Measuring Disk I/O Performance WIndows Server

I am trying to monitor some disk activity on a hyper v host machine. I see many articles that tell me the range is 1-15ms is healthy 15-25ms is warning and 26ms or more is bad.
So I set the counter as logical disk \ avg sec read and write.
So my question is what do you set the scale to. Microsoft defaults it to 1000 so I am guessing on the graph if I see 20 that is 20ms? There are so many scales you can set it to I want to make sure it is the right one. I have read a lot of people say the defaults are wrong.
Also is this the best scale to look at for disk activity? I am basing the above from a TechNet article. Would I monitor the system disk like C? or the actual disk like V that is reading off the SAN where the VMs live? Thanks!
Who is Participating?

[Product update] Infrastructure Analysis Tool is now available with Business Accounts.Learn More

I wear a lot of hats...

"The solutions and answers provided on Experts Exchange have been extremely helpful to me over the last few years. I wear a lot of hats - Developer, Database Administrator, Help Desk, etc., so I know a lot of things but not a lot about one thing. Experts Exchange gives me answers from people who do know a lot about one thing, in a easy to use platform." -Todd S.

Philip ElderTechnical Architect - HA/Compute/StorageCommented:
A telltale indicator that there is a disk I/O bottleneck would be the Disk Queue Length in Performance Monitor (via Task Manager).

Here is a blog post that gets in-depth on disk latency: TechNet: Measuring Disk Latency with Windows Performance Monitor (Perfmon).
Robin CMSenior Security and Infrastructure EngineerCommented:
You're right about the graph showing in ms if the scale is set to 1000, the clue is in the name of the counter: Average disk seconds/read (or write). Divide seconds by 1000 and you get milliseconds.
That is telling you how many seconds each read or write IO is taking. What's its not telling you is the size of the IOs, which could range from 512B to potentially 1MB. You can sort of see the sizes from the Average Disk Bytes/Read and Write counters, but as the name implies, they're averages. Perfmon is not a very easy to use tool. To see the actual size of each IO you can use SysInternals DiskMon.

As to which drive to monitor, that's for you to decide based on where the data lives that you want to monitor the disk performance of. If the data is on a SAN drive then there's no point in monitoring the local C drive.

You might find that the built in Windows Resource Monitor gives you nicer data than perfmon, depending on what you're looking for. Load it by running resmon.exe or via the performance tab of Task Manager. If you have access to your SAN management console you will also find that should be able to give you useful info too (though maybe not in a very friendly format).

My advice, dump the SAN, buy a Tintri (or several) and forget having to worry about storage performance or management ever again. I did, and the sooner I can switch the last of the old EMC kit off the better. If virtually all your IO is sub 1ms latency who cares about monitoring and configuring individual LUNs and VMs?

Experts Exchange Solution brought to you by

Your issues matter to us.

Facing a tech roadblock? Get the help and guidance you need from experienced professionals who care. Ask your question anytime, anywhere, with no hassle.

Start your 7-day free trial
valmaticAuthor Commented:
So robincm, what do you use Trinity for? Do they host your vms?
Simplify Active Directory Administration

Administration of Active Directory does not have to be hard.  Too often what should be a simple task is made more difficult than it needs to be.The solution?  Hyena from SystemTools Software.  With ease-of-use as well as powerful importing and bulk updating capabilities.

Robin CMSenior Security and Infrastructure EngineerCommented:
Tintri VMstore is a VM-aware "SSD and HDD" storage device. All writes go to SSD, and SSD is used to cache reads from HDD. It uses the SSD very efficiently such that you can expect >90% (usually >99%) of IO to be from SSD and at <1ms latency.

You can see performance stats for individual VMs, set QoS for individual VMs, or just leave it alone and forget you ever had any storage performance issues.
Philip ElderTechnical Architect - HA/Compute/StorageCommented:
We deploy Scale-Out File Server (SOFS) clusters for our larger storage backend needs. Storage Spaces provides the needed disk access and arbitration, can utilize a SSD tier, or tiers, and is scalable and resilient.

Our setup here in the shop hit 377K IOPS on one all flash JBOD. We have a configuration that allows for 1M IOPS out of that JBOD setup. When configured in a resilient disk array we can move 300K IOPS easily.

Cost per GB can't be beat by any storage vendor today.
Gerald ConnollyCommented:
We normally measure Disk Performance in two major ways

MegaBytes per Second - MB/s
IO's per second - IOPS

and then associated performance indicators of Response time and Queue length

A conventional Enterprise Spinning disk spindle will do 200MB/s or 200IOPS whereas the equivalent SSD device will do at least 500MB/s and 30K+ IOPS
It's more than this solution.Get answers and train to solve all your tech problems - anytime, anywhere.Try it for free Edge Out The Competitionfor your dream job with proven skills and certifications.Get started today Stand Outas the employee with proven skills.Start learning today for free Move Your Career Forwardwith certification training in the latest technologies.Start your trial today
Windows Server 2012

From novice to tech pro — start learning today.