Pure Storage has announced a raft of new platform enhancements at its annual user conference, Accelerate. These include updates to Fusion, an AI Copilot, AI workload support and cyber-resiliency capabilities. We dig into the details in this research note.
Background
Pure Storage has made a series of announcements that cover four areas.
- Enhancements to the Pure Storage platform
- AI support and new capabilities
- Cyber-resiliency enhancements
- Storage-as-a-Service improvements
We will discuss each area separately.
Pure Storage Platform
Pure Storage has started to use the term “platform” to describe a combination of system hardware & software and the SaaS management tools Fusion and Pure1. We believe this is a reasonable approach, as these three components are tightly integrated. The hardware solutions comprise FlashArray and FlashBlade, with Cloud Block Store for software-defined arrays in the public cloud. Pure1 provides the analytics capabilities for systems management, while Fusion delivers user and administrator management functionality.
The latest Fusion updates announced at Accelerate provide the capability to implement more targeted automation, either at the system or application layer. Why does this matter? The easiest way to understand this feature is to imagine a large storage environment that has evolved over time.
Typical enterprises will have a range of platforms, deployed at different intervals, some which may have also been upgraded in place. Generally, applications are mapped to a single hardware platform, for resiliency purposes. As almost all IT environments are dynamic, the requirements of applications change over time. Capacity increases and decreases, and performance workloads peak and trough or grow steadily with the business.
On legacy systems, storage managers were required to manually balance workloads, which can be a tedious task (although made simpler with server virtualisation). Moving applications between platforms could also require manually redefining QoS and SLA profiles, again, a tedious business in dynamic environments.
The latest Fusion updates for Purity enable system administrators to automate the tasks of workload management across platforms as well as within individual systems. A fleet of hardware can now be thought of as a pool of systems across which applications may be deployed. This abstraction is important because:
- It enables the best use of resources available to the application.
- It reduces fragmentation (wasted space and performance) on hardware and increases efficiency.
- It reduces costs, both for the vendor and the customer.
This last point is quite interesting, as any savings can be returned to or shared with the customer, making storage-as-a-service more appealing.
AI Copilot
A tech conference wouldn’t be a tech conference these days without reference to AI and, specifically, LLMs. Pure Storage has announced AI Copilot for Storage, an interactive LLM that provides the capability to use natural language as an interface for performance and security management.
It’s easy to think this type of feature may well be a gimmick. Tintri (now part of DDN), for example, looked at this technology back in 2017 and also introduced Slack integration for its VMstore hardware platform. Arguably, this implementation was constrained by the need to format requests quite precisely rather than be able to interpret natural language.
With LLMs, vendors can leverage data (such as that collected by Pure1) to query systems and (eventually) task the Copilot with making system changes. Pure Storage is introducing its AI Copilot in preview this summer. However, at Accelerate, Pure Technical Director Ivan Jibaja demonstrated exactly how the AI Copilot could be used to query and resolve performance-related issues. I recommend watching the embedded video (from 07:10 onwards), which really showcases what an AI Copilot can do.
Like VMware DRS, full autonomy for AI Copilots will take time to be accepted, but we should assume that eventually, this type of technology will be given full autonomy to review and manage systems automatically, based on human input.
AI Storage-as-a-Service
Continuing the AI theme, three AI-related services were announced at Accelerate. The first is a new service-level agreement to deliver storage-as-a-service to meet AI workload performance requirements. As part of Evergreen//One, Pure customers can now sign up to an SLA that guarantees the company will provide the resources necessary to meet AI performance without overbuying hardware.
The second feature is called Secure Application Workspaces. This capability enables customers to ringfence applications and storage with fine-grained access controls, specifically targeted at mission critical AI workloads.
The third feature is NVIDIA DGX SuperPOD certification, which is expected to be completed by the end of 2024. SuperPOD can be thought of as converged infrastructure for AI, based on the latest NVIDIA GPUs. Certification is important as it guarantees the customer-specific performance SLAs and extensive pre-testing of enterprise AI workloads.
Cyber-Resiliency
Stepping away from AI, another important area for the modern enterprise is cyber-resiliency. This term has become a definition for how businesses are able to recover in the event of a ransomware attack. Pure Storage has announced three new cyber-related features at Accelerate.
The existing Ransomware Recovery SLA has been extended with a new Cyber Recovery and Resilience SLA, which essentially offers a more customised approach to the recovery process. This means deployment of new infrastructure and professional services to manage data transfer and recovery.
The second feature is a detailed security assessment to NIST 2.0 standards. The assessment rates customers with a benchmark score, using the capabilities of AI Copilot to generate recommendations and improvements.
The third feature is further enhancements to workload anomaly detection. Specifically, this means using AI to detect changes in application behaviour that indicate a potential ransomware or other attack (such as data exfiltration). As we’ve highlighted before, ransomware is a Darwinian problem and constantly evolves. This means detection methods need to evolve too. As an example, typical ransomware attacks previously encrypted entire files, which would be relatively easy to spot. Today’s attacks target only parts of files that are critical to interpreting the content, making detection much harder.
Storage-as-a-Service
The final area of announcements covers storage-as-a-service and a new Site Rebalance SLA. This feature enables customers to perform a global rebalancing of capacity reserve across multiple sites once per calendar year. This option could be helpful, for example, when data centres are being closed down, restructured or if significant growth occurs in one geographic location.
The Architect’s View®
Rather than specific product updates (like new hardware), the announcements made at Pure Accelerate 2024 have been focused on improving the operational and cost management experience of customers. In a vendor world highly focused on selling AI, it is a welcome change to see Pure Storage not succumbing to the hype of the market.
The fundamentals of business and IT don’t change much. Companies are always looking to optimise and reduce costs, whether that be capital or operational-focused. Data growth is inevitable, far outstripping the ability to manage capacity with legacy processes.
We view the announcements from Accelerate as positive steps forward to improve the efficiency of Pure Storage customers but would highlight one important aspect. Data storage is no longer about the hardware product, but the entire ecosystem that surrounds it. This means the capability to manage at scale, deep dive into performance issues, rebalance and optimise across the fleet of hardware, and continually reduce costs.
As a result, businesses should look holistically at the entire package of services offered, not just the capabilities of hardware products. Increasingly, the ecosystem of management capabilities are the differentiators and the value-add for customers. Pure Storage continues to double-down on these improvements, addressing the needs of customers deploying on-premises storage at scale.
We recently recorded a podcast with Prakash Darji, VP and GM of Digital Experience at Pure Storage, which goes into more detail on these announcements. The recording is embedded here.
We’ve also included a link to our Pure Storage X-Ray and microsite.
X-Ray: Pure Storage, Inc. (Second Edition)
This Architecting IT report takes a deep dive into Pure Storage’s history, products, services, and future outlook. This report is only available for download via paid subscription.
Post #b77a. Copyright (c) 2024 Brookend Ltd. No reproduction in whole or part without permission.