Power management and read performance question.

[drg]

Hello, first sorry for my bad English, it is not my first language, and if I missed some similar topic on this forum when searching for this info. I've heard about this software a long time ago and finally decided to test it after I had some issues with a raid6 array failing HARD in the middle of a rebuild and heard that it could speed up read performance in some cases.

 

I currently have 1x SSD (OS), 3x 500gb HDs  running in RAID0 and 2x 4TB in RAID1 for data. My use is around 80% read 20% write and most stuff is on the RAID0 array with a daily backup to the RAID1. The few things that I want to avoid to use a backup in case of failure is in the RAID1 array and the sparse access to those files is rapidly wasting the "load/unload cycle count" of those disks (the reason I lost the previous RAID6 array).

 

From what I could understand from the manual/forum posts, the program is able to read the same file in parallel or many files for better IOPS from the duplicates right? Does it gets somewhere close to RAID0 in terms of MB/s reading a single file? And how is random IO?

Assuming I don't set the pool to duplicate my Steam folder, it would it spread all files across the disks right? Could get some better random IO if that is the case?

 

If I put those 4TB on their own pool with the asynchronous duplication, would one of the drives stay shutdown until the sync is done or would it be accessed before that?

 

Last question, do I lose anything else than shadow copy by using the pool? Could I still use my defrag program on the disks?

 

[Christopher (Drashna)]

Since you're asking about the Read Striping feature, let me link that first:

http://stablebit.com/Support/DrivePool/2.X/Manual?Section=Performance%20Options

 

 

However, no you will not see a performance boost on par with a RAID array. That is because of how the RAID controller works, vs how DrivePool works.  Specifically, because RAID is a block based solution, it's raw access speed will be faster because it's not querying the file system first. It just access the sector on the disks, directly. 

 

StableBit DrivePool operates on a file level. That means it uses Windows' standard method of accessing files, and caching them.

But because of how the Read Striping works, it will tend to get better speeds than just one disk or the other, when the files are duplicated.

 

As for the random IO, that really depends on the system, how the files are placed, and the disks. However, as DrivePool "basically" is a proxy for file system requests, it should handle random IO better.

Using the Steam folder example, if the files are placed on different drives, then when you access the bunch of files, it will be accessing different disks, getting their normal speed. Because it is distributed between multiple disks, you should definitely see better performance than if it was on a single, normal disk. And depending on how many disks are in the pool, and how spread out the files are... you may get better performance than the RAID array. Emphasis on "may".

 

 

As for the asynchronous duplication, I'm assuming you mean when you disable the real-time duplication. If that is the case, then only one copy of the file is written. Then, later (at 2am by default), it will run a duplication pass, and checking the status of all the files in the pool, and duplicating them as needed.

 

However, DrivePool writes new files to the disk with the most free space by default. That means, if your disks are all about the same size and have about the same content on them, it will basically write each file to a random disk. This will keep them up. This happens regardless of the duplication status (1x, 2x, 3x, etc) or duplicating time (real-time vs later). Though realtime duplication would write to the two disks with the most free space, instead of the single disk with the most free space.

 

 

 

And for the easy question, VSS (shadow copies) is about the only thing that you will lose by using DrivePool.

Defragmentation works just fine, as long as it's done only to the disks that make up the pool.  And actually, I use PerfectDisk on my system, without any issues.

[drg]

Thank you for the answers.

I am testing it on the 3x 500GB disks for now, not raid0 tier performance but it is very close on some tests I did, and even better in random IO for 2x/3x duplicates VMs. Going to see how well it will do during the trial.

[Christopher (Drashna)]

Well, I'm glad to hear that you're getting good results, then.

 

If you have any other questions, don't hesitate to ask.