The new Clarity based HTML5 user interface (UI) that VMware has adopted across its product portfolio has been a big win for new and existing customers. Since the debut of the new vCenter client in vSphere and vSAN 6.7, it has been quite common to hear from users who have made the transition on how quickly they've adapted to the new UI. Adapting easily to a new UI is not always a given. User experience designers are usually faced with a myriad of considerations and design questions that have neither a right or a wrong answer, which always makes introducing UI changes a challenge. The great feedback is a testament to the VMware's desire to get it right.
Not only does the new user interface provide a faster, more intuitive, and elegant operational experience for our users, but it affords our design teams much greater flexibility in the logical placement of events, and actions across the interface. Some of these adjustments show when viewing vSAN performance metrics courtesy of the performance service. Let's review a few adjustments to viewing performance-based information in the HTML5 UI found in vSAN 6.7.
Performance Service Basics
While the vSAN performance service is still an optional service for a vSAN cluster, it has grown to play a critical role in monitoring the behavior of VMs, host resources and the cluster in a vSAN powered cluster. This is one of the reasons why the performance service will be turned on by default for all new clusters using vSAN 6.7. The service runs on each ESXi host in a vSAN cluster, as shown in Figure 1.
The data collected is stored as an object, per the assigned storage policy. Keeping the performance service separate minimizes the amount of work for vCenter, as all it is responsible for is rendering the data collected from the vSAN performance service. The object can be viewed in the UI at the cluster level, underneath “Virtual Objects” as shown in Figure 2.
The beauty of the vSAN performance service is that due to vSAN’s deep integration into the hypervisor, the service can present meaningful metrics throughout the entire storage stack.
The vSAN performance service provides storage metrics at three different levels: VMs, hosts, and cluster-level metrics. Of the three levels, the host level statistics provides the largest collection of metric types, and is a wealth of information for understanding host-level behaviors. Some adjustments were made to make these host level metrics more consumable.
Host-level storage device metrics
For host-level vSAN performance metrics, disks groups and disks have been consolidated under one category, and will expose the associated metrics for three distinct items:
- The disk group as a whole
- The buffer/cache device for the selected disk group.
- The individual capacity devices of the disk group.
As shown in Figure 3, consolidating this view, paired with the new UI behavior of condensed graphs (described later in this post) gives the user the ability to identify anomalies more effectively.
The metrics presented depend on which option is selected. For instance, when a capacity device is selected, there will be two additional graphs. The "vSAN Layer IOPS" and the "vSAN Layer Latency" metrics. The vSAN performance service calls these out at the capacity disk layer for good reason. Capacity devices living across a disk group may have different levels of demand placed on them, and in some circumstances, can be extremely useful when determining areas of potential contention.
Host-level network device metrics
The host network device metrics are presented in a slightly different way for vSAN 6.7. The two menu items of "VMkernel Adapters" and "VMkernel Adapters Aggregation" found in vSAN 6.6 have been replaced with a single menu called "Host Network." This allows the user to select a specific VMkernel adapter, or all of the VMkernel adapters on a host. The UI warns the user that the statistics are representing all network I/Os processed in the network adapters used by vSAN, and that the counted network I/Os are not limited to vSAN traffic only. Figure 4 shows the selection option available for VMkernel adapters.
While not a change in the 6.7 UI, one area to emphasize is how the packet loss rate is measured. The graphs note packet loss using a permille (per thousand) sign as a percentage, or %o. This means that if a reading shows a packet loss of 23%o, it is referring to a percentage of 2.3%. Permille measurements are sometimes used when a smaller unit of measure is better suited to describe a finer level of conditions or changes. This is analogous to the use of centimeters instead of meters, or basis points used in the financial world. Take a look at Reliable Network Connectivity in Hyper-Converged Environments for a better understanding of performance degradation due to packet loss.
Other UI behavior changes
The UI will now arrange multiple graphs based on overall browser screen size. In a smaller window, multiple graphs will align in a top-to-bottom list, but when enlarging the screen, the will transform to smaller side-by-side graphs, as shown in Figure 5. This greatly reduces the need for scrolling, which is especially important when attempting to correlate variations in activity through multiple metrics.
A subtle but important change to all of these graphs is that the user is able to now drag and zoom to a targeted time period. As shown in Figure 6, upon zooming in, the "Zoom Out" button will be available to restore the view to the previous window of time. This ability to quickly adjust the time window through a drag and zoom action is a significant improvement in evaluating and correlating performance behaviors.
The vSAN performance service aims to improve the overall management experience of vSAN, by measuring the right data, in the right way, from the right location. The capabilities of it grow with nearly every new version of vSAN, and when paired with the new Clarity UI in vSphere and vSAN 6.7, is more powerful than ever.