Briefing Update, CIO, Data Center

Rethink HCI with Pivot3 Acuity Software and X5 Hardware Nodes

If you haven’t heard of Pivot3, you need to fire up Google and take a look at them.  For some, Pivot3 was one of the progenitors of hyperconverged infrastructure and, even before the phrase hyperconverged was invented, had an appliance series with traits eerily similar to what we now call hyperconverged.

Over the past couple of years, the company has passed a number of milestones, beginning in February 2016 with their acquisition of NexGen Storage and followed closely by a $55M funding round in March of 2016.  Since that time, Pivot3 has experienced incredible growth, with triple-digit YoY quarterly growth and high double-digit YoY growth.  And, in April 2017, Pivot3 released their Acuity and X5 Series products.

Pivot3’s research is showing that hyperconverged infrastructure is showing a lot of success with targeted workloads, with customers reporting that:

  • They’re experiencing a lower overall total cost of ownership for these kinds of workloads
  • Thanks to the simplification of the environment, there is less human error, with results in less downtime
  • They are able to focus more on business innovation rather than infrastructure management
  • It’s now possible to deploy new infrastructure and applications in less time than it took on more traditional infrastructure.

These outcomes are not surprising.  Our own research here at ActualTech Media yields the same results, with users seeing a wide range of similar benefits.  At an overarching level, we see that cost savings and operational improvements are the leading outcomes experienced by users of hyperconverged infrastructure.

That said, customers constantly demand more from their infrastructure environments.  Beyond “low hanging fruit” workloads, they want to consolidate more applications, even those that are sensitive to latency issues.  To this end, customers are asking whether hyperconverged will be able to provide appropriate levels of application performance.  They also want to be able to support mixed workloads, while maintaining ease of operations.  In addition, they want to ensure that, regardless of the kinds of workloads they deploy, they can continue to scale in an efficient way and that the platform provides critical data services capabilities.

Simply put, customers want to do more with hyperconvergence, but only if they can be sure that workloads will continue to operate as expected.

Enter Acuity

Depending on the solution, hyperconverged infrastructure can perform very well, but nothing compares (yet!) to NVMe.  Pivot3’s Acuity platform integrates PCIe-based flash storage into the platforms multi-tier architecture.  The whole system is then governed by an advanced quality of service engine that helps to guarantee high levels of efficiency and prioritization.  With this addition, Pivot3 indicates that their platform can now support 2x to 3x more virtual machines than other solutions, and it is 6x as fast as conventional hyperconverged infrastructure offerings.  Some of these improvements came via Pivot3’s NexGen acquisition.  This higher density and increased performance can help further reduce the TCO experienced by users of the company’s hyperconverged platform.  Pivot3 indicates that they expect customers to see 30% lower CapEx costs thanks to the increased virtual machine density.

This new solution provides relief for workloads that may have been storage-bound before.  Perhaps less intuitively, by having a more efficient data path, applications are spending less time in an I/O wait state.  So, from an application perspective, there is a lot more overall efficiency.

Acuity is the name of Pivot3’s new hyperconverged software platform.  In addition, Pivot3 released a series of “X5” nodes, all running the Acuity software.  The X5-6500 is an all-flash HCI accelerator node that includes an NVMe device whereas the X5-6000, while all flash, does not include an NVMe accelerator.  The X5-2500 is a hybrid accelerator node that adds NVMe to the data path. The X5-2000 is the hybrid node without an NVMe device included.

Thanks to NexGen, Acuity also sports full inline data reduction capabilities, a must in a world in which data growth is increasing as a staggering rate.  Acuity brings in NexGen’s deduplication engine and provides between 2:1 and 7:1, depending on the workload running in the environment.

The Architecture

Pivot3’s multitier architecture is quite interesting.  With the NVMe layer, there are three potential persistent storage tiers, and two potential caching tiers.  The entire stack consists of four layers:

  • An ultrafast read cache.
  • NVMe PCIe flash (NVMe-equipped nodes only). A very high speed persistent storage tier and caching layer.
  • A persistent storage layer.
  • HDD (hybrid nodes only). A persistent storage layer.

Architecture

For the X5 models that do not include any spinning disk, what you see here is an example of an all-flash hybrid solution.  The architecture features multiple layers of flash storage, included in such a way as to ensure that the overall solution remains affordable.  For example, few companies today need a hyperconverged system that is all NVMe.  It would be prohibitively expensive. By marrying PCIe-based NVMe flash with high capacity SSDs, the Pivot3 architecture brings customers the best of both worlds.  They get the insane performance of NVMe with the capacity benefits of SSDs.  For those that choose the hybrid nodes, it’s taken a step further, with disks running at much slower speed, but massively higher capacity spinning disk providing a place to store persistent data.

Under the hood, Acuity uses two additional methods to ensure good performance and to maximize usable capacity.  First, to maintain performance, data is written to the SSDs and HDDs using wide striping.  Wide striping means that small chunks of data are written to multiple disks at the same time, meaning that data isn’t waiting for a disk or disk queue to be available.  Second, through the use of erasure coding, this data can be broken up into these small chunks in the first place, and the platform enjoys the capacity benefits associated with erasure coding.

In order to maintain the environment in the event of a hardware failure that takes out a node, data is always written to two nodes.  Once data hits NVMe on two nodes, the write is acknowledged back to an application.

On the scale front, the various X5 models start at either 3.2TB (all-flash) or 12 TB (hybrid) of persistent storage capacity and run up to 30.7 TB (all flash) or 48 TB (hybrid) of capacity per node.  NVMe models have, at a minimum, 1.6TB of NVMe storage available.  The Acuity X5 hybrid solution can scale up to 12 nodes while the all-flash solution can scale to up to 16 total nodes.

Acuity also places renewed emphasis on quality of services.  It uses QoS to manage against performance policies and prioritization.  Higher priority data will live on higher performance media.

There are four ways that Acuity manages QoS:

  • Is there a policy in place for certain data?
  • Has the data been labeled high or low priority?
  • Where data lives – on slow disk or fast NVMe – impacts QoS.
  • These are policies around snapshots, scheduling, and retention.

Pivot3 provides some level of automation of performance policy changes.  For example, you can modify or automate performance associated with an application.  You can schedule performance changes.  For example, you may have a workload that requires more horsepower during the day than the evening.  Or, perhaps finance needs more performance at the end of the month.  You can ensure that these workload performance policies are in place to keep workloads running at the right times.  However, the platform doesn’t yet implement machine learning.  It can’t yet learn about performance patterns and take automatic action based on prior history.  But, overall performance statistics are continuously gathered.  Someday, we may see the ability for the platform to actively learn from historical performance metrics and automate performance policy changes.

If you have an existing investment in storage or servers, you’re in luck!  Pivot3 can present volumes out of Acuity to other systems via iSCSI, so you can use your existing vSphere servers with the platform. Likewise, If you have an existing SAN, you can present volumes from that system to the Pivot3 nodes.  Unfortunately, those external storage arrays don’t get the performance benefit of being in Acuity’s NVMe data path.  It’s a simple volume presentation.

Key Takeaways

If you’re in the market for hyperconverged infrastructure, Pivot3’s approach provides you with eminent flexibility, very good levels of performance, and a solid architecture.  The additional of NVMe to the data path will significantly increase performance, and the ability to leverage your current hardware makes integration of Acuity a breeze, even in the most complex environments.  With the ability to consume external LUNs and to expose internal storage to your current hosts, there are no significant worries to be had around creating islands of infrastructure.

Available in both all-flash and hybrid configurations, and with and without NVMe, there is an X5 configuration available that can meet the needs of both small and large organizations.