Using Amazon EC2 for PixInsight processing with Streaming like Google Stadia

S

scheidtmann

Member
Hi there,

I have configured an Amazon EC2 instance to use streaming (like Google Stadia) for processing in PixInsight. So far a t3.xlarge (4 vCPUs, 16 GiB RAM) gives me double the performance as given by my old core i5 laptop (with 6 GiB RAM) for the same BatchPreProcessing project (processing time is shorter by a factor 1/2). This setup should be usable with interactive processes like Blink, which currently is unusable on my laptop (but I haven't tried yet).

I described how to setup here: https://github.com/jscheidtmann/cloud-pi/blob/master/manual_setup.md
My aim is to later automate the process of setting up using e.g. terraform or ansible and then do all my astropix processing in the cloud.

I will post updates here. If you're interested and willing to try it out, please let me know here or on github.

Thanks, Jens
 
And how scalable. If you did 200 subs, can you assign 32 cpu's for example? (And how that prices out vs, oversimplifying, having 4 CPU's for 8 times as long)?
 
Great questions from Andrew and Linwood. @scheidtmann, have you captured any Benchmark data from Pixinsight? For example, would it be possible to get a 100x speedup compared with a fast desktop computer?
 
And I suppose you simply copy over your license file (perhaps from S3) every time you start up, given that you mention in the README that you disconnect the volume after every use?

I am very curious to here how you chunk up processing - fire up the instance per dataset after transferring data to S3?

Cheers!
 
You must log in or register to reply here.
Back
Top