The HiveMQ MQTT broker is a distributed, stateful application, which naturally requires a fair amount of read/write operations to the hard disc. This means that choosing the right type of storage, when using a cloud provider can have a positive impact on HiveMQ’s overall performance.
For AWS we recommend the following:
io1 instances offer constant I/O whereas gp2 capabilities increase at larger sizes.
...
According to our internal benchmarking we recommend the availability of at least 900 IOPS per node for any HiveMQ Deployment.
Amazon Web Services
When hosting your HiveMQ broker on AWS the most cost efficient way of achieving these recommended IOPS is using 300GB gp2 storage.
Of course you can use io1 instances with guaranteed 900 IOPS as well.
Azure
We recommend the use of P15 managed storage.
Google Cloud Platform
On GCP the most economic way of achieving the required IOPS is using 1000GB Standard persistent disks.
Alternatively 128 GB SSD persist disk will provide the necessary minimum of 100GB storage and 3840 IOPS.
These recommendations are based on general benchmark tests. Feel free to reach out to us, if you are seeking specific advice for your use case.