This preview shows page 1. Sign up to view the full content.
Unformatted text preview: e sharing, or secure ftp to copy files from machine to machine as we needed.
The testing procedure consisted of the following steps:
1. Execute batch files and shell scripts to clean up prior outputs on client virtual machines, the server
physical machine, and the controller.
2. Drop the database on each server VM.
3. Restore the database on each server VM.
4. Pause for 5 minutes to wait for background tasks to complete before server reboot.
5. Reboot the physical machines, and wait for a ping response from the physical machine, as well as from
all VMs involved in testing.
6. Wait 10 additional minutes for any background tasks to complete.
7. Record idle power for 2 minutes.
8. Start performance counters on client virtual machines, and start esxtop on the server under test.
9. Start the workload connections.
10. Start the workload ramp up period. Run the DBCC FREEPROCACHE command if necessary (see the
Creating the database section).
11. Start the workload.
12. Stop the workload.
13. Copy all output files to the controller. Power measurement procedure
To record each server’s power consumption during each test, we used an Extech Instruments (www.extech.com)
380803 Power Analyzer/Datalogger. We connected the power cord from the server under test to the Power
Analyzer’s output load power outlet. We then plugged the power cord from the Power Analyzer’s input voltage
connection into a power outlet. For the storage arrays, which have multiple power cords, we plugged all of the
cords for each array into a power strip and then plugged the power strip into the Power Analyzer’s output load
We used the Power Analyzer’s Data Acquisition Software (version 2.11) to capture all recordings. We installed the
software on a separate Intel processor-based PC, which we connected to the Power Analyzer via an RS-232
cable. We captured power consumption at one-second intervals.
To gauge the idle power usage, we recorded the power usage for two minutes while each system was running the
appropriate number of VMs, but otherwise idle, meaning they were not running any test workload.
We then recorded the power usage (in watts) for each system during the testing at one-second intervals. To
compute the average power usage, we averaged the power usage during the time the system was producing its
peak performance results. Because we measured the power of the server and storage separately, we added the
average power of both together to get the total power. We call this time the power measurement interval. See
Figure 6 (idle and average peak power) for the results of these measurements. Principled Technologies, Inc.: Virtualized OLTP workload performance comparison of end-to-end solutions:
Dell PowerEdge R710 with Dell EqualLogic storage vs. HP ProLiant DL385 G5 with HP StorageWorks EVA
4400 storage 17 Appendix A – Server configuration information
This appendix provides detailed configuration information about the servers, which we present in alphabetical
order. Figure 9 provides detailed system configuration information.
VMware Build 148592 caveat: All information in this disclosure regarding future direc...
View Full Document
- Fall '11