Storage controller based duplication

[browned]

Just a quick feature request, unless someone can explain how to do this with the current version.

 

I would like to place duplicated files on different storage controllers. I have 2 x HP Smart Array p410's with about the same amount of storage on each controller.

 

- from my perspective write performance for files going to different controllers should be faster than a file being duplicated onto the same controller.
- if a controller fails I have a cope of all of the data and nothing stops working.
- read performance could also be better as the file will be coming from two different paths, with two sets of controllers, cache involved.

 

The only way I can think of doing this is the create a pool of drives for each controller, then mirror them with either Windows or a third party tool or even create a nest drive pool out of the two other pools.

 

Thanks

 

[Christopher (Drashna)]

You *could* use File Placement rules to micromanage this, but that's not really a good solution.  

And neither is creating two pools and using some other software to sync the files over, in this case.

 

 

However, we do have a feature request for "duplication grouping" (in a large part because of StableBit CloudDrive).  However, this feature would essentially do exactly what you want (set one controller as one group, and the other controller in another group). 

 

 

As for the controllers, even just one should be enough for you. These cards are PCI-Express 2.0 8x (8 lane) cards, and are capable of up to 4GB/s throughput (well, that's what the controller is capable of). 

And you should be able get 20 drives running at 150MB/s, at one time.  But considering that you will likely never see that many drives saturated at the same time, there shouldn't be an issue for the controller's performance. Let alone the need for two.

 

As for the controller failing, if these drives are just being passed through, there shouldn't be any issues regardless.

However, if they are in arrays, then replacing the controller is all you'd need to do. The new controller should pick up the configuration automatically, I believe, or at worst, require you to re-setup the arrays and preserve the data. 

 

But I do understand your concerns here, especially with the failed controller.  

And as I said, the "grouping" feature has already been requested and is defnitely on our to-do list already. 

[browned]

I don't use the raid features on these cards as the mix of hard drives I have doesn't allow it. I do have about 7 drives attached to each controller in varying  sizes. Each drive is configured as a raid 0 array. The cache is set to read only and is purely there to allow for the amount of raid 0 arrays I have. The smart arrays can not do JBOD or pass through.

 

If a drive fails or is showing signs of failure I can pull it and plug it in via USB. It doesn't have any special raid configuration or disk format so I can get at the data easily.

 

I had a cache module fail the other week which took the whole controller offline until I replaced it. Removing the faulty cache disabled the arrays and therefore disk access. This is something I hadn't foreseen but luckily I had a spare cache module. I now have two more controllers on the way with cache and batteries attached.

 

I suppose my position is to be resilient as possible.

 

- If I drop a disk, replace it.

- if I drop a controller or the cache module, replace it but have downtime while I wait for parts.

 

If a feature is developed it would be great as the controller being offline will not stop access to any file.

 

Thanks

[Christopher (Drashna)]

Ah, okay. I can understand the request then. 

 

And it sounds like the P410's don't allow you to pass the disks through to the OS directly (eg act as a pure HBA card). 

 

If you wanted, you could replace the cards with LSI 9240-8i or 9207-8i cards, that may or may not be cheaper, but are pure HBA cards, and won't have this issue.