Help

Front End Ports dashboard
The Front End Ports dashboard
The Front End Ports dashboard shows performance metrics for all metro node front-end ports, sorted by the busiest total
IOPS ports to the least busiest ports. You can sort each column in ascending or descending order.
Note that historical data is not available for ports. The dashboard automatically refreshes every five seconds, displaying data
from the last five second period.
For each front-end port, the dashboard shows the following performance statistics:
● Director—Displays data for all directors or a specific director in the cluster.
● Port—Displays data for all ports or for a specific port.
● IOPs (Total) — Total read and write operations per second.
● Read (IOPS) — Read operations per second.
● Write (IOPS) — Write operations per second.
● Queue Depth (count) — Number of outstanding operations.
● Reads (KB/s) — Bandwidth for read operations.
● Writes (KB/s) — Bandwidth for write operations.
● Read Avg Latency (usec) — Average latency or response time for read operations.
● Write Avg Latency (usec) — Average latency or response time for write operations.
NOTE: Metro node back-end ports metrics are not currently available.
Guidelines
● It is best practice to evenly balance all front-end ports whenever possible.
● For Queue Depth:
○ Queue Depth is sometimes referred to as number of outstanding operations. Do not confuse this with operations per
second (IOPS).
○ Excessively high queue depth values on a port tends to result in high (poor) host response time.
○ Maximum per port queue depth is around 1100 operations.
○ Front-end ports with consistently high per-port queue depths with high latency should be addressed. Spread the
workload across additional front-end ports.
○ Per port queue depths depends heavily upon application workloads and HBA adapter settings.
○ Follow the Dell EMC recommended values for host and HBA queue depths. Refer to the Host Connectivity Guides,
available on Dell EMC Online Support.
● For Latency statistics:
○ Satisfactory latency or response time depends heavily on the application's requirements.
○ It is difficult to give absolute recommended values for front-end port latency since it depends largely on back-end
latency.
○ Read or write latency values under 10msec are good, and greater than 100msec is usually cause for concern.
○
High latency ports should have host workload moved or shared across other lower latency ports, if the port's workload is
the problem instead of another underlying problem.
● For throughput (IOPS) and Bandwidth (KB/s) statistics:
○ Monitor per port MB/s maximum usage to avoid bandwidth saturation (for example, one 8Gbps FC port's maximum
available bandwidth is typically 800 MB/s or less.)
○ If values for these metrics are unsatisfactory, be aware of resource bottlenecks such as over-saturated front-end ports
or host initiators, or over-utilized metro node directors.
○ Identify performance-intensive applications such as nightly backups, or data warehouse applications that might cause
other latency-sensitive applications to suffer.
○ Check the Virtual Volumes dashboard to identify busy volumes.
Corrective actions
● Whenever possible, use all available front-end ports.
54
Monitoring the system