Released: September 2017
This release presents new functionality on long term cluster trends and server redundancy.
It is often desirable, depending on the appetite for risk within the environment, to ensure that a number of devices within a cluster are redundant. That is, should these devices fail, there is still capacity in a cluster to ensure continued operation of the machines in that cluster. Redundancy is normally expressed in the form of N+x, where ‘x’ is the number of backup components required. In the case of virtualisation clusters, N is the number of devices needed to support the demand within the cluster adequately, and ‘x’ refers to the number of additional hosts. Therefore, when a cluster has a ‘high availability’ policy of ‘N+2’, it can withstand the failure of 2 hosts and still have sufficient capacity to support the aggregate workload of all the virtual machines in that cluster.
A new feature has been added to the Headroom Summary available from the right click menu at any point of the sunburst.
For each metric being used in headroom calculations, Capacity Planner will determine the cut-over points for server redundancy. The following is an example of a cluster level headroom summary for CPU utilisation.
There are now markers on the summary chart indicating ‘cut-off’ points for HA. These indicate the points at which, were demand greater than the level at the marker, there would less than 3, 2 or 1 redundant servers. This is calculated by subtracting the capacity of servers in the cluster from largest to smallest. Capacity Planner assumes a ‘worst case scenario’ regarding HA and to determine the cut- off point for less than one redundant server (<N+1), the capacity of the largest server in the cluster is subtracted from the clusters operational capacity. This is then repeated to determine the <N+2 cut-off by subtracting the capacity of the next largest server in the cluster and so forth.
In the example above, the total CPU capacity of this cluster, as an aggregate of normalised capacity of the devices in that cluster, is 92.4GHz. This cluster has set aside an ‘operational reserve’ of 20% giving an operational capacity of the cluster of 73.9GHz. The capacity of the largest host in the cluster is then subtracted from this operational capacity to provide a <N+1 cut-off point of 46GHz. It can be seen that the aggregate demand has passed the <N+2 marker, therefore the cluster has less than 2 redundant servers. The text to the right of the summary indicates how many redundant servers are available based on this metric.
This calculation is repeated for every metric that is used in calculating capacity headroom and will be presented for each metric.
At the summary level for the cluster, the metric with the least number of redundant servers will be used to determine the overall cluster redundancy.
A new event type has been introduced showing long term trends towards cluster capacity. They are available in a new tab in the events dialog window.
As with all events, they are ordered chronologically by default. Each event will indicate when a cluster will have less than 2 redundant servers (<N+2), less than 1 redundant server (<N+1) or cluster saturation. This is when the demand in the cluster will reach the current operational capacity of the cluster (total capacity – reserve headroom %age).
As with all events, the time-series will show the current demand and the forecast trend to the point where the threshold is crossed.
Click the links below to view other versions of Capacity Planner release notes.
|Capacity Planner Release Notes 100||Released: May 2019|
|Capacity Planner Release Notes 99.2||Released: March 2019|
|Capacity Planner Release Notes 99||Released: December 2018|
|Capacity Planner Release Notes 95||Released: May 2018|
|Capacity Planner Release Notes 94||Released: March 2018|
|Capacity Planner Release Notes 93||Released: January 2018|
|Capacity Planner Release Notes 92||Released: December 2017|
|Capacity Planner Release Notes 91||Released: October 2017|
|Capacity Planner Release Notes 90||Released: September 2017|