.

Some of the result was a bit odd and Im unshure if mpstat is realiable in a multicore environment because the values was sometimes inconclusive.