I’m a child of the ‘80s, which means I grew up with the obligatory posters of the Ferrari F40, Lamborghini Countach, and Porsche 959 on my wall. I remember driving along one day and nearly losing my mind because a Ferrari F40 was stopped at a red light up the road. I was so excited to see this rare exotic car in real life. When I pulled up next to it something just didn’t seem right but I couldn’t quite place it. As the light turned green the Ferrari pulled away and I didn’t hear the roaring twin turbo V8 I was expecting, but something rather anemic. Then it hit me. This was no Ferrari F40 – it was a Pontiac Fiero with an F40 body kit.
There are a lot of all-flash arrays on the market. There are no posters, but like supercars of yore, IT administrators, storage engineers, DBAs, and application owners are lusting (yes, all-flash arrays are sexy!) after them and dreaming about piloting one of these beauties in their data center. The question is – do you know how to tell the Ferrari from the Fiero?
Today we released XtremIO version 3.0, adding inline compression and substantial performance improvements to what was already the #1, fastest, and most game-changing all-flash array on the market. Best of all, these are software updates available to our existing customers to run on their existing XtremIO arrays at no additional cost. Inline compression augments our existing inline global deduplication technology and together these allow XtremIO arrays to commonly fit six times more capacity into the array than the physical usable flash would otherwise allow, making the economics of owning XtremIO extraordinarily compelling.
Being #1 in a market means you have a big target on your back. Many vendors want to position their products as competitive to XtremIO. They do this by playing the “checklist game”. Make a big list of features and call out all the things they have just like XtremIO – thin provisioning, deduplication, compression, encryption, snapshots, VMware integration, scale-out, and more. Toss in some performance hero numbers and voila, they’ve got an XtremIO-beater. But just like putting an F40 body on a Fiero didn’t yield a Ferrari or fool a 16-year old, these tactics shouldn’t fool you. So here’s a quick reference guide for distinguishing the F40 from the Fieros in the all-flash array market. Things may look the same at a glance, but inspect a bit and they aren’t even close.
|Why should you care?|
|Thin Provisioning||Always thin provisioned. Full performance on thin LUNs. Kilobyte granular allocation size.||Thin may be disabled. Thin reduces performance when enabled. Many MBs to many GBs allocation size||Ensures only the amount of flash needed for incoming I/Os is allocated, leaving the most amount of capacity free at all times.|
|Deduplication||Always on, always inline, global across all LUNs, unlimited history. Never slows or stops. No duplicate block ever touches the flash.||Post-processed. May slow or stop. Duplicates land on flash. Works only within a volume.||Inline data services ensure consistent latency response and maximize flash endurance.|
|Compression||Always on, always inline. Never slows or stops. Only fully compressed blocks land on flash.||Post-processed. May slow or stop. Partially compressed or uncompressed blocks land on flash.||Inline data services ensure consistent latency response and maximize flash endurance.|
|Encryption||Always on, always inline. Never slows or stops.||Post-processed. May slow or stop.||Inline data services ensure consistent latency response.|
|Snapshots||Inherently writeable. Space efficient for user data and metadata. Same performance and data services as parent volumes. Instantaneous operations.||Lose significant performance. Metadata consumes lots of flash capacity. Operations slow with advanced topologies.||Leverage snapshots as a productivity tool for high performance active volume clones. They are not simply point-in-time copies for backups.|
|VMware Integration||Supports all VAAI verbs with metadata enhanced XCOPY handling.||Supports none or only some VAAI verbs. XCOPY is handled by brute force.||Allows dynamic and rapid VM cloning while the array is under high load.|
|Scale-Out||Linear performance scaling, with dedicated Remote Direct Memory Access between controller nodes.||Scale-up (dual controller only) or scale-out with non-linear performance gains and network redirects between nodes.||Allows the array to handle consolidated mixed workloads and to deliver the full performance potential of flash media.|
|Balance||Inherent. Achieved as data enters the array.||Post-processed. Achieved manually, or by automated back-end processes.||No hot spots. The array always dedicates the most resources to host I/O and not to internal re-balancing processes.|
|Integrated Flash-Specific Software||No vestiges of disk-array design. Everything optimized for flash.||Legacy disk design such as log-structured file systems, system level garbage collection, and RAID.||Consistent and predictable performance at scale, under high and constant load. Superior flash endurance because of lower write amplification.|
|Performance||Delivered under constant heavy load, on full arrays, with fully random workloads, over extended time periods.||Lower peak and sustained performance with wide swings as internal array processes contend with host I/O.||Consistency and predictability are just as important as peak performance.|
|Metadata||In-memory processing||De-staged to SSD||Ability to handle any random or sequential workload with equal performance. Vastly superior performance for metadata-heavy operations such as deduplication, VM cloning, and snapshots.|
|Data Protection||Flash-specific||RAID||Superior performance, endurance, and capacity utilization.|
|Capacity Efficiency||82% Raw:Usable||~50% Raw:Usable||Better economics. More usable flash for a given raw capacity purchase.|
|Flash Endurance||Enterprise-grade MLC SSDs and flash specific endurance management techniques such as content addressing with inherent balance, always inline data services, flash-specific data protection and in-memory metadata processing.||Consumer grade flash with 1/10th the native endurance often used. Post-processing data services and array balancing consume endurance cycles.||Endurance should never be a concern under any workload for the lifetime of the array.|
|Management||Simple. A couple of hours from shipping container to production. No configuration parameters or tuning required.||Must select RAID levels and optimize the array for the expected workloads.||Focus on value-add projects rather than storage setup and administration.|
|Integration||Vast support within the EMC Ecosystem of products, including converged infrastructure in VSPEX and VCE Vblock configurations.||Limited or none.||XtremIO’s value extends to PowerPath, ViPR SRM, ViPR Controllers, VPLEX, RecoverPoint, and more.|
Our motto at XtremIO is Xpect More and our mantra is Architecture Matters. XtremIO isn’t just a storage array filled with flash that goes fast. It is a new tool for optimizing application environments (read about how we do this for Oracle here) to become more efficient and powerful than ever before. It takes a unique and innovative design to break new ground. If you learn how to distinguish the Ferrari from the Fieros in the all-flash array market, you’ll transform your operations just as XtremIO customers already have. In their own words:
“I am a strong believer in XtremIO and the impact it is having in our virtual environment. You have been an incredible partner and I look forward to the future. It’s only going to get better.”
“The installation was nothing short of amazing.”
“We cloned a 75GB VM which normally takes us 20-25 minutes and it came back and said it was done in 30 seconds. After some research we found that that was no fluke – it actually takes less than 30 seconds. Just amazing amazing speed.“
“We started toying with running other workloads on XtremIO. We had some database performance issues and moved the temp files and log files to XtremIO. The call we got back from the CRM team was ‘What did you guys do? This is running better than it ever has.’ It made all the difference for them and what they were seeing as far as throughput.”
“I am looking for Star Trek technology. If we are going to get to that level, we can’t be twittling around in the details. We need things that just work. Between virtualization and XtremIO we have boxes we just plug in. I think that is what’s going to allow us to get to the next level.”
“XtremIO delivers upon what matters the most: Availability and Performance. Neither are worthwhile without the other. Our customers demand solutions that are always on with the lowest latency possible.”