The Storage Team Blog about file services and storage features in Windows and Windows Server.
This post is a part of the nine-part “What’s New in Windows Server & System Center 2012 R2” series that is featured on Brad Anderson’s In the Cloud blog. Today’s blog post covers storage management through SMI-s and how it applies to the larger topic of “Transform the Datacenter.” To read that post and see the other technologies discussed, read today’s post: “What’s New in 2012 R2: IaaS Innovations.”
Although my day job is no longer working on SMI-S, I did want to let folks know that a lot of work went into the Standards-Based Storage Management Service (aka Storage Service) in the upcoming Windows Server release. Here are the highlights:
Discovery - Discovery is what happens when you register (Register-SmisProvider) or update (Update-StorageProviderCache): the storage service tries to find out as much information as you wanted (there are four levels of discovery) and that information resides in the service’s cache so you don’t need to constantly go out to the provider to get the data. In Windows Server 2012, the way the information was discovered was through “walking the model”, which is to say, starting with an object and then following associations to get additional information. Unfortunately, this can take a long time on all but the smallest storage configurations.
For Windows Server 2012 R2, the mechanisms have been changed. Instead of model-walking, the service will do enumerations of objects and then figure out (in memory) how they inter-relate. This turns out to decrease discovery times up to 90%! We worked with vendors to make sure providers will work well with this change, and for some of those vendors, you will need to get updated providers.
Updates (through Indications) - I recently blogged about the indication support for SMI-S. The internals of this are much improved in Windows Server 2012 R2 and more provider changes will be caught through indications so that rediscovery won’t be needed as often. The information in the older blog still applies, except that the firewall rule will already be in place (the provided script will still run unchanged).
Secure connections (using HTTPS) - In my first posting, I advised against using Mutual Authentication with the storage service. For Windows Server 2012 R2, this has been improved. This applies to indications as well as normal SMI-S traffic. Follow the Indications blog for configuration information (you need to have the certificates in place). Not all providers will work well with mutual auth. I will post more about security in the near future.
Resiliency Settings - When creating pools and volumes using the Storage Management cmdlets, you could easily specify various parameters that were just never going to work. SMI-S is too generic here. This has been simplified - stick with vendor defined settings and you should be fine. For Windows Server 2012 R2, we’ll give you an error if you try to override any of the parameters.
Pull Operations - One of the efficiency improvements is to change how enumerations are done by using a newer mechanism called “pull” operations. (You can read more about this here.) This allows chunking of the data coming back from the provider, and generally will lower the memory required on the provider side for large configurations. This works now with EMC providers; others will update in the future. To enable pull operations, you will need to modify the registry value
HKEY_LOCAL_MACHINE\Software\Microsoft\Windows\Current Version\Storage Management\PullOperationCount
Set it to something like 100 which tells the provider to send 100 instances of a particular class at a time. Through PowerShell, this would look like:
To turn it off, just set the value back to 0. (Leaving it enabled for a provider that does not support pull operations will only cause a small performance degradation so you might want to leave this set if you have more than one provider and either supports pull operations.)
Registering a provider - You can now register a provider that does not have any arrays to manage (yet). An Update-StorageProviderCache cmdlet will find any new devices at a later time. More significantly, if for some reason the provider lost contact with an array, this would result in errors that could be difficult to recover from.
Node/Port address confusion- In Windows Server 2012, we had these defined backwards from what is mandated by standards or used by things like the iSCSI initiator. This has been corrected but may require provider updates because we found some bugs in existing implementations.
Snapshot/Volume deletion - Under some conditions, it would not be possible to delete a volume or a snapshot. For example, we sometimes thought a volume was exposed to a host when it wasn’t, or a snapshot was in the wrong state for deletion. This has been improved.
Snapshot target pool - Specifying a -TargetStoragePoolName for snapshots is now supported, that is, if the provider/array allows it. However, be careful when you have more than one array with pools of the same name (which might be common).
Masking operations. There have been cases where masking/unmasking operations can take a long time to complete, particularly when SC VMM issues multiple requests in parallel; these could result in timeouts. Also with the current SMI-S model, masking operations performed by Windows might require multiple method calls to providers. Windows Server 2012 R2 now uses jobs for such operations instead of making all masking operations synchronous.
To see all of the posts in this series, check out the What’s New in Windows Server & System Center 2012 R2 archive.