Enterprise-level Server Deployment on AWS

[bryanbumgardner]

This is a slightly different thread than others asking about AWS because we are considering a large, enterprise-level deployment and I was hoping to get some feedback on our idea.

We're looking at setting up a Server to serve up to 75 users over the next year or so. We're going to host this server on AWS, but I'm confused about the best sort of EC2 instance to handle this work.

The workflows will be run daily by any number of users, who will mostly need lots of joins and complex sorting, sometimes from databases but mostly from uploaded assets. 

That makes me lean toward this option: 

m3.xlarge:

8 x vCPU

30 GiB RAM

2 x 80 SSD Storage

Problem is, I have no understanding of the impact of multiple users on a server. Right now I'm focused on this hardware because that's the immediate ask, if I need to add more Server Workers/Controllers that can happen later.

Is this a solid choice for 50-75 users trying to run light workflows?

Thanks!

[jrgo · Accepted solution]

Hi @bryanbumgardner,

The only spec I'd suggest modifying is your storage. 500GB is the least I'd recommend, especially if there will be a lot of uploaded files/assets, but also for all the temp/work files when the job is processing, result outputs files, log files and - most importantly - the persistence (MongoDB) database.

Difficult to say if that setup will be sufficient because it's unsure how many jobs it'll be asked to run at any given time or how many users will be on the Gallery (if enabled) concurrently. However, this setup (with the larger HD) will be a good starting point. When you run though the Alteryx system settings, you'll see an option to set how many simultaneous jobs are allowed to run. This is where you want to start conservative... 2-4. You'll need to closely monitor the jobs and see if you can up this number based on whether jobs are getting hung up. 10 jobs running 2 at a time will finish faster than if they were running all at the same time.

This document, Alteryx Tech Overview, contains much more detailed information, specifically starting on page 11, "Scaling out Alteryx".

Hope this helps!

Best,

Jimmy

[StephenR · Accepted solution]

We have a theoretical user base of about 400, but a daily user base of about 50-100 (we have yet to see more than 1 or 2 workflows queued at a time for more than a minute), and we have a number of intensive spatial/reporting workflows that use large files.  That being said, we are running an m4.xlarge with 4 vCPUs and 16 Gb RAM, and it handles things pretty well.  One thing to note is that the server license is per 4 CPUs, so to do 8 CPUs you would need to purchase a second license, and I would argue that you would be better off adding another node at that point than getting a larger server.

 

We have it set to allow 4 workflows simultaneously.  So far, the only real area that we see affecting speed is network bandwidth and file size.  We've handled that by placing some of our large, often used files on the server itself so that r/w is all local.  Smaller files associated with only one workflow we let Alteryx Designer upload along with the workflow.  Because of that we have two 1TB hard drives, one for Alteryx and all the temp, etc. files, and one for our data files.