Sippy Softswitch Hardware Specification Guidelines


Stick to reputable name brand hardware.

KVM access is required for Sippy Support to access your installation for installation & upgrades.


** A note about Real CPU cores, versus Hyper-Threads or Virtual Cores

Our documentation states “physical cores”. This excludes virtual or hyperthreading cores.


For example, the Intel E5-2690 processor is marketed as having 8 physical cores and 8 virtual cores or “Threads”. This CPU counts as having 8 physical cores.




Sippy Softswitch

The figures below are based on our synthetic and real-case performance tests results with signalling traffic only and disabled features that decrease calls processing performance. The Sippy Softswitch does have built in media relays however utilizing this feature comes with a trade off with capacity.  This is an important consideration before procurement of your hardware.

Media Relay Scaling
In our baseline testing we have calculated that using the built in media relay and SIP signalling, 4 physical CPU cores are fully utilized by RTPproxy processes approaching limits of 1000 CC and 20 All Out CPS. When this is scaled to 2000 CC and 30 CPS we suggest 16 physical Cores.




Conversational traffic pattern with low-to-medium CPS rate (high ACD and ASR stats)



2,000 Concurrent Calls, 75 CPS Spec guideline


CPU
8 Core, 2.0GHz, Intel Xeon E5 or better or comparable AMD**
RAM
16GB or more (may depend on number rates/routes/environments)
Disk

Hardware RAID with Write-Back Cache (WBC) and Battery Backup Unit (BBU) support :
    - RAID 1 based on 2 X 240+ GB SSDs is a best fit ;
    - RAID10 based on 4 X 1 TB SATA drives will also work here, but less preferable ;
    - RAID 1 based on 2 X 1 TB SATA drives may be used as last resort.


Optionally RAID 1 based on 2 X 1 TB SATA drives for backups and system logs.

Software RAID is NOT recommended, as it could be causing the performance degradation under the load.
Network Interface
GigE Network Interface Intel(R) PRO/1000 PCI Express Gigabit Ethernet. Bonded (FEC) interface is recommended
Bandwidth
less than 10 MBit/sec


3,000 Concurrent Calls, 90 CPS Spec guideline


CPU
8 Core, 2.5GHz, Intel Xeon E5 or better or comparable AMD**
RAM
24GB or more (may depend on number rates/routes/environments)
Disk

Hardware RAID with Write-Back Cache (WBC) and Battery Backup Unit (BBU) support :
    - RAID 10 based on 4 X 240+ GB SSDs is a best fit ;
    - RAID 1 based on 2 X 480+ GB SSDs provides acceptable performance ;
    - RAID 10 based on 4 X 1 TB SATA drives may be used as last resort.


Optionally RAID 1 based on 2 X 1 TB SATA drives for backups and system logs.


Software RAID is NOT recommended, as it could be causing the performance degradation under the load.
Network Interface
GigE Network Interface Intel(R) PRO/1000 PCI Express Gigabit Ethernet. Bonded (FEC) interface is recommended
Bandwidth
less than 100MBit/sec

 

6,000 Concurrent Calls, 135 CPS Spec guideline


CPU
12 Core, 2.0GHz, Intel Xeon E5 or better or comparable AMD**
RAM
32GB or more (may depend on number rates/routes/environments)
Disk

Hardware RAID with Write-Back Cache (WBC) and Battery Backup Unit (BBU) support :
    - RAID 10 based on 4 X 240+ GB SSDs is a best fit ;
    - RAID 1 based on 2 X 480+ GB SSDs provides acceptable performance ;
  


Optionally RAID 1 based on 4 X 1 TB SATA drives for backups and system logs.


Software RAID is NOT recommended, as it could be causing the performance degradation under the load.
Network Interface
GigE Network Interface Intel(R) PRO/1000 PCI Express Gigabit Ethernet. Bonded (FEC) interface is recommended
Bandwidth
less than 100MBit/sec



8,000 Concurrent Calls, 200 CPS Spec guideline


CPU
16 Core, 2.5GHz, Intel Xeon E5 or better or comparable AMD**
RAM
32GB or more (may depend on number rates/routes/environments)
Disk

Hardware RAID with Write-Back Cache (WBC) and Battery Backup Unit (BBU) support :
    - RAID 10 based on 4 X 480+ GB SSDs is a best fit


RAID 10 based on 4 X 1 TB SATA drives for backups and system logs.

