ThoughtSpot engineering has performed extensive testing of the ThoughtSpot appliance on various Amazon Elastic Compute Cloud (EC2) and Amazon Elastic Block Store (EBS) configurations for best performance, load balancing, scalability, and reliability.

You can find information here on which configuration of memory, CPU, storage, and networking capacity you should be running for your instances. There are also details on how to configure your placement groups.

ThoughtSpot AWS instance types

Use case



Data shape

Total cluster
data size

Per VM
Data capacity

Recommended
Instance type

vCPU/RAM

Boot volume

Data volumes

Standard

(1KB/row)

Up to 2 TB

250 GB

r4.16xlargea

64/488

200 GB

2X 1 TB

>2 TB

384 GB (Large)

r5.24xlarge

96/768

200 GB

2X 1.5 TB

Up to 100 GB

100 GB

r4.8xlargeb

32/244

200 GB

2X 400 GB

Up to 20 GB

20 GB

r4.4xlargeb

16/122

200 GB

2X 400 GB

Thin rows

(<300 bytes/row)

Any

192 GB

m5.24xlarge

96/384

200 GB

2X 1 TB

(a) Use the sizing calculators on each cloud provider to plug in expected customer discounts to arrive at the proper recommended cloud instance type.

(b) Use the small and medium instance-type configuration. Refer to: Use small and medium instance types.

Regions, availability zones, and placement groups

AWS instances are configured to a location with regard to where the computing resources are physically located. You must specify a region, an availability zone, and below that, a placement group.

AWS nodes in a ThoughtSpot cluster must be in the same availability zone (and, therefore, also in the same region).

A placement group is a logical grouping of instances within a single availability zone. Placement groups are recommended for applications that benefit from low network latency, high network throughput, or both.

ThoughtSpot relies on high connectivity between nodes of a cluster, which is why creating a placement group is recommended. Being in same placement group will give you the best shot at the highest bandwidth across AWS EC2 instances and the lowest latencies. This will make the node-node network reach the closest AWS promised specs. Our default recommendation for a multi-instance setup requires a placement group since it works best for our application performance. Also, AWS will provide jumbo frames (9000 MTU) support in such situations, and they don’t charge extra for being in the same placement group.

Having said that, ThoughtSpot will still work with EC2s in the cluster across placement groups in an availability zone.