Like most in the storage industry, I’ve been paying close attention to EMC World 2015. In the absence of a genuine industry technical conference, EMC World provides the market with insight and perspective from the industry’s king-pin; an unveiling of the efforts of EMC engineering and a glimpse of where their sales and marketing efforts will be focused.
While the EMC product portfolio spans a broad gamut, from cloud software to development platforms, enterprise storage remains EMC’s core competency. Through all of product updates, releases and technical tidbits, the message is clear: the next-generation of storage infrastructures will be composed of all-flash storage architectures. In support of this message EMC highlighted futures like DSSD and all-flash hyper-converged offerings, updates to XtremIO and results from the recent Gartner Market Share Analysis: SSDs and Solid-State Arrays, Worldwide, 2014 (where EMC and Pure Storage rank #1 & #2 respectively).
The focus of all-flash messaging centered on the updates coming to the XtremIO platform including larger capacity X-bricks, larger scale clusters, greater performance and new data services and features. I’d like to share my 5 takeaways on the announced XtremIO 4.0 (aka ‘The Beast’) from EMC World…
- XtremIO is hardware-defined and cannot scale-out with different capacity X-bricks: The non-disruptive scaling is limited to identical X-bricks. Thus a customer who deploys XtremIO today must choose between future expansion based on old hardware or a disruptive and expensive forklift upgrade to modern X-brick hardware. Compounding this disruption is the dependence of applications on array-based data services like data reduction, replication, and clones. These features can not be non-disruptively migrated at the application layer (i.e. VMware Storage VMotion, Oracle ASM, SQL Availability groups, etc) – leading to the same old lengthy forklift upgrade process.
- 18 months post initial launch, non-disruptive scale-out and dual parity data protection finally arrive: I bet no one is more happy for the arrival of these two features than EMC sales teams and partners as this update will allow technical validations and proof of concepts of XtremIO to at least partially catch up to EMC’s marketing claims. These updates are also good for customers, as they receive enterprise grade data protection and will help technical evaluations move beyond simply puling two drives.
- Native replication does not mean array-to-array replication and requires additional infrastructure: Much like non-disruptive scaling, XtremIO 4.0 replication has some important caveats to understand. The upside is XtremIO customers will no longer have to use VPLEX to provide the RecoverPoint IO splitter for XtremIO volumes. Beyond that it’s business as usual, customers are still required to purchase RecoverPoint hardware appliances, software licenses and additional storage capacity for the local and recovery journal volumes.
- Comprehensive data Reduction continues to be critical in enabling the adoption of all-flash arrays: Data reduction has allowed all-flash to replace disk storage at scale, yet many are still skeptical. Kudos to EMC for sharing the data reduction results of the XtremIO install base. When you compare them to the results published real-time by Pure Storage on our homepage, I think it is pretty clear that we’re out-delivering XtremIO significantly in terms of data reduction effectiveness (more on this in a forthcoming blog post) – and by extension total cost of ownership. Transparency helps instill confidence in the market and maybe this will help advance the storage industry to come to a consensus on reporting data reduction results.
- EMC continues to promote unrealistic ‘Millions of IOPs’ claims: Performance capabilities is no longer the primary driver of the adoption of flash. The consistent sub-millisecond latency ‘modernizes’ the responsiveness every application – from demanding databases to batch processing to responsive virtual desktops. Case in point is a recent post from my friend Chad Sakac, where he highlights an XtremIO customer who has an ‘extreme’ workload defined as 125,000 8KB IOPs. His point was absent as to the gains all-flash provided the business process> From an all-flash perspective the IO load our entry level FlashArray FA-405 is rated at 100,000 32KB IOPs and can easily sustain ~200,000 8K IOPS. If this example is ‘extreme’ then who’s the target customer for a 2 million IOPs ‘Beast’ configuration?I’m sure some will read this post as a negative response from a competitor… which is a fair claim.
Before you label this post as, “haters are gonna hate” understand that these details often get swept under the rug until Pure & EMC engage in a competitive situation. Transparency is the new norm; it’s best for building long-term relationships with customers.
Pure Storage has set out to reinvent storage; this goal is greater than the nuances within this post or architectural debates around scale-up vs scale-out. The FlashArray delivers above and beyond the capabilities of the XtremIO as it…
- Delivers consistent sub-millisecond performance through hardware failures and maintenance windows
- Provides demonstrably better data reduction results that make flash affordable for all performance-centric application workloads
- Eliminates forklift upgrades and data migrations, allowing for future scaling with new hardware technologies while preserving all array-based data services and the business operations they support
- Eliminates the additional infrastructure and software costs associated with external data services like data replication and array management.
- Independently scales performance and capacity up and/or out to keep in line with evolving customer needs, without any unnecessary (and expensive) overhead.
These are just some of the ways Pure Storage, a 5-year-old private company, has reinvented storage and is consistently rated alongside the 35-year-old industry leader, EMC. Kudos to all at EMC for the advancements made across the entire portfolio. We may only compete for a portion of the overall market’s storage needs, but it looks like we will continue to battle for those looking to deploy the next-generation of their storage infrastructure.