Cloud Player Agent?

We have deployed one EC2 instance (m1.medium) and we have deployed only 8 transactions to this location thus far. CPU utilization on the instance (m1.medium - is single core) is basically pegged while transactions are running. The average duration of the 8 transactions at this location is 4 - 4.5 second range. I tried opening a support case to get some EC2 sizing recommendations but I never really got a straight answer. Does anyone have any real life experience deploying players in the EC2 environment and what they have seen from a performance and reliability standpoint?

Thanks!

Joe

Find more posts tagged with

seum_playback

seum

Accepted answers

All comments

aLTeReGo

SeUM includes the Cloud Instance sizing calculator which is based on our own internal testing and best practices. However it's difficult to make gross approximations since not all transactions are created equal, and no two transactions are the same. The number of steps in a given transaction, the technologies being used on the page, and page complexity all play a role in calculating how many transactions can be reasonably managed by a single player instance. In the next release of SeUM we include a Player load indicator (current and historical) for just this reason. These gauges will allow customers to understand precisely how much capacity their players are consuming, as well as how much unused capacity they have free. If you'd like to sign-up for the beta and give it a try you can do so here.

jmanek

There is also a way how to change the instance type of an existing instance (in case the existing configuration is no longer sufficient etc.).

All you need to do is:

1. Go to the Amazon AWS site (http://aws.amazon.com/console)

2. Log in (My Account / Console - AWS Management Console) and go to the EC2

3. Display existing Instances

4. Select existing instance and Stop it (right click / Stop)

5. Right click and select Change instance type

Depending on the original instance type you can convert:

Small instance (m1.small) to Micro, Small, Medium (t1.micro, m1.small, c1.medium, m1.medium)

Medium instance (c1.medium) to Micro, Small, Medium, Large, 2XLarge, 4XLarge, 8XLarge (t1.micro, m1.small, c1.medium, m1.medium, m1.large,

m1.xlarge, m2.xlarge, m2.2xlarge, m2.4xlarge, c1.xlarge, cc1.4xlarge, cc2.8xlarge, cg1.4xlarge, hi1.4xlarge)

Large instance (m1.large) to Micro, Small, Medium, Large, 2XLarge, 4XLarge, 8XLarge (t1.micro, m1.small, c1.medium, m1.medium, m1.large, m1.xlarge, m2.xlarge, m2.2xlarge, m2.4xlarge, c1.xlarge, cc1.4xlarge, cc2.8xlarge, cg1.4xlarge, hi1.4xlarge)

klinejw

We are running the 2.0 beta and while I love the concept of the player load indicator - we are seeing that only a couple invocations of recordings cause that indicator to show 100+ percent. I agree with your logic on the fact that transactions are not created equal. We are going to shuffle around some of our recording assignments to the EC2 instance and see if we can narrow this down to certain recordings.

We are still wanting to hear from anyone that is using EC2 and what experience they are having in that environment.

jiri.tomek

Load indicator shows current load so if you assign few recordings to agent at once, it can show high values. For average load over time use Player Load chart. It will give you better overview.