DPACK Basics: Read / Write Ratio
A Read/ Write Ratio compares the percentage of read operations to write operations between the server and its storage devices. This can be seen in many performance-related measurements, such as IOPS, Disk Throughput, latency, etc. Understanding this ratio is critical in storage device and array design.
DPACK displays a stacked graph of read (indicated in magenta) and write (indicated in blue) operations for showing storage device performance. This means that the total number of operations are comprised of a percentage of reads and writes. Typical observed workloads reflect a “70 / 30” Read/ Write ratio, meaning that, on average, 70% of operations consist of reads while 30% consist of writes. Read/ write ratio is applied to IOPS, but it can be applied to other storage device performance metrics, as well.
DPACK will publish the numerical value of the R/W ratio for the top 10% of observed IO period within a collector run.
Note: you may also see Read/ Write ratio written as: R:W::x:y (e.g. R:W::70:30, where R = 70% and W = 30%)
General rules of thumb:
- The observed average R/W ratio globally is approximately R:W::70:30
- Being different from the average isn’t wrong. It is just different.
- Common workloads where W > R: portions of a database (specifically log files, “scratch area”, tempDB), non-persistent VDI
- Common workloads where R:W::50:50: Backups, clustered applications, cloud
- Common workloads where R > W: many normal workloads including reporting
- Most commonly used with IOPS, but is also used for design and troubleshooting in conjunction with IO Transfer size, Disk Throughput, and latency
IO Transfer Size