Lately, we have been engaged in situations where Disk Magic was used improperly. Disk Magic is an incredible tool, that is very precise and powerful at what it does. However, it is not uncommon for it to be misunderstood, and misused. When this happens, there is a high propensity for the tool to provide an invalid, inconclusive, or irrelevant results. Disk Magic may make the situation worse than before the analysis started by providing misleading information for objectives that it was never designed to address.
Let us first define what Disk Magic is:
Disk Magic is a pre-sales sizing tool that predicts average service times for specific points in time. It uses inputs from existing or theoretical workloads to validate that a proposed solution, when deployed according to best practice and properly balanced, is able to provide adequate performance.
From that definition, we can clearly see that Disk Magic is designed to be used as a pre-sales tool that will let you predict how a Disk System will behave when put against steady state workload profiles from specific points in time.
Typically, one chooses the Peak Transaction period, the Peak Throughput period and the Peak Write Throughput period and is trying to answer one or more of the following:
Compare different solution configurations for new hardware
Size IBM solutions to replace competitive environments
Combine multiple workloads on to new hardware
Show growth estimates for the new hardware
Also, keep in mind, that Disk Magic assumes a perfectly balanced system. Disk Magic will output full box averages with respect to service times and component utilizations from a specific interval that would be uniquely valid for the workload profile during that interval. It does not guarantee not to exceed maximum service times.
That takes us to what Disk Magic is NOT:
NOT a troubleshooting tool.
Disk Magic can not predict and model unbalanced environments
It can't predict implementation problems nor identify where don't follow best practices were not followed
It can't predict determine causes for performance problems
It can't debug performance issues
It can't expose issues around replication bandwidth or lack thereof
NOT a tool to predict absolute peak service times
NOT a tool to model workload profiles that have large bursts during the reviewed interval
NOT a tool that can provide insights around impacts of back-end activities such as Flashcopy or volume mirroring
NOT a tool that can provide a before and after look at a hardware remedy for a performance problem that it would have assumed away by design.
Hopefully, this will help you to identify when to use Disk Magic, and when to not use Disk Magic.
Starting with Disk Magic version 9.23.0 it is possible to select different skew levels for IBM i workloads. This can be a very useful feature when there is enough data to calculate the skew level of a workload. That can be done through the use of the STAT tool for the SVC, Storwize and DS8000 families. The STAT tool can be downloaded from: http://www-01.ibm.com/support/docview.wss?uid=ssg1S4000935
However, if you do not have a model that can calculate a skew level and you want to use Easy Tier in your solution, it is highly recommend to use the pre-defined Very Low skew level (2.00).
We are proud to announce that the Techline Disk Study team will present a Birds of a Feather session on Friday morning during the IBM Storage Masters events happening in Gaitherburg, MD during July and August 2015.
The session will focus on how to use Disk Magic, provide tips and hints on how to better leverage the tool, discuss use cases and provide an open discussion forum for Questions and Answer.
If you are attending the Masters, please stop by, say hi to Brannen (July session) and Andre (August session), and enjoy a good conversation focused on Disk Magic.
XiV Real-time Compression (RtC) was added to Disk Magic in version 9.22.0.
In recent communication to the Disk Magic community Joe Bacco has asked to avoid using Disk Magic to size XiV RtC solution. For the time being, if you need to size an XiV solution with RtC pleas contact XIV_RTC@il.ibm.com
Edit columns (hold ‘Alt’ key and use the mouse to select), and remove all traces of the r_await and w_await columns.
Note that any editor capable of removing columns will work. Notepad++ is used as just an example that can quickly remove these columns.
Load the modified file into Disk Magic
Don’t forget, if you want Disk Magic to understand the start time and date of samples, you can either add a couple lines to the top of the IOSTAT file as shown below, or override the actual date and time using the Disk Magic multi-file-open menus. See the Disk Magic help for the details.