eG Monitoring
 

Measures reported by XIOClusPerfTest

To external applications, XtremIO appears and behaves like a standard block storage array. However, due to its unique architecture, it takes a fundamentally different approach to internal data organization. Instead of using logical addresses, XtremIO uses data blocks internally to store data. According to the block content, XtremIO decides where exactly to place the data blocks.

In a write operation, any data chunks that are larger than the native block size are broken down into standard blocks when they first enter the storage array. The system calculates a unique fingerprint for each of the incoming data blocks, using a special mathematical algorithm. This unique fingerprint is used for two primary purposes:

  • To determine where the data block is placed within the array

  • Inline Data Reduction

Because of the way the fingerprinting algorithm works, the ID numbers appear completely random and are evenly distributed over the possible range of fingerprint values. This results in an even distribution of data blocks across the entire cluster and all SSDs within the array. In other words, with XtremIO it is neither necessary to check the space utilization levels on different SSDs, nor to actively manage equal data writes to every SSD. XtremIO inherently provides even distribution of data by placing the blocks based on their unique IDs. To achieve peak performance of any storage array, administrators should constantly keep a check on the read/write operations on the clusters, disks, SSDs of the storage array. The shorter the time taken to perform the read/write operations, the better is the performance of the storage array. In order to improve the performance of the storage array and figure out any processing bottlenecks in the storage array, it is necessary for the administrator to carefully monitor the processing capability of each cluster in the EMC XtremIO with respect to the size of the data blocks. The XIOClusPerfTest test helps administrators in this regard!

This test reports the I/O processing capability of each cluster with respect to the data block size. For each data block size, this test reports how well the data block is processed, the time taken to process each data block size etc. Using this test, administrators can identify processing bottlenecks and initiate pre-emptive actions.

Outputs of the test : One set of results for each cluster:data block on the EMC XtremIO Storage array

The measures made by this test are as follows:

Measurement Description Measurement Unit Interpretation
readBandwidth Indicates the rate at which data blocks of this size were read from this cluster. MB/Sec  
writeBandwidth Indicates the rate at which data blocks of this size were written to this cluster. MB/Sec  
readLatency Indicates the time taken to read data blocks of this size. msecs Ideally, this value should be low. A high value could indicate that read/write operations are slowing down for some reason.
writeLatency Indicates the time taken to write data blocks of this size. msecs
avgLatency Indicates the average time taken to read/write data blocks of this size. msecs A value close to 100 indicates that the SSDs in the cluster are running out of space.
readIOPS Indicates the rate of read I/O operations measured for the data blocks of this size during the last measurement period. IOPS  
writeIOPS Indicates the rate of write I/O operations measured for the data blocks of this size during the last measurement period. IOPS