HPE Nimble with Infosight – second part

In the first episode I promised that we will go trough what Infosight can do and to show lovely bits under “Labs” tab. Before we start let me remind you to always integrate your vCenter Server(s) with your Nimble array(s). This will open up a new horizon for sure, worth it!

If you have done this already, then your ” Infrastructure -> Virtualization” tab should be full with VM entities.

Your mileage may vary, but under “Datacenters” section your vCenter datacenter objects are being shown, including cluster, host, datastore and VM count. I only have one real one, the other “Datacenter X” is a demo datacenter.

The “Clusters” tab is a bit more useful since this is showing not just inventory, but operational data as well. If you have an over-provisioned host in a cluster, this will tell you.

We are getting there, where the real differentiation spots are. Welcome to the “ESXi hosts” area, why Infosight is unique. It loads the datastore objects from vCenter’s inventory and matches that with the volume names from the Nimble array. So no more LUN/datastore mismatch. I’ve seen many occurrences when VMware team said to increase the space for a given datastore and the storage team added space to an another one. Also some when impact was a little greated, since the incorrect datastore was unexported….. This will not happen anymore. This window also shows the VMFS version, the block size, the used and total capacity.

Finally the “VM” view. It can look like a mess at first sight, but spend some time with it.

It gets clear when a specific VM entity is selected as that opens the performance view for that VM. Below you can see that it is reporting the latency, VM generated write and read IO counts and also that the capacity inside the VM – needs VMtools.

Click to the third tab in here, called “VMDKs” and open up your beer. This gives you the details of read/write IOPS, read/write latency, read/write throughput for the last 24 hours and generic performance. I have only one VMDK below shown, but this is neat!

Let’s go back to the “VMs” tab for a second. If you click and check what is under this menu you can see one called like “VM I/O Contention Trend”.

After selecting a datastore the following is being shown. This diagram tells me when a VM was hitting my volume – and my array – with how hard and the same info about the latency the VM had at that time. Ok. I can see that a VM below – the same VM by the way – is generating lots of IOs and its latency is around 19.2ms average.

If you select “VM CPU Contention Trend” the very same diagram will show VM CPU usage and CPU ready on a given host.

Capacity managers on your side now if you want to purchase a Nimble array and slowly operations teams also.

Let go into the “Labs

There is one report called “Inter-Volume Performance & Contention“. After generating a report covering the last 7 days it tells exactly which volumes are the most utilized in terms of IOPSs and how their latency is looking like.

My Nimble-VMFS2 datastore is working the hardest and my Nimble-VMFS1 is having the highest latency in certain periods. If you hover the mouse to the colorful stripes it will tell exactly the value.

Recurring Performance Patterns” is something you can get a picture – literally – about when and how hard the array is working.

Let’s see random read IOPS in the last month.

Each cell is 5 minutes average. You can see this can be run against Sequential read, random write, sequential write also. Also data can be downloaded in a CSV, so you can feed your systems you want to feed with this. What this is for? This for capacity related planning and management, with a sip of troubleshooting option, like if you experience slowness in every Sunday 10PM, this can draw that to you why.

Next one which is interesting is “Pool Performance Telemetry”. This is the data which feeds Infosight AI for Performance recommendations. (Don’t look on my cache hit rate, this is an all flash array, so does not use read cache).

The killer feature “Resource Planner”

You have an array and you want to add new workload or extend current ones. This will help you to calculate if your array can take it or not.

Let’s say I will add an Oracle DB for my CRM with 1TB dataset and this will consume 2800 IOPS in bursty fashion.

This will display how my current array will look like after I deploy this volume. AF1000’s CPU will bump up to 28% ulitization and I’ll be still be ok with the capacity, as no impact is being told. Also showing some larger models how they would look like from CPU utilization perspective.

Now let’s increase the IOPS requirement for this CRM oracle to 12800. Spot on! AF1000 CPU would run up to 108% utilization, so better to go for a higher model as current array of mind would not be able to host this.

You can also run existing workload based models, like what happens if your existing volume would need to be a destination in a P2V project, doubling the capacity and performance need.

It reports I’ll be still ok to put that onto this array.


I guess now you feel how you can squeeze to most out of a Nimble array, but not just that, but also you VMware infrastructure. Infosight can help you getting some sponsors in a pursuit for a new storage array.

There will be a third post with some other reports, stay tuned!