| Author |
|
simplikate
Intermediate
Joined: June 20 2007
Location: United States
Posts: 7
|
| Posted: June 20 2007 at 10:48am | IP Logged
|
|
|
We are looking to initially deploy two 512 licenses of your product - to give us 1024 lines. We may then upgrade to 2048 lines in the next 18 months. Our goal is to provide a VOIP-enabled Customer Service application nation wide to about 3000 communities in the US, which will reach up to 2 million residents.
If possible, can you advise if your platform in the following hardware configuration would work? You guys are obviously experts in multi-line deployments - so before I commit to a 5K purchase of two of your 512 line licenses, I'd like to get some quick advice on how it would be deployed.
We are planning on using two servers - each a Quad Core Xeon Processor 4MB Cache, 2.0GHz, 1333MHz FSB, and 4GB RAM server - as our main work-horse servers. We have others (4 more) we could put into service if needed, but are not quite as powerful. We have a burst-able DS3 40Mbit circuit for Internet connectivity for each data center for bandwidth.
Is there a recommended configuration for hosting 512 lines? With the server listed above - should we limit each server to 250 lines, and put the other servers into service for the remaining 512 lines? Any assistance would be a great help.
|
| Back to Top |
|
| |
support
Administrator
Joined: January 26 2005
Location: United States
Posts: 1666
|
| Posted: June 21 2007 at 12:39pm | IP Logged
|
|
|
Hi simplikate,
Thanks for waiting while we gather information for this response.
We do not publish “hard” performance data for the products due to the huge number of deployment variations that exist. As a rule, deploy on the fastest host machine you can afford and using the largest network data pipe you can get. We know this is a somewhat “lame approach” so we will try to discuss additional information that may be useful to you and other users when making your deployment decisions.
Previous posts:
There are a few previous posts that discuss deployment. The topics discussed in these posts are useful.
Important question about Media Engine - Deployment
What are the Max lines and limitations
Server capacity
Performance benchmarks
Deploying ISTP of 5000 users - Questions
The primary limiting factor when deploying a VOIP app built around the media engine is how well the host machine and the app handle media.
For example: Let’s assume you will deploy a VOIP app capable of using full duplex G729 for all call media. G729 currently has the lowest bandwidth requirements for media. 74 raw bytes of data every 20Ms goes over the Ethernet wire (148 bytes full duplex). This can be verified using a network sniffer (ethereal).
This is 1184 bits full duplex per 20Ms or 59,200 bits per second. If you are using 100 megabit Ethernet, the theoretical total number of calls that can exist is 100,000,000/59,200 which is roughly 1689 active calls.
In reality though, there will be Ethernet collisions for placing packets on the network and other errors. So conservatively we cut this total call volume by 50% to 844 active full duplex G729 VOIP calls.
Note:
When using uLaw or aLaw media for the calls, the numbers are as such:
214 raw bytes per Ethernet frame per 20Ms…
same as 428 bytes full duplex…
same as 3424 bits per 20Ms full duplex…
same as 171,200 bits per second full duplex.
Using 100 megabit Ethernet and using the logic from the G729 computation above, we can achieve approximately 292 calls at the same time.
This gives you some kind of indication what may be possible. The other limiting factor is can the host server machine handle the network I/O for that call load effectively. That is a question each end user has to determine based on the hardware being deployed. At this time, most vanilla server machines and network gear seem to be OK.
Support
|
| Back to Top |
|
| |
|
|
Active Topics 
Member List 
Search 
Help 
Register 
Login
LanScape Support Forum -> LanScape VOIP Media Engine™ - Pre-Sales Technical Support
Topic: Recommended Hardware Configuration Question
