Home > Storage > PowerScale (Isilon) > Product Documentation > Cloud > Introduction to APEX File Storage for Azure > Test methodology
To understand the performance characteristics of APEX File Storage for Microsoft Azure, a standard benchmark tool was used. It is used for conducting sequential read tests with a request size of 128 KiB and sequential write tests with a request size of 512 KiB, with NFS version 3.
Regarding the sequential read/write tests, here are a few specifics:
The OneFS inline compression and inline deduplication features were left at their defaults (enabled) during all tests. The benchmark workload specified 0% compressible or deduplicated data blocks. The data reduction ratio for the dataset is 1.0 in the tests.
Note: There is a known issue that the inline deduplication and inline compression feature is not enabled by default at the disk pool level. The first MR (Major Release) will fix it after GA (General Availability). You can also reach out to your sales representative if you need an immediate fix.
In the tests, we used Standard_D48s_v5 clients to generate I/O to the OneFS clusters. The ratio of clients to nodes is 2:1. Each client instance contained 48 vCPU cores and 192 GiB memory, and the network bandwidth was 24 Gbps. For each test, we followed the rule of ensuring that the aggregate bandwidth of data disks is sufficient at the VM level. See Appendix C: recommended data disks configuration details for optimal performance.
All performance tests in this document were performed in the South Central US location.