This section provides guidance how to extract data from various Vendors and products (and the commands to run or tools to use) for analysis in WorkloadCentral and contains information on what metrics can and should be made available from various vendor storage arrays. Note that there might be changes based on array firmware releases and the various tool releases. It's important to note that WorkloadCentral Analysis Policies can be easily created for any vendor or product.
The WorkloadCentral Data Importer Format
WorkloadCentral does require a specific set of workload I/O metrics in order to be able to analyze, visualize and create workloads. Each row in the data needs to represent a distinct time sample, with all metrics represented as columns within the CSV file. The columns do not need to be in a specific order.
To simply the process WorkloadCentral provides multiple Analysis Policies for different vendor's data. When you upload data it will automatically attempt to recognize the format and recommend a set of policies that match the data format. An Analysis Policy is essentially a "conversion" template that converts the imported data into the specific set of metrics that WorkloadCentral requires to perform analysis. The metrics required to perform analysis include, but not limited to, Read / Write percentages, Read / Write IO sizes, and IOPS.
Analysis Policies are designed to be extensible and customizable. Therefore, you can either edit an Analysis Policy so that it can interpret the Input Data from the .csv, or you can edit the import .csv column labels to so that it can be interpreted by an available Analysis Policy.
As different storage arrays keep track of workload data differently, and produce .csv in different structures, the Analysis Policy also provides mechanisms to perform calculations on the data in the imported .csv to derive the metrics WorkloadCentral requires to perform analysis on the .csv. For example, if a storage array provides data on Throughput and IOPS, but does not provide data on average Block Size, then the Analysis Policy can derive the average Block Size by dividing Throughput by IOPS.
As an example, here is a truncated screenshot of an imported .csv from a storage array:
For more information on how to get workload data for various leading vendors see the following:
- Virtual Instruments® Fibre Channel Probes
- Virtual Instruments® Software Probe (for VMWare® vSphere)
- HP Enterprise®