Very High CPU Utilization (EC2 Instance - m.medium)
I'm using an Amazon EC2 Wowza 3 instance (ver 3.1.2) (ami-b4f9bfe6). Without any stream (inbound/outbound), the CPU utilization is around 2-3%. But when I started to stream a live event using FFsplit (at 600 kbps bitrate), it shoots up to 70%, with only two (2) viewers. Then when I streamed another live event using RTMPWorld (@ 400 kbps bitrate), from 70% CPU utilization, it further shoots up to 95%, with a total of only three (3) viewers. I'm wondering why the server behaves like that?
According to Richard of Wowza Support, any pre-built AMI's need not be tuned (based on the Performance Tuning guide provided by Wowza) since they are automatically tuned. So, where does the server's problem possibly come from?
Also, is there a way for me to know if there were changes made on my server that deviate from the configuration of an automatically tuned server? A standard configuration file for a tuned pre-built AMI (ami-b4f9bfe6) would be a great help so I can have it compared to my server.
I'm now planning to migrate to Amazon EC2 High-CPU Extra-Large Instance (c1.xlarge) so that it could support my two (2) channels with three (3) transcodes each. What if the number of channels has increased and my single c1.xlarge EC2 instance could no longer support these number of channels?
Let's stick to your example above. You've said that I will be needing seven (7) c1.xlarge EC2 instances to support 30 channels, even 10 of which are pass-through, leaving 20 that will require transcodes. How should I configure all these seven (7) c1.xlarge EC2 instances to allow simultaneous 'live' applications? Or what is the best solution for this?