As of March 14th, this thread remains open, but the IX2 giveaway is closed (30 given away). Thanks, and please continue to provide data!!!
This is a discussion thread where customer who have used the following freely provided, as-is, powershell scripts to gather their vSCSIstats and ESXtop (and optionally their EMC back-end storage) performance data can share their results.
While these will be there for the overall community to use (as this forum is completely public), within EMC this data will be used to further improve our storage platforms, and ensure we are testing against "customer like" scenarioes.
Any customer who posts their ESXtop and vSCSIstats (from their "heavy hitting VMs") performance data (if you are an EMC customer, the EMC array stats are bonus), I will provide an iomega IX2 to as a "thank you"
Please post your environment info and the output from the scripts (can be large, but is VERY compressible, please zip) on this thread.
As you post your results by continuing the discussion thread, please note some details about your environment:
The scripts to simplify gathering your data are below:
As always thank you for being VMware and EMC customers!
The data is zipped using 7-Zip 9.2 using one of the new algorithms that can use more than 2 concurrent cores for compression.
Attached the output of the esxtop and vscsistats script and get_unified_nas_perf with a interval of 300.
I hop this is of any value.
Added the output of get_unified_nas_perf in csv format.
Maybe its also worth to mention that all datastoreses are created with the VSI Unified Storage Management vCenter plug-in
I was able to get the esxtop info (I hope I did it right). I'm still working on the vscsiStats. I keep getting a message "not running anymore so quit". I just got it running. I'm not sure why but when I run it on my laptop, it works. Both are Win 7 x64... hmmm
Anyway, here are the vscsiStats too. I hope this can help, and if I did anything wrong and you need new data, just say so. Thanks.
I will attach the files as they become available. I am selecting a sample of the entire environment, and will include our heaviest I/O application (SAS) for the scsistats.
See my attached doc entitiled Environment for more details on setup and timing of data collection.
Couldn't get the ESXTop script to run so I just ran it in batch mode for 4 hours, 30 second intervals.
Data attached in CSV files.
The vscsistats script doesn't seem to work with ESX 4.0, I made some edits so the script would at least run and collect some data. The target was an Autonomy IDOL indexer that appears to be working on a reindex today.
vSCSStats is still running, I will upload shortly. IO workload is SQL 2008 R2 production servers.
Windows 2000, 2003 and 2008 og some linux servers. Theres Exchange 2003, sql2000, 2005 and 2008 servers. Emailxtender indexserver, vcenter servers, windows AD/DNS, and a lot of web and application servers
Remove the first run and added a weeks data from collect_esxtop
Attatched unifited_perf CX3-40 Primary Production SAN being migrated to CX4-120, CX3-40f is just a replica of CX3-40, using RPA for replication.
1. how many ESX/ESXi hosts?
4 ESXi Hosts
3 (esx1-3) - HP DL380 G5, Dual Quad Core 2.66 GHz Intel Xeon x5355, 32 GB RAM
1 (esx4) - HP DL360 G7, Deual Quad Core 2.66 GHz Intel Xeon x5640, 64GB RAM
2. how many VMs?
58 VMs - Mix of Windows 2003, 2008, and Ubuntu
vCenter is also running as a Guest VM
3. what vSphere version?
ESXi 4.1.0, 260247
4. VMFS, NFS, Both?
VMFS for all production VMs. NFS storage for testing and iso/image storage.
5. What type of array backs the datastores?
Celerra NS20 - Fiber Channel and SATA Shelves.
Most data storage on SATA Drives with most (not all) OS partitions installed on FC
The NFS Server is built on HP DL380G5 running Ubuntu
The esxtop stats were collected over a 24 hour period.
The vscsistats were collected for all VMs over about an 8 hour period, I started out collecting every 60 seconds but had issues and switched to 300 seconds. The results are cumulative over that 8 hour period. Not sure if I collected these correctly, just let me know if other information is needed.