All flash and hybrid flash arrays are becoming more the norm than the exception in the datacenter. The cost of the drives goes down while the capacity increases, almost monthly. It is tempting to assume that the blinding speed that SSDs deliver eliminates the need to keep an eye on basic storage monitoring.
Speed is the paint that covers many things, but it can mask other issues. This can make the performance from your flash drives not so… flashy. The bottleneck can often shift to another point in the system.
Where did my Flash go?
The below group of IntelliMagic Vison minicharts shows a VSP G1500 showing high read and write throughput accompanied by high read response times. These are indicated by the black arrows.
The red and yellow borders around each chart show the level of risk associated with each chart. Red indicates a high risk, yellow a moderate risk, and green means no risk during the time period.
A knee-jerk reaction would be to blame the back-end disks, but a quick drilldown into the drive busy rating chart, shown in Figure 2, indicates low to moderate disk utilization.
They are only roughly 20% busy on average and the peaks on the graph do not correlate with the observed peaks in throughput or response time shown in Figure 1. Furthermore, the green border of the chart indicates that the drives are operating within specifications.
There it is…
Further inspection of the minicharts shows a peak in port response time, indicated by the red arrow in Figure 3, that corresponds to the peaks in read response times previously noted.
Drilling down from the minichart to the port response time graph (Figure 4) gives more detail and shows the timing matches exactly.
Drilling into the Port Response Time graph into the individual ports (Figure 5) shows that the four host ports are heavily utilized during this time, with port CL1-D being most heavily hit.
Looking at the negotiated port rate for each host port in Figure 6 shows that each of the ports negotiated at 8 Gbps while they are capable of 16 Gbps, rendering them unable to perform at their full potential.
The Brocade switches to which they are connected are only capable of 8 Gbps, so the host ports are constrained by connected infrastructure. The bottleneck has moved from the array to the switch.
Resolving this would require migrating the host ports to a switch capable of 16 Gbps. Obviously, while feasible, this is not something that can be quickly corrected.
|Link ID||WWPN for Port||Data rate for port||Port technology||Operational Status for Port|
Figure 6: Connected Port Rate
We’re not out of options, however. If we drill down further we can see the host groups that contribute to the port utilization during the peak time. For example, Figure 7 shows the host groups associated with port CL2-C.
After compiling the information for each port, we found that each port has two masking views, for a total of eight masking views. The hosts associated with these asking views had a simultaneous spike in throughput resulting in constrained front-end ports. A graph for one of the masking views is shown below in Figure 8.
Since these eight masking views simultaneously increased throughput during the same time period with roughly the same intensity, the workloads are most likely related, and the server owners should be contacted to see what could be done to mitigate spikes of this nature.
Restoring the sparkle to your Flash
This brief blog shows how even though you may have an all flash/hybrid flash array, sometimes the bottleneck can move elsewhere in the data stream. IntelliMagic Vision makes it easy to isolate these issues and create an action plan for resolution.
Subscribe to our Blogs
Subscribe to our newsletter and receive monthly updates about the latest industry news and high quality content, like webinars, blogs, white papers, and more.
HDS G1500/F1500 Series - Architecture and Performance Analysis
This webinar looks at the key physical and logical components that make up the architecture of the HDS VSP G1500/F1500 series and provides insights into the key related performance metrics.
Storage Performance Analysis for HDS VSP G1000/G1500/F1500 Series
This whitepaper provides a brief overview of the HDS VSP G1500/F1500 series hardware and discusses some best practices for configuring your HDS VSP for optimal performance within an actual SAN infrastructure.
NetApp C-Mode Architecture and Performance Analysis 101
This webinar looks at performance and capacity for your NetApp arrays running ONTAP Cluster-mode.