Data Mover Cache Test

Data movers (also known as X-Blades) contain the defined file system. They access data from the back end and provide host access using the same UltraFlex I/O Technology that supports the NFS, CFS, MPFS, and pNFS protocols. Each data mover has a large cache, which serves NFS I/O requests and reduces the I/O going into the storage processor. If users to VNX complain that their NFS requests take time being serviced, you may want to check cache usage to see if it has been used effectively or not. The Data Mover Cache test can be used to perform this periodic cache usage check. For each data mover in the DME of a VNX system, this test monitors the directory name lookup, openfile, and buffer caches, and reports the percentage of requests that have been serviced by each of these caches. Deficiencies in cache usage are thus revealed, thereby enabling administrators to assess its impact on storage system performance and to investigate the reasons for the bad cache usage.

Target of the test : An EMC VNX Unified Storage system

Agent deploying the test : A remote agent

Outputs of the test : One set of results for each data mover in the DME of the EMC VNX Unified Storage system.

Configurable parameters for the test

Parameter	Description
Test Period	How often should the test be executed.
Host	The IP address of the storage device for which this test is to be configured.
Controller Station IP	The Control Station is the management station for the VNX for File system, and enables control and configuration of the system. The eG agent uses the CLI that runs on the Control Station to monitor and manage the performance of the VNX for File system. To enable the eG agent to use this CLI, specify the IP address of the Control Station in the Controller Station IP text box. By default, the IP address of the host will be displayed here.

Measurements made by the test

Measurement	Description	Measurement Unit	Interpretation
DNLC hit ratio	Indicates the percentage of time the requested directory pathname was readily available in the Directory Name Lookup Cache (DNLC) without requiring to perform disk I/O on this data mover.	Percent	A high value of this measure (close to 80%) indicates that the Directory Name Lookup Cache is being utilized well. A smaller value is a cause for concern, as it not only indicates poor cache usage, but also implies that direct disk accesses are more. One of the key reasons for a low value for this measure is an improperly sized directory name lookup cache. If this is the case, then, you may want to increase the DNLC size to improve cache usage and reduce the I/O going straight to the disk.
Open file hit ratio	Indicates the percentage of time the requested file is readily available in the open file cache without requiring to perform disk I/O on this data mover.	Percent	A high value of this measure (close to 80%) indicates that the open file cache is being utilized well. A smaller value is a cause for concern, as it not only indicates poor cache usage, but also implies that direct disk accesses are more. One of the key reasons for a low value for this measure is an improperly sized open file cache. If this is the case, then, you may want to increase the size of this cache to improve cache usage and reduce the I/O going straight to the disk.
Kernel buffer hit ratio	Indicates the percentage of requested pages that were readily available in the buffer memory of this data mover without performing disk I/O.	Percent	A high value of this measure (close to 80%) indicates that the kernel buffer cache is being utilized well. A smaller value is a cause for concern, as it not only indicates poor cache usage, but also implies that direct disk accesses are more. One of the key reasons for a low value for this measure is an improperly sized open file cache. If this is the case, then, you may want to increase the size of this cache to improve cache usage and reduce the I/O going straight to the disk.