Comments on: Write-back cache: Battery vs Disk http://storageadvisors.adaptec.com/2007/06/28/write-back-cache-battery-vs-disk/ Storage Solutions for Real World IT Professionals Sat, 05 Jul 2008 07:56:49 +0000 http://wordpress.org/?v=1.5.2 Comment on Write-back cache: Battery vs Disk by: SJS http://storageadvisors.adaptec.com/2007/06/28/write-back-cache-battery-vs-disk/#comment-84221 Mon, 29 Oct 2007 21:46:21 +0000 http://storageadvisors.adaptec.com/2007/06/28/write-back-cache-battery-vs-disk/#comment-84221 Why not implement this storage on something like Compact Flash that will survive a power outage? Why not implement this storage on something like Compact Flash that will survive a power outage?

]]> Comment on Write-back cache: Battery vs Disk by: Neelesh http://storageadvisors.adaptec.com/2007/06/28/write-back-cache-battery-vs-disk/#comment-76890 Wed, 26 Sep 2007 11:01:37 +0000 http://storageadvisors.adaptec.com/2007/06/28/write-back-cache-battery-vs-disk/#comment-76890 Does IBM and Dell support 512MB battery-backed write cache for RAID 5 or more. I need to find whether they support or not and if possible some references to websites or URL's where i can locate the info. Does IBM and Dell support 512MB battery-backed write cache for RAID 5 or more. I need to find whether they support or not and if possible some references to websites or URL’s where i can locate the info.

]]> Comment on Write-back cache: Battery vs Disk by: Tom http://storageadvisors.adaptec.com/2007/06/28/write-back-cache-battery-vs-disk/#comment-70574 Mon, 13 Aug 2007 19:10:54 +0000 http://storageadvisors.adaptec.com/2007/06/28/write-back-cache-battery-vs-disk/#comment-70574 Mao, if I understand your question, yes, only the write data is synchronized to the other controller. However most of what I've been talking about is synchronization only for the sake of redundancy, i.e., in case a controller fails with dirty data in it's cache. If you take dual controllers to the next level then you can start talking about redundant and concurrent paths to the same data. In other words, a host can issue a read command to either of the controllers - perhaps basing the decision of which controller on its view of which controller is least busy. If the controller pair supports this mode of operation, then every read will have to be "checked" against the other controller to see if a more recent version is in it's cache and perhaps hasn't made it to the other controller yet. As you might imagine, supporting this mode can be extremely difficult, and I don't believe it's very common in "normal" storage. I say "normal" because super-high-end, crazy-ass storage on mainframes may support this read-snooping, but that's out of my league. TT Mao, if I understand your question, yes, only the write data is synchronized to the other controller.

However most of what I’ve been talking about is synchronization only for the sake of redundancy, i.e., in case a controller fails with dirty data in it’s cache. If you take dual controllers to the next level then you can start talking about redundant and concurrent paths to the same data. In other words, a host can issue a read command to either of the controllers - perhaps basing the decision of which controller on its view of which controller is least busy.

If the controller pair supports this mode of operation, then every read will have to be “checked” against the other controller to see if a more recent version is in it’s cache and perhaps hasn’t made it to the other controller yet. As you might imagine, supporting this mode can be extremely difficult, and I don’t believe it’s very common in “normal” storage. I say “normal” because super-high-end, crazy-ass storage on mainframes may support this read-snooping, but that’s out of my league.

TT

]]> Comment on Write-back cache: Battery vs Disk by: maobo http://storageadvisors.adaptec.com/2007/06/28/write-back-cache-battery-vs-disk/#comment-70494 Mon, 13 Aug 2007 07:42:58 +0000 http://storageadvisors.adaptec.com/2007/06/28/write-back-cache-battery-vs-disk/#comment-70494 Now one more question: the cache only reserve the write data and commands. The others(for example the read command) are not synchronous, are they? Thank you very much. Now one more question: the cache only reserve the write data and commands. The others(for example the read command) are not synchronous, are they? Thank you very much.

]]> Comment on Write-back cache: Battery vs Disk by: Tom http://storageadvisors.adaptec.com/2007/06/28/write-back-cache-battery-vs-disk/#comment-63358 Mon, 02 Jul 2007 13:29:21 +0000 http://storageadvisors.adaptec.com/2007/06/28/write-back-cache-battery-vs-disk/#comment-63358 Yes, Mao. There should be a high speed channel betweeen the controllers for active/active. It's common to use the same back-end channel as the drive interconnet, such as SAS. But some designs use the front-end channel, such as iSCSI or FC. GbE is probably too slow. PCIe would be nice, and it's starting to show up in some dual-controller boxes, such defined by the <a href="http://www.sbbwg.org/home">Storage Bridge Bay</a> spec. As far as the technology, it's all very vendor unique. RDMA would certainly be a nice choice, but a more common choice would be block storage interface commands. TT Yes, Mao. There should be a high speed channel betweeen the controllers for active/active. It’s common to use the same back-end channel as the drive interconnet, such as SAS. But some designs use the front-end channel, such as iSCSI or FC. GbE is probably too slow. PCIe would be nice, and it’s starting to show up in some dual-controller boxes, such defined by the Storage Bridge Bay spec.

As far as the technology, it’s all very vendor unique. RDMA would certainly be a nice choice, but a more common choice would be block storage interface commands.

TT

]]>