What are the System Requirements for a Metric Insights Server?
This article lists the minimum system requirements for the Metric Insights server. This assumes a single node for a Simple Install. It can also apply to a single node for an Orchestrated Deployment (multi-node environment). The system requirements for Orchestrated Environments are listed for each pod (Kubernetes, OpenShift), the same applies for each service (Docker Swarm), or task (AWS ECS).
System Requirements for Production Instances
Simple Installation
MI application server:
- 8-Core CPU
- 32 GB RAM
- 1 TB+ disk total
- If the disk is split into several partitions, please ensure enough space is allocated to the following directories:
/var-
/optor/app /home/tmp
- Docker lives in
/var/lib/dockerby default. The MI filesystem lives in/opt/miby default (or/app/mialternatively)
- If the disk is split into several partitions, please ensure enough space is allocated to the following directories:
On AWS, this equates to an m5.2xlarge EC2 instance type.
On Azure, this equates to a D8ds v4 VM instance size.
On GCP, this equates to an e2-standard-8 instance type.
Orchestrated Environments
RAM required by pod:
- Web Master 16 GB
- Web Slave 8 GB
- Data Processor 16 GB
- Data Analyzer 8 GB
- Image Generator 4 GB
- Console 1 GB
- Redis 1 GB
CPU is not less than 1 core per each pod, which is equivalent to 1000m in Kubernetes.
System Requirements for MySQL Server
The MySQL database must be running on a separate server, preferably on a database cluster managed by a DBA team or a native cloud solution like Amazon RDS.
- 8-Core CPU
- 32 GB RAM
- 1 TB+ HD total
- If there are different volumes for specific filesystems, please ensure enough space is allocated to the following:
/var
- MySQL data files by default live in
/var/lib/mysql
- If there are different volumes for specific filesystems, please ensure enough space is allocated to the following:
- In Version 7.0.1a: We support MySQL 8.0.37
On AWS, this equates to a db.m5.2xlarge RDS instance type.
On Azure, this equates to an 8 core Compute Gen 5 server.
System Requirements for Pilot Instances
Simple Installation
Single Server running both the application and MySQL:
- 4-Core CPU
- 16 GB RAM
- 500 GB disk total
- If the disk is split into several partitions, please ensure enough space is allocated to the following directories:
/var-
/optor/app /home/tmp
- Docker lives in
/var/lib/dockerby default. The MI filesystem lives in/opt/miby default (or/app/mialternatively)
- If the disk is split into several partitions, please ensure enough space is allocated to the following directories:
On AWS, this equates to an m5.xlarge EC2 instance type.
On Azure, this equates to a D4ds v4 VM instance size.
On GCP, this equates to an e2-standard-4 instance type.
Orchestrated Environments
RAM required by pod:
- Web Master 1 GB
- Web Slave 1 GB
- Data Processor 2 GB
- Data Analyzer 2 GB
- Image Generator 1 GB
- Console 512 MB
- Redis 512 MB
CPU is not less than 1 core per each pod, which is equivalent to 1000m in Kubernetes.
For pilot instances, the specs above should be more than sufficient. If there are plans to run large datasets and/or large bursts as part of the pilot, we recommend moving to an 8 core / 32 GB RAM server for better performance.
A Note About Virtual Machines
See Deploy Metric Insights as a Virtual Appliance for more information.
Metric Insights can be deployed as a virtual appliance, for pilot instances as well as for production instances. If you choose this option, then your virtual machine should meet the requirements laid out above.
It's important to understand that the Metric Insights virtual appliance is a virtual machine. You do not need to create your own virtual machine to serve as a host. In fact, the host machine should be a physical machine, not a virtual machine. If your IT department offers to provide a virtual server to serve as a host for Metric Insights, then they must certainly be running one of the common virtualization environments. In that case, simply send the Metric Insights virtual appliance to them and ask them to 'import' it.