Today EMC announced their new and improved mid-range storage at EMCWorld: say hello to Unity! It will be replacing the VNX1 and VNX2 systems with simple, modern flexible and affordable new platform and can be purchased in either an all-flash or hybrid configuration. There are quite a lot of differences between both the new Unity and the old VNX systems, with a number of improvements (available at GA or later) that make me really, really happy. Let’s dive in!
All-flash or Hybrid
As said, you can configure the Unity systems either in an all-flash or hybrid configuration. In the datacenter itself you can spot the different types based on the all-chrome Unity logo for the all-flash array, and the chrome+blue logo for the hybrid arrays.
All Unity systems (starting at the smallest Unity 300 all the way to the Unity 600) are 2U in size and contain 2 storage processors. Each processor has an Intel CPU (core count goes up on the faster systems) and a varying amount of RAM. What’s different compared to the VNX is the Max # of I/O modules: for each system, this is 2.
Drive counts for each system start at 4 drives (the bare minimum), going up to 150 disks for the Unity 300 and up to 500 drives for the Unity 600. This gives you 900TB of capacity on the smaller system and up to 3PB on the biggest Unity box.
Connectivity of the storage processors to additional disk enclosures is done using SAS 12Gb/s cabling, starting with the onboard ports. The SAS controllers have native Data at Rest encryption (D@RE) onboard. Take note: you will only be using the onboard SAS controllers on the Unity 300 and 400. An expansion module containing 4 additional SAS ports is only available on Unity 500 and 600.
Other I/O modules includes 16Gb/s FC and 10GbE Ethernet (either copper or fiber). Take note (and this is one of the big changes compared to VNX: file and block data will use the same Ethernet ports!
There’s different types of flash drives being used in the Unity system: SAS Flash-2 (currently up to 1.6TB/disk) rated at 10 Writes per Day (10WPD) and SAS Flash-3 (currently up to 3.2TB/disk), rated at 3WPD. The latter can only be used in all-flash storage pools, and not in hybrid, FAST VP pools!
If you don’t want a physical Unity system in your datacenter, you can also download the Unity VSA and run it in your VMware environment (e.g. on VNX storage or local VMware storage). Good use cases for this is either for training (like I’ll be doing the next couple of days) or for small branch offices where you don’t want to put a real, physical, expensive box.
File and Block now DO play together nicely
Can I put this out here? FINALLY!!!!!
I’ve been deploying CLARiiON and VNX arrays for ages and I absolutely LOATHED the Unified systems, with their datamovers, control stations, non-functioning GUI and all kinds of “file vs block” restrictions. I have advised customers more than once to either buy an Isilon, or to just keep it at a Windows VM. It was apparent that Celerra was an old beast and could not keep up. Well, no more!
The “data movers” are now virtual, running in a Docker container on the storage processors themselves. From a hardware perspective, this means that 2U DPE is all you need for either file or block connectivity.
The Ethernet ports (2 onboard, rest on expansion I/O modules) will service both the file (SMB and NFS) traffic and the block (iSCSI) traffic. Of course neatly separated using different VLANs and subnets.
Storage pools no longer have to be separate for block and file: you can create one storage pool and use it for both purposes. These pools can still be homogeneous (one type of disks) or heterogeneous (up to 3 types of disks, with Fast VP moving data to the correct tier).
Unity systems are all thin provisioned. There’s no option to thick provision a LUN anymore.
Sneak peak: while not available at GA, the ability to shrink a storage pool is on the roadmap for the Unity system! I’ve sent in so many feature requests for this.. if a customer wants to move disks from one oversized storage pool to another, undersized storage pool, this should finally be possible. I haven’t heard any official availability dates just yet, only rumors of Q4 2016 / Q1 2017. We’ll see and keep our fingers crossed!
By the way: FAST Cache can be shrunk straight from the start (one RAID1 pair of disks at a time) without taking the entire FAST Cache offline in the process.
HTML5 & Unity Management
Bye Java; Unity will be using an HTML5 interface.
As you can see on the screenshot, the new interface looks a lot more user friendly compared to the old Java-based Unisphere beast. There’s also a new Unisphere CLI to manage the systems, a REST API and something new called CloudIQ which uses cloud-based analytics to proactively resolve potential issues in a customer’s system. Unfortunately this information is not accessible to EMC Partners; just the customer itself.
Gone is the concept of storage domains in Unity: the Unisphere HTML5 GUI and CLI are aimed at managing one system at a time. If you need to manage more Unity systems at the same time, you can add them to Unisphere Central.
The Unity system supports pointer-based snapshots, clones and remote replication (synchronous and asynchronous).
Replication is either native file/block between two Unity arrays, or you can glue a RecoverPoint cluster to the Unity system and go cross-platform. Keep in mind that native replication in the Unity systems is not compatible with the VNX1 and VNX2 systems: so for migrations from VNX to Unity, it’s not as simple as replicating the datastores over.
My thoughts on Unity…
So far the list of features looks promising. Unity tackles quite a lot of the gripes I’ve had with VNX systems (Java based management, clunky file/block integration, storage pool limitations) and further standardizes the hardware.
During EMC World I’ll be attending a couple of tech deep dive sessions to get some more details on Unity and to play with some of the Unity VSA systems in the vLabs. Stay tuned for more! And in the mean time, check this EMC news release for more info on Unity.
