The VMAX3 (100K, 200K, 400K), the latest evolution in our high-end storage arrays (Symmetrix), officially went GA today. This platform has been some time in the making and it took the combined effort and long hours of many individuals in engineering to bring it to fruition. As there will be many other blogs/announcements concerning VMAX3 with HYPERMAX OS 5977, I’ll just focus on what I consider to be the biggest advancement in data placement since the introduction of FAST many Enginuity versions back. The VMAX3 presents a new storage paradigm – that of SLO management or Service Level Objectives (not SLAs though give it some time…). SLOs advance us past the existing FAST VP model by providing average response times tied to a particular service. Whereas FAST VP is concerned with moving extents between thin pools to get hot data on say EFD and cold data on SATA according to the policy on the storage group, SLOs provide the customer the ability to choose different objectives based on required response times of their applications. I think this picture from Unisphere demonstrates it better than I can explain it:
So depending on the physical disks in your array (e,g, EFD, FC 10K 15K, etc.) you are presented a number of different SLOs from which to choose. You can even assign a workload type to the SLO (OLTP, DSS) which will help the FAST engine (think hinting). The average response times range from 15.5 ms down to less than 1 ms. When you provision storage to your ESXi cluster, you simply select one of these SLOs that match the requirements of your application. Have all sorts of different applications with different response time requirements (um who doesn’t?), no problem. Our improved cascaded storage groups make this a breeze:
You can see I am using Platinum, Silver, and Bronze SLOs in my example (more detailed in my TechBook – link below). I am also taking advantage of the “catch-all” SLO, System Optimized. This SLO lets the FAST engine work in a similar fashion as FAST VP does today on the current VMAX (though granted the engine is much improved). The reason I picked that SLO is I am unsure yet what the application accessing that device will need. As soon as the requirements become apparent, I’m just going into Unisphere and modifying it:
Note I can also move individual volumes from the storage group into another one with a different SLO:
Very straightforward, very simple with the new Unisphere interface. Speaking of which, here’s the new SLO dashboard:
BTW this model ties together very nicely with vVols, but let’s not get ahead of ourselves. There are obviously plenty of other great things about the VMAX3 – fully thin, pre-configured in the factory, an internal hypervisor which contains embedded services, smaller physical footprint, no more metavolumes, flash vaulting, multi-port mapping (no longer just 0 and 1), and the new Unisphere/Solutions Enabler software to mention a few. The product documentation will have all the details.
I’ve updated all my VMAX/VMware documentation for the VMAX3. I previously updated my VMAX Content Pack for VMware Log Insight for VMAX3 support so that is also ready to go. Here are the links. I’ve included both the emc.com and support.emc.com links. You can tell if you have the right doc by either the date saying “August/September 2014” on the title page, or for the TechBook look for version 10.0.
Using EMC Symmetrix Storage in VMware vSphere Environments
Using VMware vStorage APIs for Array Integration with EMC Symmetrix VMAX
Implementing VMware vStorage API for Storage Awareness with Symmetrix Storage Arrays
Using the EMC VMAX Content Pack for VMware vCenter Log Insight