Dell EMC PowerStore hands on

Thanks to Ingram had the opportunity to put my hands on a real – and remote – Dell EMC Powerstore array. Picture taken by me in our dataceter are all off now, the system is 600km away from me. The environment it was sitting consist two Dell EMC PowerEdge R740 servers that are connected ot redundant SAN fabric and to the mentioned Dell EMC PowerStore 5000T box.

It is quite a properly configured one, so it is not having just low amount of drive:

  • 21 x 3,8TB NVMe SSD
  • 4 x 8.5GB-os NVMe NVRAM

Above picture shows that starting from the left there are 21 NVMe SSD drives and then the last four bays in the enclosure hosts the NVMRAM modules. Some models have two, but this one as it sits just at the middle of the product line, has four. These NVRAM drives are serving as the cache of the controllers.

In my last article I was guessing the type of the processor(s) in certain models and I was right when was suspecting they are Skylake. The model I have here has two Intel(R) Xeon(R) Gold 6130 CPU @ 2.10GHz CPUs accompanied by 576GB RAM. The last is surprising since this is a “T” model, not an “X” one, so this one cannot run ESXi and workload VMs, but still has nice amount of RAM for some reason. In each controller – all PowerStore kits coming with two and only two controllers – there are two M2 connected SSD, where I believe the operating system is booting, which is based on CentOS.

Looking at the system from the back – disregard the warning that it has different SFP modules installed – a top/bottom design with 180 degree turn design is visible. Both controllers are totally the same and they have the power supplies as well. No need to worry although one power supply can deliver its services to the other controller if that for some reason looses inlet AC.

The look and feel is clean and tidy, only important metrics are shown.

Dashboard

Top left “Overview” can show multiple arrays in a single cluster and their health – I have only one here. The “Capacity” speaks for itself, so anything about used/consumed/free space is behing that menu. Performance is the one which most of the clients interested out there and so as you. By clicking on any of them a whole cascade of information is displayed.

The watchlist is a great idea! The things that a particular admin is interested can be pinned here and access to those entities can be instant. These can be Volume/Volume group/File system/Virtual Machine(in case of model X) and VVols.

Monitoring

Alerts are events are aggregated here, not that much to talk about.

Compute

Host and host groups are managed here. Naturally hosts must be added here in order to allow them to use a volume. This is enhanced by allowing to define groups that can have multiple hosts – like a VMware cluster surely need to have access to the same datastore LUN – having access to the same set of volumes.

VVols are shown here, the VMs that use them for sure and this is where vCenter can be registered.

Storage

No question the top most important place if a volume or a NAS/NFS export is about to be managed/created.

  • Volumes: for managing exports of block volumes
  • Volume groups: for managing the group of block exports
  • Storage Containers: VVols can’t live without “storage container” and that is what can be created here
  • File Systems
  • NAS Servers

If the array is installed by selecting unified mode, than the last two options have importance, otherwise forget them. The creation of a file system must involve first creating a NAS server, by granting an IP (subnet,vlan,GW) to it and defining the protocols we want to use to deliver services These can be SMB(3)/NFSv3/NFSv4. After selecting SMB the NAS server can be added to an Active Directory domain so ACLs can be used as you would normally do on a regular share. Workgroup is an option, but that is not what an enterprise love.

If there is already a NAS server one or multiple file systems can be created.

It is really easy, just name must be told, size and the SMB3 relevant options.

Protection policy is where the snapshot/replication can be configured and scheduled.

What is interesting that Dell EMC does not require, nor ask for the information what the volume will be hosting – which would define the block size/deco if needed – which as last resort could tell them in their cloud analytics platform the the achieveable reduction levels from real life. Some other vendors does this so real life data can be used for sizing and planning. One single thing can be configured on a volume and that is the performance policy. It has three values, high/medium and low. Whatever those mean.

Clicking on a volume or a file system opens up a view zoomed into that entity.

The exported and consumed capacity is shown, further detailing the ratio and qantity(GB) of thin provisioning and snapshots. Capacity planning is also possible since the history of the consumption also available. Real time data is ingested also, current performance can be measured hitting that volume, so as the protection policy it has and as well as how many hosts/host groups have access to it.

Protection

The place where the async replication can be configured (no sync rep so far) between PowerStore arrays and the protection policies I mentioned before.

A single protection policy can contain multiple protection rules. Each rule can tell the array when to create a snapshot, how long it must be retained. When it comes to replication the RPO can be defined. I failed to understand although how a snapshot can be made consistent, as the array is not talking to the application (VMware or Windows/Linux if bare metal) at all.

The max snapshot quantity is 256 in any combination and the minimum RPO can be low as 5 minutes, the highest is 24 hours.

Migration

A part that you will not use that regularily since this is how to import volumes sourced from Dell EMC PS,SC,Unity and VNX2 systems, all without disruption.

Summary

The Powerstore interface is looking great and fresh – thanks for leaving all that old and rusty UI behind – however it needs some revision still as it is interesting that block size is displayed at some places, but the array gives no option to configure it for volumes. Then why is it showing it?

I also hope that he 256 snapshot limit will be increased and somehow consistent snapshot option will be included by allowing the array to tell vCenter – and VMware Tools indirectly – when a snap is created. Same if a volume is used by bare metal Windows that it will be able tell to VSS. It is still a question of mine what is the reason of putting the cache into drive bays as those consume precious space where otherwise capacity would be sitting.

I’d wait till next revision and see if RAID6 will be added and sync replication enabled, but even with it’s first release PowerStore has a place in the market, but I doubt that in the segment Dell EMC wishes at day 0.