How to tell drive queue depth?

[smkout]

Dear forum members, I was wondering if there is a way to deterimine 'I/O drive queue depth' in Windows Vista? Such as in task manager or such.
The question is arising from reading the article about the OCZ IBIS drive, where the verdict was that unless your are dealing with a queue depth of 5 or more, you are not going to see an improvement.
Within Task Manager>Process, I can view I/O reads and writes, but these are showing numbers in the hundreds of thousands, so I'm figuring that's not the same thing.
Just have a Velociraptor at the moment, but am looking for some sort of reference of what I'm dealing with my current applications. I use various CAD programs and saving of files can take considerable time on larger projects.
Any suggestions or insight would be appreciated (the simplier the better)
Thank you.

 

[Emulex]

good rule of thumb in perfmon is to never sustain a queue of 1.0 or more. only time i've seen that happen is when the cache disabled (not by choice, by failure) on raid set.

 

[smkout]

Thank you Emulex for the reply.
So by the sounds of it you been able to observe queue depth in perfom. Can you please let me know where/how, as I am just getting familar with it?
Thank you again!

 

[Emulex]

i just use perfmon, it can analyze so many aspects of your system (especially sql server) that you can learn it and understand the bottlenecks - far too complex depending on your goals to explain in an email (google it). There are apps that just sit around and read the data and accumulate information and report to you - some companies have appliances that can watch your whole network/systems and give assessments. i suspect you are looking for something more finite as a bottleneck?

i usually use the vmware san thread for performance measuring but that's heavy on the server side - it can crush the biggest san arrays (and make them fail) if not properly configured. I've also seen people raid two x25-M 160gb drives and run the same benchmarks - in raid-0 to show numbers (not reliability) that slaughters $100K san storage in iops and linear throughput. i'll dig up the thread. it's pretty odd to see a $100K san behave so much more poorly than a small server with direct attach drives but of course that might be a 5 year old 100K san or a misconfigured one or the flexibility (synchronous replication across 100km) may be the reason. the simple benchmarks don't hold a candle to these benchmarks but i can appreciate a simple report sometimes. to me it is important to break it down, block size, queue depth, % cpu loading, random performance (read/write), linear performance (peak read/write), combinations (60/40), etc. the amount of data laid down on an excel spreadsheet allows you to really see the minute effects of disk alignment, or the system build that accomplishes the goal.

CAD would be more larger linear files and assuming you have sufficient ram (or not) paging by the application or operating system. I don't know much about autocad other than to use a freaking certified graphics cards (no consumer junk) and gobs of ram and video ram. I just assumed if you throw enough at it the load/unload was mostly linear read/writes

 

[smkout]

Thank you again for the response Emulex.
I was hoping that there was a way of viewing a column or such that would just state the queue depth, but I guess I will have to investage the help section (and Google) of perfmon a bit. Sure to find lots of items I never knew about along the way.
The CAD program I mostly use, is not autodesk's Autocad, although I use autocad for importing and exporting. The program I use (autosprinkvr), is a computer aided drafting program, specifically for the fire sprinkler industry. It does not require the use of the Quadro cards (thank God). However since it uses a different architecture of programming to ACAD, it is difficult to compare hardware review to how it will perform with this program in particular, and with multitasking while using this program, as it works the system differently to ACAD.
There are a few different 'things' going on when I save a file with this program. There is a usb security dongle, that the program looks for to enable a save to be performed (which differenciates from the demo version), it also reads and writes to the database (information on products used in system, etc) as well as saving the file. This seems to create a 'random' reading/writing situation as when I watch the disk activity in task manager I see various files exchanging 'the top spot' in the disk read/write columns of task manager. And seems one has to be verified before proceeding with the next 'stage of the save'.
Anyways, just figured I'd throw it out there is case it was one of those easy fix things. I'll do a bit of digging in perfmon. (This option seems to provide more options than the performance monitor accessed through task manager, which is what I had previous been using, so thanks for that tip!)

 

[Emulex]

copy protection lol. it's hard to tell - photoshop uses(used) it's own swap file. hell they could use the usb drive for that - if its an older program that written in the XP days it may not be designed for win7.

fire up ida pro and remove the protection then you could see if that is the hangup . usb eats MONSTROUS amounts of cpu in some cases. I'm guessing a security dongle may not be as fast as a modern high speed usb flash.

