PowerFlex 5.0 - Performance, scalability & simplicity redefined

At boxxe, we’re always exploring the innovations that help organisations modernise their infrastructure and get more from their tech investments. Dell Technologies latest release — PowerFlex 5.0 — is a great example of that.

In this blog, Ian Moore, our technical expert, takes a closer look at what’s new in PowerFlex 5.0 following its unveiling at Dell Technologies World 2025. From enhanced flexibility to improved automation and security, Ian breaks down the standout features, explains why PowerFlex continues to be a smart choice for modern IT environments, and explores the options available when designing a PowerFlex solution.

PowerFlex is a software-defined storage platform from Dell with an architecture focused on performance, flexibility at scale and resiliency.  I like to think of PowerFlex as combining the best of both worlds from 3-tier architecture and Hyper Converged Infrastructure.  It's a solution comprised of separate storage and compute nodes which provide the flexibility of 3-tier architecture with a single management plane & lifecycle services typically found in HCI solutions. 

When considering a storage and compute platform the most important thing to understand is 'why?'.  There are multiple storage & compute solutions within the Dell range alone without considering solutions from competing vendors, so what makes PowerFlex stand out from the competition?  Well, PowerFlex stands out in several areas with the first one being speed.  If you need storage I/O at blistering speed PowerFlex is for you.  This thing is fast…..really fast.  Think Millenium Falcon in hyperspace fast, without the reliability issues J.  PowerFlex has exceptional reliability built in and we will get to that bit later. 

Another area in which PowerFlex stands out is in the name.  It's flexible to the point there are many supported Operating Systems that can be installed, & multiple can be run on the same platform at the same time.  PowerFlex is all about datacentre consolidation.  If you're currently running multiple products on different hardware, you can bring this all under one management plane by consolidating workloads onto PowerFlex.  It's even possible run multiple competing hypervisor solutions side by side on the same PowerFlex platform. I can't think of a reason why you would want to do this, apart from maybe evaluation purposes, however it is a good example of the flexibility of the product.  Supported products include Linux OS, Windows OS, VMware vSphere, Oracle, Citrix XenServer & Nutanix.

Product Tracks (4.x & 5.0)

There are two product tracks for PowerFlex, with 5.0 being the latest release.  It's not possible to perform an in-place upgrade from 4.x to 5.0 (the architecture is just too different to support this) so PowerFlex 5.0 is for greenfield deployments only.  Deciding between version 4.x & 5.0 is a bit like choosing a track of switch firmware.  There are two options with a different feature set; all of which are supported and updated with product & security updates.  Choose the one which is the best fit for your requirements.  For those who want to migrate from 4.x to 5.0 Dell will provide a migration tool in later releases, however it's important to understand this is a migration to new hardware and not an in-place upgrade on existing hardware.   

What's new in PowerFlex 5.0?

With PowerFlex 5.0, Dell have introduced a new storage architecture that eliminates the trade-offs between scale, performance, efficiency and availability.  This new architecture is built on erasure coding, which is a proven architecture for efficiency and resiliency used in many storage products & is a welcome addition to the PowerFlex solution.  It's designed for native block and implemented as a distributed scale out solution, providing resource that scales across drives and nodes.  They’re calling it the Scalable Availability Engine.

Advanced Fault Resilience

PowerFlex 4.x was designed for up to 6 9's of availability with single fault tolerance.  PowerFlex 5.0 takes this up to 10 9's for availability and dual fault tolerance with very fast rebuild.  Both are extremely impressive with 6 9's equating to 31.5 seconds of downtime per year.  10 9's takes this to 0.00315 seconds of downtime per year.  It's important to understand availability numbers are dependent on the number of disks and nodes in the solution with most PowerFlex 5.0 combinations falling between 8 & 9 9's. 

PowerFlex 5.0 provides seamless reconstructs of data in the face of multiple node failures ensuring uninterrupted access by using one of two erasure coding protection schemes.  2+2 protection scheme requires 5 nodes.  The 8+2 protection scheme requires 11 nodes.  There is one node of spare capacity in each protection scheme.  Taking the 8+2 protection scheme as an example, every 8+2 stripe will be written to 10 of the 11 nodes.  A node’s worth of spare capacity is evenly distributed throughout the protection domain & enables the cluster to rebuild the failed capacity to restore the health of the system completely.  All nodes in a cluster are active, with pseudorandom distribution for spare capacity, which results in evenly placed spare capacity on all 11 of the nodes.

