PowerMax Attack of the Clones
Fri, 22 Jul 2022 20:27:07 -0000
|Read Time: 0 minutes
In typical blog fashion, I tried to come up with a clever title, and this title seemed obvious. But if we are going with a movie title, Back to the Future may be more appropriate.
The announcement of PowerMax 2500 and 8500 arrays include many new features, including in our local replication software suite TimeFinder. Most improvements are transparent to the user interface, including cache and storage efficiency, target volume performance, and consolidate the benefits of the various modes users previously had to choose from. But another key change, and the reason for this blog, is that we are putting renewed focus on a feature that we had planned to deprecate: TimeFinder Clones. Here is the behind-the-scenes story.
When SnapVX was introduced with the VMAX3 platform, we felt it could eventually replace all other TimeFinder options: Mirror, Snap, Clone, and VP Snap. SnapVX immediately replaced TimeFinder/Snap which has never been supported in Solutions Enabler on VMAX3 and later. SnapVX targetless snapshots are lighter on metadata than TimeFinder/Snap ever was, allowing for many more snaps: up to 1,024 per volume and millions per array.
SnapVX snapshots use fewer resources because they do not use target volumes to capture a point-in-time. And snapshots may never need to be accessed. Metadata efficiency was a key driver for developing targetless snaps to begin with. We heard many times from customers that most of their snaps and clones were never directly accessed and were only in place in case data recovery was needed. But those copies consumed volume ids, which in turn consumed metadata, neither of which are infinite resources.
SnapVX targetless snapshots not only save resources, but they also save a ton of time! There is no need to create the copy volumes, groups, and source-target pairs that were traditionally required for snaps and clones. When an array is deployed you can immediately protect an application with a single command, REST call, or UI operation. This capability was a great advancement at the time, and we have enhanced it with improvements and new features such as metadata efficiencies, zDP, and Snapshot Policies.
Host access to a snap or clone is often needed. This brings us to how SnapVX was intended to replace the rest of the TF options.
Host access to a SnapVX targetless snapshot is provided by linking a target volume (or target storage group) to a snapshot. Like a traditional snap or clone, the application host could read from, and write to, a point-in-time image without affecting the source. The intent was for customers to use linked targets to provide copies of the same or different point-in-time to many users for test and development, online backups, and other uses. And the targetless snapshots themselves provide application protection and restore capabilities with no target volume required. Many customers have been using SnapVX this way with great success.
We expected linked targets to allow us to eventually deprecate Clone, Mirror, and VP Snap. However, as has been the story of TimeFinder since it was first developed decades ago, our customers find other ways to use it, and so we adjust accordingly.
While targetless snapshots provide excellent application protection and capability to restore point-in-time data, there is more customer demand to restore modified target data than we had anticipated. Doing so with SnapVX linked targets requires that a new relationship be established between a snapshot of the linked target and the original source volume. While PowerMax data efficiency features avoid a full copy locally in these situations, a full remote copy is required if the original source also has SRDF protection. Clone can differentially restore from a target volume and is an excellent option when this scenario is required for a solution. (The full copy is only required when restoring modified data from a target volume, for example, during failback after running production on linked targets of SRDF R2 snapshots. Restoring a point-in-time from a SnapVX snapshot is differential).
There are times when only a few target volumes are needed for host access. In these situations, it can be easier to “cut out the middleman” and create a Clone instead of creating a snapshot and then linking a target. Targetless snapshots provide many benefits, but they are an entity of their own that need to be managed and will use system resources regardless of a linked target. Management is not a bad thing by any means. Targetless snapshots can be the right tool for the job in many situations. But Clone is another useful tool to have in your toolbox. We have talked with customers who created a targetless snapshot to link a target, and then immediately unlink and terminate the snapshot, and continue working with the (now unlinked) target. This workflow certainly will get the job done. But, in those situations, it can be easier to create a clone.
As SnapVX usage increased across our customers, we recognized that there are still valid uses for Clone. There are plenty of situations when all you need from SnapVX is the targetless snapshots, and plenty of other situations that call for linking a target to a snap. But sometimes you may only need a target volume. We could have built that functionality into SnapVX, for example, if we added an option to allow linking a target directly from the source volumes. But we decided to continue Clone because it allows customers to keep using their existing scripts.
Continuing Clone also saved development time on our end, and we used that time to improve Clone. Enhancements include increased scalability (256 clones per volume, previously 16), storage efficiency like SnapVX linked targets, and the option for Unisphere to create the target volumes during the Clone operation.
We have deprecated Mirror in PowerMax 2500 and 8500. But we are keeping BCV volumes around as an option for customers who want to use them for Clones and SnapVX linked targets. BCVs make it easy to identify your copy volumes and do not change any feature functionality.
VP Snap is also being deprecated. We are applying the data efficiencies of VP Snap to Clone, so there is no need for VP Snap.
Deciding not to deprecate Clone is a change of direction on our part, and that is not always easy to explain, but it is the right thing to do. We are happy to discuss this decision with our customers because it gives them another tool that is best for some jobs. SnapVX and Clone make for a complete TimeFinder solution, and they are not mutually exclusive. Some jobs will benefit from all options. SnapVX snapshots can protect the application, while linked targets provide host-access to point-in-time data, and clones quickly create on-demand, host-accessible copies of the latest application data.
Author: Mike Bresnahan, Senior Principal Technical Marketing Engineer