Today we announced a new set of platforms, the PowerMax 2500 & 8500. For ease of reference and to avoid confusion, I’ll refer to this as the V4 platform to distinguish it from the previous PowerMax which I’ll call the V3. The boxes run a new operating system, PowerMax OS 10. This OS is unique to the V4 platform. I know in the last release of PowerMax, that OS could be put on some VMAX arrays. Unfortunately the new OS 10 is hand-in-hand with the hardware so it is impossible to apply this to the PowerMax 2000 or 8000. As this is Dell’s announcement of the product, don’t go looking for all the new software yet. Posting of software and shipping of the V4 will come a little later.
With this new PowerMax V4 release, Dell, in an effort to simplify naming for customers, has begun removing “EMC” from the lexicon. Therefore it is the Dell PowerMax, not Dell EMC PowerMax. Realistically, such a change cannot be accomplished all at once as we have so many documents and products out there. Even the small library of documents I write are not completely transformed because all the products are not. For example, it is now Dell Unisphere 10 for PowerMax, yet it is still Dell EMC Solutions Enabler. We’ll get there.
Disclaimer: This is a GA announcement for software which is not yet available. My posts, as well as updated documentation, contain material that as of this date, is accurate. There is the chance that one or more things I write about will change before the software is posted. These would not be major; and I will endeavor to update the posts and documents as soon as possible after learning about them, but if I miss something feel free to correct me.
I’m going to discuss in brief (as much as I can be) new features, changes, and futures. As I work on applications, I’m not going to cover hardware (hint: it’s faster, more scale, more secure, etc.) or every single change, only those I think are most impactful to our VMware integration. I’ve done some deep dives of new features and I’ll link to those for people who want more detail.
You may recall our previous NVMeoF release of FC-NVMe on the PowerMax V3 platform. This is of the same family of NVMe over fabrics. NVMe/TCP is predicted to be a huge growth area , particularly as IP networks become faster and faster. Like FC-NVMe, NVMe/TCP is at its infancy – currently the only platform fully qualified is ESXi 7.0 U3. Certainly more will follow but these things take time. For more detail on this feature see this post.
Local replication – TimeFinder/Clone
Ever since the introduction of SnapVX, we have continued to support the TimeFinder modes of clone, mirror and VP Snap in emulation mode. This was done so that customers could continue to use their existing scripts while making their way to SnapVX. Through customer feedback, we learned that the previous TimeFinder/Clone technology, not simply an emulation, was very useful. For example, when we think of SnapVX we mostly consider them as targetless local backups. They provide a point-in-time without the need for extra devices. These snapshots might be manual or created through policies. If you need a test environment you simply link the snapshot to target devices and present them to the environment. Unisphere or the REST API will do this for you. For quick tests this is efficient but if I wanted to keep my copy around for a while and make changes, the efficiency goes down because during the testing that targetless snapshot is keeping track of changes of the source volumes, think array resources. This is where a clone would be better. A clone is a point-in-time copy of the production device with no ability to rewind (relink) like a snapshot. There is no tether, thus no extra resources being used. It is still efficient since it is copy on write, only bringing over data when it has to.
A couple things to remember about clones. You cannot take a snapshot of a clone, essentially treating the clone like a gold copy. For those of you who test databases let’s say, such a capability would be useful. Perhaps in another release. But the most important thing to remember about clones is that they are NOT consistent by default, unlike SnapVX. You must pass the flag -consistent when creating it. If using Unisphere it is under the advanced options. That will hopefully change in the next release also as my colleagues and I push for it.
The PowerMax 2500/8500 no longer has eNAS, rather a new feature called PowerMax File is taking its place. File is integrated directly into the array like the other containers, unlike eNAS which was essentially VNX code running on the PowerMax. That meant using a different GUI interface to configure and dealing with concepts like control centers and data movers, along with limited replication capabilities. With PowerMax OS 10.0, the new operating system, File can be setup and accessed directly in the embedded Unisphere for PowerMax. You can find File Configuration under the System menu as shown below.
Replication is now handled with SRDF, using SRDF groups very much like vVols in fact. You can do both synchronous and asynchronous. The wizards make the setup a step-by-step process. For those steps check out my other post.
We’ve finally increased the number of SRDF groups from 250 to 2000. This one has been a long time coming and helps us in the future vVol space.
The big change with vVols in this release is the deprecation of the external VASA Provider 2.0. The new PowerMax platform will only support embedded VASA (EVASA), whether or not you plan on using replication. Beyond that, there are no new features though plenty of code changes to aid performance and scalability and even fix a few bugs. Management has changed a little which I’ll now cover.
For both the external VASA Provider and EVASA prior to this release, any management such as log retrieval or starting and stopping VASA, was done in the management interface which was accessed via https :// < VASA IP > :5480. With the new PowerMax, management is embedded in Unisphere in an area called Serviceability.
From here you can click the hyperlink and you’ll see similar areas as the vApp to make changes to the VP parameters or download logs.
As always, some features are bound to either be deprecated, delayed, or perhaps even renamed. We got all three here:
- Symmetrix Access Controls (SYMACLs) are gone in V4. Unisphere moved to a RBAC model for management. This works great in Unisphere, but has no impact to our SYMAPI calls in the SRA. So if you use the SRA with the new V4 platform, this security option is no longer.
- As I mentioned, eNAS is deprecated.
- The Solutions Enabler Virtual Appliance is deprecated. Lots of VMware customers used it, so they will need to move to a regular VM OS solution or use the embedded Solutions Enabler on the array.
- FC-NVMe is not planned for the GA release. Customers wanted focus on NVMe/TCP but I would expect FC-NVMe will be back.
- Non-disruptive Migration or NDM is now called Data Mobility
The Veeam plug-in may not be available when the new platform ships. This is common with plug-ins that they lag array releases. As I’ve explained before, plug-ins have their own release cadence so they don’t always line up with other releases like platforms. My best guess, after discussing the schedule with the developer, is that the update will be out within a month or two. I’m testing it currently.
REST 10 is likely to be supported in the first half of calendar year 2023.
For those who have had challenges with VAAI in their environment, I regret that this isn’t the release to alleviate those. Any changes we’ve made have been fairly minor and are unlikely to make a big difference in any environment. For example manual UNMAP is faster, which also means automatic UNMAP will be a bit faster. The current area of focus for future changes is with RDF as that has caused the most performance issues. VMware will be making some adjustments to UNMAP in their next major release which I hope will benefit our array, but only testing will tell and we’re not there yet. The changes revolve around reducing the MB/s below the default of 25 MB/s. I have done an overhaul on the VAAI paper, however, removing all the old use cases and providing new numbers on the platform.
I’ve updated a few whitepapers and a TechBook for the GA announcement of the V4 and there will be more on the way. The one caution I’ll add (see disclaimer) is that because the availability of the platform does not coincide with the announcement, i.e., you cannot get a V4 today, I may make updates to the papers over the next couple months. These won’t fundamentally alter the technical content, rather it would be updates to names or screenshots. Really just cosmetic, but my experience has been that if a customer sees a different screen than the one I have in the document, it raises concern.