locked
Better redundancy for storage. RRS feed

  • Question

  • While the core features of WHS are good, the redundancy aspect of the storage architecture leaves a bit to be desired, and no, folder duplication is not the answer.

    It would be nice to see some ZFS like features in WHS where you have blcok level redundancy and auto scrubbing (background process), for constantly monitoring blocks/sectors on hard drives.

    If not, WHS should have a way to designate "certain" drives for storage and not others, and this has to include RAID arrays. Along the same lines, it would be nice to give the drives a priority where WHS handles drives in terms of reading and writing based on that priority. If the priority is not possible, then a way to just designate certain drives as "not in storage pool" (especially for the system drive) would be good.

    Sunday, February 25, 2007 11:57 PM

All replies

  • Hmm, I don't know. This seems too much like a 'pro' backup solution and not one for home users. I don't see your average family setting up RAID arrays or designating and prioritizing drives for storage. But I may be missing the point of your post.

    Why, by the way, is folder duplication not the answer?

    Monday, February 26, 2007 10:52 AM

  •  kapone wrote:

    While the core features of WHS are good, the redundancy aspect of the storage architecture leaves a bit to be desired, and no, folder duplication is not the answer.

    It would be nice to see some ZFS like features in WHS where you have blcok level redundancy and auto scrubbing (background process), for constantly monitoring blocks/sectors on hard drives.

    If not, WHS should have a way to designate "certain" drives for storage and not others, and this has to include RAID arrays. Along the same lines, it would be nice to give the drives a priority where WHS handles drives in terms of reading and writing based on that priority. If the priority is not possible, then a way to just designate certain drives as "not in storage pool" (especially for the system drive) would be good.




    I really have to agree with this post.  I realize the reply of it possibly being a pro feature, but my guess is that these boxes are going to come pre-built by the novices, and they won't really be touching it, but for people that buy the software themselves, are going to be pretty tech savvy.  So I think in the long run it would be a good addition.
    Monday, February 26, 2007 12:32 PM
  •  kapone wrote:

    While the core features of WHS are good, the redundancy aspect of the storage architecture leaves a bit to be desired, and no, folder duplication is not the answer.

    It would be nice to see some ZFS like features in WHS where you have blcok level redundancy and auto scrubbing (background process), for constantly monitoring blocks/sectors on hard drives.

    If not, WHS should have a way to designate "certain" drives for storage and not others, and this has to include RAID arrays. Along the same lines, it would be nice to give the drives a priority where WHS handles drives in terms of reading and writing based on that priority. If the priority is not possible, then a way to just designate certain drives as "not in storage pool" (especially for the system drive) would be good.


    For the system drive, I can see absolutely no reason why one would not want the entire drive in the storage pool. The size of the system drive has an enormous impact on functionality right now (go look at the threads from people who started with a 40 GB system drive), and I don't expect that impact to go away entirely with the RTM version. Marking some of the space on the system drive as unavailable would just exacerbate that problem.

    And as for designating a drive as "not in storage pool" that's easy to do. Just don't add it to the pool. The drive won't be usable from anywhere but the WHS desktop, but it won't be in the pool.

    I agree that a more robust data protection feature might be nice, though.
    Monday, February 26, 2007 1:26 PM
    Moderator
  •  Ken Warren wrote:
     kapone wrote:

    While the core features of WHS are good, the redundancy aspect of the storage architecture leaves a bit to be desired, and no, folder duplication is not the answer.

    It would be nice to see some ZFS like features in WHS where you have blcok level redundancy and auto scrubbing (background process), for constantly monitoring blocks/sectors on hard drives.

    If not, WHS should have a way to designate "certain" drives for storage and not others, and this has to include RAID arrays. Along the same lines, it would be nice to give the drives a priority where WHS handles drives in terms of reading and writing based on that priority. If the priority is not possible, then a way to just designate certain drives as "not in storage pool" (especially for the system drive) would be good.


    For the system drive, I can see absolutely no reason why one would not want the entire drive in the storage pool. The size of the system drive has an enormous impact on functionality right now (go look at the threads from people who started with a 40 GB system drive), and I don't expect that impact to go away entirely with the RTM version. Marking some of the space on the system drive as unavailable would just exacerbate that problem.

    And as for designating a drive as "not in storage pool" that's easy to do. Just don't add it to the pool. The drive won't be usable from anywhere but the WHS desktop, but it won't be in the pool.

    I agree that a more robust data protection feature might be nice, though.

    Here's why the size of the primary hard drive matters. I suspect the pagefile is on that drive. I suspect the "lazy write" feature of WHS uses space on the primary drive, while it's waiting to write to to the secondary drives. Since it uses a 10GB partition for the O/S itself, a typical page file would be around 2-3GB, with a 40GB HD you don't have a whole lot left to play with. Hence the performance drop.

    I have a Tyan motherboard based  server with dual 3.2GHZ Xeons, 2GB RAM, and 10 250GB SATA II drives on the server (connected via Sil 3124 HBA and Sil 3726 port multipliers) for data, and an 80GB drive as the primary. WHS transfers large files to these drives at ~35MBps sustained over gigabit LAN, and FROM these drives to a client PC at almost 43MBps. So, I'm not really seeing a performance issue.

    However, the point of my post was that file systems have moved beyond mere mirroring for redundancy. There's various types of RAID available now, which goes beyond just RAID 5. WHS doesn't even need to do RAID 5 (or some other flavor), what it needs to implement is some sort of parity check for data and mirror that parity information across all drives. In this scenario if a drive fails, you remove it from the configuration, with no effect (since the data can be reconstructed from the parity information). WHS could easily just reconstruct the data and write it to one of the other drives, it does'nt even have to wait for a new drive to be installed. And this "reconstructing" could certainly eb done in the background, and yes, the performance will be degraded a bit while it's doing it, but would be more robust AND cost effective than simple mirroring..aka folder duplication.

    Monday, February 26, 2007 3:58 PM