Pure Nonsense – Separating Fact From Fiction in Flash
At Dell we have been designing, developing and supporting Tier-1 mission critical storage environments for decades with industry-leading solutions. Although each customer defines “Tier-1” based on their unique needs, there are some common requirements everyone agrees must be offered for an all-flash storage portfolio to be considered for use in a Tier-1 mission critical environment.
Mission Critical Storage Checklist:
- Ability to scale capacity while maintaining consistent & predictable performance
- Proven remote replication providing any recovery SLA – from asynchronous and synchronous support up to continuous availability (active-active for zero RPO/RTO)
- Intelligent and efficient copy (snapshot) technology, and integrated copy data management, with application integration for simple and fast copy creation
- Simple and non-disruptive migration tools to move workloads between arrays and migrate to future arrays as needed
- Ability to consolidate mixed workloads while maintaining performance and protection, including block and file on the same array if required
- Data-at-rest encryption and other security methods, like secure snapshots
Many vendors claim to have Tier-1 mission critical storage, but knowledgeable customers can quickly see through the hype. Just saying your storage is for Tier-1 or mission critical environments doesn’t make it true, you need the proven architecture and data services to support it.
For example, Pure Storage has recently made the following claim:
“FlashArray//X now represents a higher-performance tier for mission-critical databases, top of rack flash deployments, and Tier 1 application consolidation”
The announcement of their new FlashArray//X (//X) has forced Pure Storage to make this bold (and questionable) claim and relegate their previous generation, the FlashArray//M (//M) series, as “most economic all-flash consolidation”. But what does that mean?
The implication is that the new //X array has something new, or at least improved that makes it more suitable for mission critical/Tier-1 application environments than the previous //M series. However, all we can see is that it should be faster than the FlashArray //M thanks to Pure’s proprietary NVMe drives (aka – flash modules). That’s it… there’s no evidence of new or improved Tier-1 mission critical data services!
After much research on Pure’s website, it turns out that the only real difference between the //X and the //M series is the use of proprietary NVMe ‘Direct Flash’ modules and ‘Direct Flash’ software, which Pure Storage claims will make the //X faster than the current //M series. However, as the //X isn’t generally available yet, and since Pure Storage isn’t publishing any performance numbers, no one really know what that means.
Fun fact: Pure Storage recently removed all FlashArray performance metrics from its website and data sheets – perhaps they don’t want you comparing FlashArray to other industry-leading all-flash arrays?
So the question is, how does an unspecified performance improvement, and no new or improved data services, make the FlashArray//X more appropriate for mission critical/Tier-1 environments? The obvious answer – It doesn’t!
However, adding customized NVMe flash modules does make the array more expensive (to cover the cost of all the custom-built proprietary hardware), which means they need to make it sound more capable to justify the cost premium. In fact, Pure Storage has been clear that the FlashArray//X needs to be positioned into mission critical Tier-1 application environments because these environments have larger budgets.
Since Pure Storage now positions its previous generation, the //M series, for “general-purpose consolidation”, it’s clear even they understand it’s not built for mission critical environments. Since it’s all they’ve been selling, they don’t seem to understand the true requirements of mission critical storage. Luckily for you, Dell does… so, make sure you consider the following before putting FlashArray//X into your mission critical data centers.
Consider the following before considering FlashArray//X for your Tier-1 mission critical environments
Controller Architecture
Pure Storage’s FlashArray//X still uses the same active-passive (backend is passive) dual controller architecture
- You cannot scale performance without replacing the controller with a faster controller.
- Capacity will always be limited – one fully active controller can only do so much. Currently the //X can only scale up to 182TB of raw capacity using 9.1TB proprietary flash modules (which are supposed to come later in the year).
- Since only one controller is fully active at any given time, you’re paying (a premium) for a second controller that is idle most of the time.
- What is the //X array overhead for RAID-3D, metadata and other drive management activities? If it’s the same as past generation FlashArrays (~43%) wouldn’t that reduce their current total usable capacity (before data reduction) to about 104TB per FlashArray//X, using their currently shipping 9.1TB flash modules?
- As stated by Pure Storage, 1PB effective capacity requires 5:1 data reduction and the currently unavailable 18.3TB proprietary flash modules. As others in the industry have pointed out, this is very optimistic as it is generally not possible to achieve the 5:1 data reduction in a typical OLTP environment (a common Tier 1 use case).
Dell arrays leverage dual fully active and multi-controller architectures
- Dell mid-range all-flash arrays offer a dual fully active controller architecture, which allows greater scalability while maintaining performance with a dual controller architecture.
- Dell high-end all-flash arrays leverage a multi-controller architecture which allows all controllers to be active and shared at all times. This provides the ability to add controllers to scale performance and capacity as the environment grows.
- All of the Dell all-flash arrays can scale to multiple petabytes while maintaining consistent performance.
Mission Critical Availability
Pure Storage is marketing their new FlashArray//X as being always-on and having six-nines availability before the first units have even been installed! The calculation here is questionable at best. Dell offers at least six-nines of availability with an architecture that has been proven over decades of use, in the world’s most demanding environments.
Mission Critical Tier-1 Data Services
FlashArray//X is lacking key data services:
Insufficient Mission Critical Remote Replication
- No current synchronous remote replication, which limits the recovery SLAs the array can offer – introducing risk of data loss in the event of a disaster at the main data center.
- We hear that synchronous replication is coming, but most customers will be very leery about using version 1 replication to protect their mission critical data.
- No active-active replication for ‘continuous availability’ with zero RPO, zero RTO.
- Limited multisite remote configurations – no star or cascaded configurations.
Limited Snapshot Technology and NO Integrated Copy Data Management
- FlashArray//X snapshot capabilities remain the same as the //M series and there is no way to provide ‘secure’ snapshots.
- No integrated copy data management (iCDM) tool to help automate and simplify leveraging snap copies.
No non-disruptive data migration tools
- When you max out a FlashArray’s capacity, or performance, how can you move workloads to other boxes without impact to the users?
Cannot Consolidate Block and File on the Same Array
- Most tier-1 application environments are leveraging some ‘file’ workloads in addition to their ‘block’ workloads. Having to use different arrays for your block and file workloads dramatically adds costs and complexities.
Dell arrays have a host of proven mission critical data services and tools that have been in mission critical production environments for years
- Proven remote replication tools providing async, sync, multi-site, active-active, and more…
- Fast, efficient snapshot technology to create and use snapshots without impacting performance or capacity.
- Non-Disruptive Migration (NDM) tools for simple workload migration between boxes, with no impact to production applications.
- Unified block and file on the same array.
- Data at Rest Encryption (DARE) and secure snapshots for increased security of data.
- And many more features you can read about here.
I hope you’ve enjoyed this blog and watch for more as we focus on other relevant enterprise storage themes.
You may be wondering… what about NVMe – isn’t it a requirement for enterprise storage today? Not exactly, for now see here and stay tuned for more on that topic in an upcoming blog!
