or How I learned to like The Workload Concentration Bubble
As an avid user of live optics to analyse environments for sizing, I was recently discussing the outputs with a peer, I realised there was a lot more to this particular graph than at first appears.
For those that don’t know, Live Optics is a free online tool by Dell to collect and analyse data about your IT workloads and environment – www.liveoptics.com . It allows you to create some really great graphs of your workloads in Windows, VMware, Linux, UNIX. You can run it on non-Dell environments as well.
One graph that appears on the executive summary report is the Workload Concentration Bubble.
This is basically a snapshot of the performance your disk system.
I really like the Workload Concentration Bubble (WCB) graph, because in one view you can see the multiple items of data about the disk system. Each circle represents a Disk (or LUN) and shows the following:
- Quantity of disks (number on graph)
- Each disk is first sorted by its 95th percentile IOPS number from highest to lowest and its performance plotted in the y axis
- Relative Capacity (size of the circle)
- Latency or QoS (colour of circle)
- Distribution of capacity (position in x axis)
Let’s delve in the example above for some assumptions/analysis:
- There are 19 disks (number of circles)
- They range in size from 30GB to 3TB
- 20% of the capacity does 50% of the work
- The most performing disk – the Red one – (@ 1200 IOPs) also has latency > 20ms
- The second most performing disk – the yellow one – (@ 800 IOPs) also has latency > 10ms
The big thing to look for is any yellow or red circles as they do indicate a potential disk IO bottleneck. A little bit more investigation will be required (Live Optics has many more graphs you can drill down on), but I would look at SSDs or NVMe as possible solutions. If we are having latency > 20ms at 1200 IOPs then spinning disks are probably not the best answer.
So, in summary, a good quick overview of the disk system performance and my first go to when looking at designing Storage and Hyper Converged solutions.