We've discussed the capacity-related features in Aria Operations before: Assess, Reclaim, and Rightsize. But we've never gone into detail on the Capacity Remaining details, let's do it.
Capacity Remaining as an idea is intuitive, in practice its a projected metric calculated against Datacenters, Clusters, ESXi Hosts, VMs, Datastores, and more. Let's take a look at Cluster Capacity Remaining.
The Capacity Remaining tile in the middle shows Capacity Remaining at 9% (5.45TB) and indicates that's how much Disk Space is left, which represents the most constrained resource in the Cluster (the other two being CPU and Memory). Clicking the Capacity Remaining tile takes you to the Capacity tab and gives more detail.
You'll now see CPU, Memory, and Disk details for capacity remaining. As you can see, Disk Space Remaining (Allocation) has the lowest amount of capacity remaining (9%) and as such is assigned as the Cluster Capacity Remaining metric.
How is this being calculated? If you click the How Capacity Remaining is Calculated link you'll be given a visual.
Keep in mind that you can define the Buffer levels in your Policies, mine are set to 0%.
Once buffer is taken off the top, Capacity Remaining is calculated based on "usable capacity - projected peak usage over the next 3 days". For Disk Space the Allocation and Demand Totals and Usable are the same (my overcommit ratio is 1:1).
Total and Usable Disk Space are metrics available on the Cluster, as shown here.
Next, Capacity Used is based on the projected peak usage over the next 3 days. For Allocation we use the Total Provisioned Disk Space with Overhead metric and for Demand we use the Utilization metric, shown below.
As such, the capacity analytics engine is projecting the peak usage over the next 3 days to be: 28% (16.71TB) for Demand and 91% (54.89TB) for Allocation.
Now that we have our Usable (actual) and Used (projected) metrics we can calculate our Remaining metric "remaining = usable - used (projected)". In my case the Disk Space Remaining for Allocation is 9% (5.45.TB) and 72% (43.64%) for Demand. This 9% remaining represents the minimum of the big 3 (CPU, Memory, and Disk) and as such is considered to be how much capacity we have remaining in the Cluster.
Now let's look at CPU.
CPU Allocation Total and Allocation Usable are based on Cluster Metrics based on a 3:1 Overcommit Ratio, mine look like this.
CPU Demand Total and Demand Usable are based on these two Cluster Metrics.
Demand and Allocation CPU Used is based on the projected peak usage of both over the next 3 days, specifically these metrics.
As indicated, the projections are: 42% (211 vCPU) for Allocation and 36% (120.51 GHz) for Demand. Leaving us with 58% and 64% remaining respectively.
Finally, for memory we have the following.
Total is 3TB, 2.25TB available after HA, and 9TB total on the Allocation model based on a 3:1 Overcommit Ratio. These are actual Cluster metrics.
We are then calculating our projected peak used metrics based on these:
For Allocation Memory we project 13% (874.74GB) and for Demand Memory we project 40% (914.85GB). Based on these we project capacity remaining to be 87% (Allocation) and 60% (Demand).
You can do this same exercise for Datacenters, ESXi Hosts, VMs, and more. Thanks to Broadcom Engineer Tigran Mkrtchyan for his insight on this blog!
Comentarios