IOPS- Storage performance & bandwidth Calculator for Virtualized environments

Technologies and computing paradigms have evolved but the general laws of physics and mathematics still apply when assessing performance and bandwidth on virtualized or cloud environments.
Performance and bandwidth in any computing environment still depends upon how many compute operations can be performed in a slice of time and how much information can be transferred in and out of compute, storage and network sub-systems.
The most cited reasons for performance issues in virtualized and cloud environments are related to storage performance and throughput. Thus, storage performance, throughput and bandwidth capacity planning become critical to when rolling out workloads in virtualization and cloud IT infrastructure landscape. Storage IO performance and bandwidth planning exercise becomes especially important when deploying high IO demanding ERP applications like Oracle , SAP , any large or mission critical dataset backups and restores.
The aim of this article is to aide virtualization / cloud application owners and infrastructure architects to collaborate with storage specialists to cater for the data IO (Input/Output) performance and bandwidth requirements for workloads (User<->Application<->VM/Workload <-> Host <-> LUN <-> N/W <-> Storage).
Sample IO performance and bandwidth calculation walkthrough looks like this:
• Let us start by defining what is IOPS ; IOPS stands for I/Os per second or Input/Output operations per second
• Subsequently to calculate the amount of data transferred from any sub-system per second (data transferred / second ) can be calculated as = IOPS * IO transfer unit size typically 4KB or 8KB but can be up to 64k
o Storage IO data block size typically is determined by the block size that is defined when formatting the storage volumes or sub-system before first use
• Typically a single 15K SCSI disk supports =180 IOPS
• Therefore data transfer rate of a single 15K SCSI disk will be:
o 180 IOPs * 4K data block = 720KB/s Or = 703.13 KiB/s Or = 43.20 MB/min Or = 5.76 Mbps Or = 2.59 GB/h
For example; if a production database backup is running on 10 Spindle LUN (each spindle = 15K SCSI disk) then it will support the following data transfer rate:
(180 IOPS) * (10 Spindles) * (4KB data block) = 7200KB/s
7200KB/s = 57.60 Mbps Or = 432.00 MB/min Or = 7.20 MB/s Or = 6.87 MiB/s Or = 25.92 GB/h
So, the maximum data you could backup in the example above would be 25.92 GB/h and hope that meets or beats your backup window.
Check out the free online tool / service for your own bandwidth conversion calculations at: http://web.forret.com/tools/bandwidth.asp
In a virtualized or private cloud environment you would be easily able to find out what the storage and network sub-system configurations look like to perform the performance and bandwidth calculations BUT in public cloud environments not all cloud hosting providers provide the maximum IOPS per VM/workload or any sort of IOPS benchmarks for their environment and you may as well end up benchmarking performance and bandwidth metrics yourself which may vary per public cloud or per cloud VM instance type.
In a nutshell, know your target infrastructure environment well, ask the right questions and do the IOPS calculations before deploying applications to avoid future complications. It is always significantly harder to fix IOPS related performance and bandwidth issue after the deployment is in place.
Harry Sangha - Inventor, entrepreneur and cloud computing evangelist. Founder of Cloudcreo - a Cloud Management Platform and Appliance Factory.