Quadro cards are the ECC of video cards - you don't really want to have errors designing a building or something - think about how they cherry pick the best intel chips and make them "special" - likewise nvidia did the same - you can't have your video card doing 128bit math and throwing an error or bad results - but in a game you'd never notice since the video card can restart in the event of fault.

 

[smkout]

yeah, I dunno about the disabling the protection, would be a conciderable savings but the support, etc is a large part of the yearly lease of the program ($3,600).
I dunno if they will maybe change as USB 3.0 becomes more the norm.
I can't imagine the actual reading from the doggle being such a 'large read' though the jumping back and forward in the Disk section of Resource Manager, is indicating to me that the jumping back and forward of the various files is 'beating up' the hard drive.
Some of the numbers I'm seeing in the disk activity for various files seems scary (in the 'Write B/min' column I am seeing numbers approaching 200,000,000).
Up until recently I had been dealing with file sizes of 5-10MB. With the BIM (Building Information Modeling, Revit and Bentley products) becoming more the norm, the file sizes are going to explode in the upcoming years.
Working for a smaller firm in difficult economic times. Upgrades need to be worthwhile and keep to minimum. Apart from the cost of parts, there is some time loss when upgrading a designer's computer, and the IT guy is also a cost. I've therefore been holding off upgrading to the SSDs. Had been more because of concern of reliability (other than the intel but wasn't going with the 100MBs write speed for a $400,160GB drive). Seems now though that the Sandforce are able to provide speed (random and seq), and reliability, so thinking about pulling the trigger, and want it to be like the deer hunter - 'one shot'
Given the improvement when these things are RAID'd together, I am interested in that option. However the lack of TRIM suport in RAID is a killer for me. Computer downtime and troubleshooting in the Design department would negate the saving from the improvement in performance of the new drive. It was originally the Revo drive I was looking at, but the garbage collection doesn't seem to be effective, whereas with the IBIS drives, the garbage collection seems to be very effective. The IBIS are conciderably more expensive, however if I was going to get the benifit of it, for a long period, given that it's how I put food on the table, it is not 'unattainable'. However, given Anand's review indicated that only a certain segment will see the benifit, I'm not looking to just throw money away...Which is what was getting me around to looking for an indication of my current queue depth. Some research in my future.
Yeah those Quadros seem to be the ferrai of video cards. It seems Nvidia makes these (as apposed to EVGA, BFG...) and also provide the support and custom drivers. Those drivers are a key part. With my GTX 260 (waiting for Kempler unless I get alot more 3D work before that), will be fine in ACAD until I start trying to move around in a full shaded render mode, and then it hardly works. With Quadro cards, there seems to be some juice (code) in there that allow users to move around almost the same as wireframe. My program, can render scences in 3D with the GTX way easier that ACAD, though ACAD can produce much more photo relistic rendering and ACAD can save the file in 5-10 seconds, where I can be waiting nearly a minute on autosprink.
Anyways Emulex, didn't mean for this to be so long, but figured I'd paint the picture in case it brought up any insight/thought, etc.
Thanks again for the input

 

[tweakboy]

You guys should look into. Winbootinfo 30 day trial . Also there is a process manager free app that tells you more stuff.

google her in. Thanks and sanks, gg and gb

 

[smkout]

Thanks for the suggestion Tweakboy. Downloaded it, and it is interesting, but unless I'm mistaken Winbootinfo only provides information on the boot up. I'm really trying to find my queue depth when I'm saving/opening files on this autosprink program.
Interesting that it has my boot time @ 191.1 seconds, and the cpu idle time as 174.5 secs. (I guess this is due to the CPU waiting on the hard drive?)
Have you been able to use this program to monitor disk I/O queue depth or other monitoring outside of the boot process?

 

[mtnd3vil]

Not 100% sure vista has QDepth in resource monitor, but 7 does.

 

[tweakboy]

Good post mtn3vil ,, You sound like a virus name LOL. Theres free apps that do it as well,,, juts fyi,,,thanks gg and gb

 

[smkout]

Thanks for that insight mtnd3vil! I'm going to go ahead and get Windows 7. Was going to be getting it fairly soon anyway, so this is another reason to do sooner rather than latter.