amazon ec2 - getting started with EC2 for compute-intensive (non-web) parallel application -
i'm using libsvm regression analysis. works champ. 3-parameter grid search optimize parameters model maxes out 4 cores on 2.66 ghz intel box, , still have wait couple of hours generate single model.
this seems job amazon ec2.
i've seen plenty of tutorials , introductory material on using ec2 web-related tasks.
but if have small compute-intensive custom ansi-c program want run multiple instances of on ec2? can provide pointers on how (or buzzwords search for)?
i don't think quest different of web application. stack different of course, regardless – the principles remain same.
as commented on question: elastic map reduce might you're looking parallelize work easily, etc.. if limited, cloudera. ready-to-rumble hadoop distribution support ec2 well.
if map-reduce not liking, need setup own instance. speaking, keypoints follows:
- you want figure out way start ec2 instances.
- you want figure out way bootstrap , configure them.
- cluster/network?
starting ec2 instances
if don't require auto-scaling or custom interface, aws console extremely job. have select ami (amazon machine image) suitable project. i'd either official ami or ubuntu-based (if remember correctly, ubuntu used linux on ec2).
but , liking. (and don't know enough project.)
once figured out setup works you, easiest way clone work setup own ami , start instances it, etc..
bootstrapping
bootstrapping can using ec2 calls user-script
. allows pass shell script instance, execute calls setup stack, etc.. i'm not sure required in case, etc.. in case comment or extend answer, go detail here.
cluster/networking
this wild guess since i'm not sure code does, or how works, etc.. if it's not necessary, i'd scale out using single instance first. can lot of cores , ram provisioned ec2. depending if work requires more ram or cpu, high-cpu , high-memory instance types.
you can start off t1.micro
, can get free even , go there.
let me know if helps!
Comments
Post a Comment