Deploying Server on AWS: EC2 and EBS recommendations?

Question

Does anyone have any experience running Alteryx Server on AWS?

The MTR for Server currently calls for 1TB of disk, 16 GB RAM, and 4 CPUs (2.5GHz approx) for the performance server. That level of spec puts me at a m4.xlarge (4 vCPUs, 16GB) or c4.2xlarge (8 vCPUs, 16GB).

I found a deck from an Alteryx partner that stated they only used a c3.large ( 4 vCPUs, 7.5GB Ram). CherryCasino: A Modern Cloud Based BI Architecture .

EC2 Questions:

* Are the Compute Optimized Instances better suited to Alteryx server?
* Is the standard deployment (8GB Ram) usually sufficient?

Regarding the 1TB of disk. I assume SSD, but do I need provisioned IOPS or can I just go with general purpose?

This server will likey be used heavily once per week for a bunch of extracts from a RedShift cluster. I think the extracts will run between 1-5GB.

adrianloong · Answer

I wanted to share with everyone how Alteryx & Tableau server could be setup with AWS

See slide 13 below:

https://adrianloong.files.wordpress.com/2016/03/amaysim-tableau-conference_feb2016.pdf

Am also finding that the load on Alteryx server is much higher than Tableau server as Alteryx does most of the workflows scheduling and pushes to Tableau server

adrianloong · Answer

I'm running nearly 2-3 workflows per hour on my alteryx server hitting redshift and producing a tableau tde.

Our current config is a r3.2xlarge instance. (8 VCPU, 61Gb memory, 1x160 SSD) running extracts from redshift into alteryx workflow and tableau tde.

Remember to set your alteryx scratch disk to the attached SSD volume especially when processing large datasets, you will see a significant performance improvement in alteryx) as the attached volume is much faster than other EBS volumes.

KoryC · Answer

Our own Alteryx Gallery is running in AWS, configured in a multi-node deployment spread across several instances.

We certainly recommend following the tech spec guidelines, which does indeed place you in the m4.xlarge range, but your experience/needs may vary. Alteryx Server will still function on lower hardware specs (such as a c3.large as you've referenced), but there may be a degredation in performance depending on the types of workflows and load that you expect to run.

The workflow type (data being consumed, tools being used, etc.) will dictate if the workflow leans more towards being compute-intensive, RAM-intensive, HD-intensive, network-intensize, or a combination of the 4. For example, if you're doing a lot of drivetime processing, that's typically more compute-intensive, wheras a lot of sorting and joining will be more memory intensive. Large data reads or writes to a local datasource may find the disk being the bottleneck, and reading/writing across a network, or consuming web services, will often show the network being the slowest point.