With 10 nodes (of 8+2 EC) it’s possible to lose up to two nodes and still have full access to data.  An EC rebuild is now a backend operation that requires very little CPU.  When recovering from a node failure, most of the reads are being reconstructed from parity, which is an expensive overhead on reads.  Including a spare node’s worth of evenly distributed capacity as a best practice provides enough extra space to complete rebuilds and performs as a ‘fault unit’ to hold the full stripe writes if a node fails. 

Improvements to Clones and Snapshots

In version 5.0 there is a firm distinction between snapshots and clones.  A clone is a read/write copy of an existing volume which is persistent if the source volume is deleted.  A snapshot is a read only copy of a volume or clone.  It's a point in time copy that cannot be mapped.  Deleting the volume/clone deletes all associated snapshots.  PowerFlex 5.0 introduces eight times more snapshots and clones when compared to PowerFlex 4.x. resulting in 1022 useable in a volume tree.  This provides additional scalability for VDI environments with snapshots for extensive testing and rollback.  Additional volume clones provide flexibility and scalability. 

Storage Efficiency & Workload Consolidation

PowerFlex 5.0 includes improved capacity efficiency with an increase from 50% to 80% when compared to PowerFlex 4.x.  The introduction of erasure coding is huge when evaluating storage efficiency and is the cornerstone that provides the improvement in this number by jumping from 50% of raw to 80% of raw with improvements from FTT=1 to FTT=2.  Erasure coding plus additional efficiencies such as compression, combined with an increase in raw capacity (192TB raw capacity per storage node) result in a greater than 50% reduction in storage footprint by reducing the number of nodes and racks required to deliver the same amount of usable storage space.  In some cases, footprint reduction is as high as 75% or more!

All volumes span across all storage nodes resulting in a fully distributed storage system.  This results in a platform more resilient to disk and node failures and opens the ability to have more usable and effective capacity.

Performance

As a rule, storage systems with mirroring architecture outperform erasure coding architecture for multiple reasons.  This results in a mirroring configuration being the architecture of choice when performance is a requirement. PowerFlex 4.x includes MG (Medium Granularity) for extreme performance use cases, and FG (Fine Granularity) for capacity efficiency use cases that still require high performance, but not the absolute best. 

When running comparative testing for PowerFlex 4.x & 5.0 on the same PowerFlex hardware, the results show the new EC architecture beats the performance of PowerFlex 4.x in FG (Fine Granularity) across the board and outperforms MG (Medium Granularity) with 8k & 16k block sizes.  The introduction of a write cache in PowerFlex 5.0 plays a large part in increased performance when compared to PowerFlex 4.x which does not include this feature.  When PMEM is available in a PowerFlex 5.0 system, we end up with a scale-out RAM-backed distributed write cache unlike anything else.  Write cache can’t take all the credit for increased performance. There are other improvements that contribute to the overall performance of the PowerFlex 5.0 system including a new light-weight metadata structure, the switch to 8k granularity (from 4k in FG) & much improved CPU threading. 

When comparing 8+2 & 2+2 protection scheme's they provide similar performance.  The only noticeable difference is with large writes at which point the 8+2 scheme absorbs writes faster.  It’s also worth noting there are now two basic node options (performance and extreme performance) to choose from when designing a PowerFlex 5.0 storage solution.

What's not included in PowerFlex 5.0

It's worth touching on a couple of features which are not included in 5.0.  If there are any features listed below are currently in use and are a requirement for your business, this may provide a reason to stay on the 4.x track which is still supported & under development. 

• File services have been removed from the upcoming release of 4.8 & 5.0

• There is no longer a HCI deployment model with PowerFlex.  The focus is on the 2-layer or disaggregated topology, where this product really stands out from the competition

• Asynchronous Replication is scheduled for inclusion in PowerFlex 5.0 next year

• Volume Migration may be included in a later release if considered a requirement based on feedback from PowerFlex customers

• Fault Sets will be reintroduced in upcoming versions of PowerFlex 5.x

In summary

PowerFlex 5.0 marks a real step forward — bringing performance, flexibility & resilience together in a way that helps organisations simplify operations and get more from their infrastructure. With major architectural updates and a stronger feature set, it delivers the scalability and stability modern workloads demand.