Simplicity with Kaminario Healthshield & QoS

Kaminario LogoKaminario presented about there all-flash array back at Storage Field Day 7. Back then, I have to admit I wasn’t too impressed: it felt like just another all-flash array with few differentiating factors. For me, that changed after their presentation at Storage Field Day 10. The two components they’ll be adding are a policy based QoS system and a cloud based monitoring system called Healthshield. Both should help simplify storage management and look beyond merely the storage array itself. But first, a quick recap on Kaminario: who are they and what do they do?

Kaminario is originally an Israeli company, whose HQ has recently moved to Boston, MA. They’ve got several offices around the world, including NYC, London and Israel (where they still do R&D). It’s a one product company (their K2 all-flash array, more on that later), with a partner driven sales model. Numbers wise, they’ve got 150+ channel partners and “hundreds of customers”, with on average 2-3 K-blocks per implementation.

The Kaminario product is the K2 array: an all-flash array consisting out of K-blocks. Each K-block contains a dual active-active controller and (initially) one disk shelf. You can either scale up these K-blocks by adding additional disk shelves, or scale out the K2 array by adding additional K-blocks (currently up to 4 per array). Which route you take depends on whether you want extra capacity or extra performance: and you can both scale out and scale up simultaneously. With version 5.5 of their product (launched in August 2015), the K2 array supports native replication.

Kaminario guarantees a 3.3:1 data reduction on all their arrays, and they mean it: if you don’t achieve this, Kaminario will add additional raw capacity at zero-cost for the customer. Their current arrays show a 5.1:1 data reduction, which is slightly lower than last years numbers but explained by the fact the Kaminario systems is now being deployed in more mixed workloads compared to the “just VDI” deployments initially.


Kaminario has good visibility into what their own arrays are doing, but limited visibility into the rest of the IT ecosystem. Their vision incorporates DataCenter Aware Storage that integrates with the other IT stacks and simplifies the management of the storage system (and related tasks). Here comes Healthshield!

Kaminario DC Aware Storage

Healthshield collects analytics of the Kaminario K2 array itself (via Data Aware APIs) and from the datacenter infrastructure via Host Datacenter Aware Agents, initially for Oracle, Openstack, Docker, Microsoft and VMware systems. All analytics collected from the datacenter infrastructure are gathered onto the Kaminario array, and then forwarded to the Healthshield system. Another simplicity aspect: this means you only need to open a set of internet facing ports for the Kaminario array, not every individual DC component that you want to connect with Healthshield.
The various statistics (1000’s collected every second) are then crunched in the Healthshield back-end (offloading the production storage systems) and presented out via a SaaS webpage.

Kaminario didn’t invent the wheel here. Nimble Storage has had InfoSight for a while now, which does the same thing really. But there’s no shame in that: I expect all vendors will be implementing a similar solution sooner rather than later. Differentiators will be how much of the DC ecosystem you can plug into it and the quality of advice you get back.

Policy based QoS

Kaminario also demoed their QoS services, which will be added to the array functionality in the 2nd half of 2016. They state that most QoS solutions nowadays are too complex: and I wholeheartedly agree!

Kaminario QoS

To improve on this, the K2 QoS will be policy driven. The theory behind this is that the customer should be setting a preference via policies, and not be tweaking individual knobs or switches. There’s just too much storage that needs to managed; an admin can’t be setting rate limits, queue depth limits and 10 different toggles for several PBs of data: it quickly turns into a full-time, reactive job not suited for quickly changing workloads.

Thus, an admin will be able to set a customer defined policy (such as gold/silver, production/test, etc) on volumes in a hierarchical level (application -> volume group -> volume).

QoS integrates with Healthshield and will generate recommendations to change QoS policies for a specific application or volume group. It will also look a bit further than just the advantage for that specific volume group and predict the negative impact for other volumes that now have less resources to compete for. Hence a user will be able to balance benfits and downsides relatively quickly.

My thoughts on Kaminario

A typical storage admin has to balance his time between so many things to keep everything up and running: nobody has time to tweak QoS parameters all day. And even if he/she had the time, there’s often no way to fully understand the impact of a QoS change on the entire storage ecosystem. Healthshield and the policy based QoS have the potential to make QoS something that actually adds value, instead of just using the rate limit option to push down the greedy volumes. I’m looking forward to see the real-life results as soon as both go GA.

You can watch Kaminario’s presentation over here. And Dan wrote about them already; you can find his post over here.

Disclaimer: GestaltIT paid for the flight, hotel and various other expenses to make it possible for me to attend SFD10. I was however not compensated for my time and there is no requirement to blog or tweet about any of the presentations. Everything I post is of my own accord.