How to ramp up and down with servers that have dependencies on each other?

[hpclover]

Hi,

I have a system that runs in EC2 but I have never used pool manager.
And recently we start to talk about how to automate ramping up and
down computing power based on load and I started to look for a
instance pool management solution.

At bare minimum, our system requires 4 instances to run. 3 of them
need to be started first (can be independently started) and fully
functional before the 4th one can be started. Let's call the 4th
instance "processing node". When load increases in the work queue, we
can add more processing nodes to certain point. Beyond that point, we
need to either launch another set of 4+ instances (3 of them to be
prerequisite instances + several processing nodes) or we can swap out
less powerful instances and put in higher performing instances (e.g.,
change an instance from m1.large to m1.xlarge or to m2.4xlarge to take
advantage of larger memory thus less disk IO). Basically in our
system, instances are assigned different roles and each role runs
specific services. instances must be ramped up proportionally and
some need to be launched before others can be launched. For exmaple,
if instance of role A needs to be ramped up by 4, then we must ramp up
instance of role B by 1 first. Or If an instance of role A needs to
be bumped up to c1.xlarge, then we must bump up instance of role B to
m2.xlarge.

Is there a way to do this kind of pool management?

Thanks.

[rajashekar.s]

Hi,

You want to use Lifeguard or can switch over to Auto Scaling feature of AWS??

 

Best regards