Software RAID is NOT recommended, as it could be causing the performance degradation under the load.
Network Interface
GigE Network Interface Intel(R) PRO/1000 PCI Express Gigabit Ethernet. Bonded (FEC) interface is recommended
Bandwidth
less than 100MBit/sec


10,000+ Concurrent Calls, 300+ CPS Spec guideline


CPU
24+ Core, 2.5GHz, Intel Xeon E5 or better or comparable AMD**
RAM
64GB or more (may depend on number rates/routes/environments)
Disk

Hardware RAID with Write-Back Cache (WBC) and Battery Backup Unit (BBU) support :
    - RAID 10 based on 4 X 960+ GB SSDs is a best fit

RAID 10 based on 4 X 1 TB SATA drives for backups and system logs.


Software RAID is NOT recommended, as it could be causing the performance degradation under the load.
Network Interface
GigE Network Interface Intel(R) PRO/1000 PCI Express Gigabit Ethernet. Bonded (FEC) interface is recommended
Bandwidth
less than 100MBit/sec



International and Call Center traffic pattern with medium-to-high CPS (low ACD and ASR stats)


1,200 Concurrent Calls, 150 'All In' 225 'All Out' CPS Spec guideline


CPU
12 Core, 2.0GHz, Intel Xeon E5 or better or comparable AMD**
RAM
16GB or more (may depend on number rates/routes/environments)
Disk

Hardware RAID with Write-Back Cache (WBC) and Battery Backup Unit (BBU) support :
    - RAID 10 based on 4 X 480+ GB SSDs is a best fit ;
    - RAID 1 based on 2 X 960+ GB SSDs provides tolerable performance ;


RAID 10 based on 4 X 1 TB SATA drives for backups and system logs.

Software RAID is NOT recommended, as it could be causing the performance degradation under the load.
Network Interface
GigE Network Interface Intel(R) PRO/1000 PCI Express Gigabit Ethernet. Bonded (FEC) interface is recommended
Bandwidth
less than 10 MBit/sec




Sippy Media Gateway (SMG) hardware guidelines

SMG Cluster is a beta product, and only available to enterprise customers participating in our beta program. If you wish to take part in this program, please contact sales@sippysoft.com


The relay or proxying of media through an intermediary system is useful for some specific scenarios, such as NAT helping and network topology masking. Relaying media will increase the network requirements, bandwidth resources and CPU resources of your system.


Media Relay causes the RTP stream between your client system and your vendor to travel via your Sippy Softswitch infrastructure.


The type of codec used by your client & vendor has a large bearing on the resources consumed. For example, here is the example of relative sizes between a selection of popular codecs:


Codec / Bandwidth usage examples



G.711
G.729/G.723
2,000CC / 30CPS
250 MBit/sec
40 MBit/sec
3,000CC / 50CPS
375 MBit/sec
60 MBit/sec
6,000CC / 100CPS
750 MBit/sec
120 MBit/sec
10,000 CC / 150 CPS
1.25 GBit/sec
300MBit/sec


Furthermore, the characteristics of traffic using the same codec can vary.


For example, the packetization window of a codec at 10msec will have a throughput of X and bandwidth usage of Y. If the packetization window is changed to 20msec, then the throughput changes to ~A and bandwidth usage to ~B.


The point to remember here is that all media traffic, even using the same codec is not always equal in terms of the resources required to relay it.


Low spec SMG hardware suitable up to 2,000 Concurrent Calls, 40 'All Out' CPS


CPU
8 Core, 2.0GHz, Intel Xeon E5 or better or comparable AMD**
RAM
8GB
Disk
Mirrored SATA hard drives or better
Network Interface
GigE Network Interface Intel(R) PRO/1000 PCI Express Gigabit Ethernet. Bonded (FEC) interface is recommended  


Example of SMG hardware specs stable under 3,000 Concurrent Calls, 60 'All Out' CPS, 180 sec ACD, 240,000 PPS and 128 MBit/sec


CPU
2 X CPU E5-2620 v2 @ 2.10GHz (12 physical cores in total)
RAM
8GB
Disk
RAID1 based on 2 X 1 TB SATA drives
Network Interface
Intel(R) PRO/1000 (3 vectors \ 2 processing threads) with Link Aggregation LACP setup 



Sippy Media Gateway (SMG) Cluster solution


Several Sippy Media Gateways (SMG) can be aggregated into the SMG Cluster solution that provides redundancy and load balancing between SMG nodes. Case one SMG node goes down due to hardware fault, network issues or there is a necessity to perform maintenance on that specific SMG node, SMG Cluster will continue its operation by proxying media sessions as usual whereas the load will be shared between existing SMG nodes available in SMG Cluster. Also, in case the SMG Cluster Maximum CC capacity is reached or all SMG nodes went down, then excess\remaining media sessions will be proxied through the Built-In RTPproxy service available on the Sippy SoftSwitch itself till SMG Cluster is back online.