XtremIO Directions (AKA A Sneak Peek Into Project-F)

At DellEMC World 2016, together with Chris Ratcliffe (SVP, CTD Marketing) and Dan Inbar (GM XtremIO & FluidFS), we gave a session on what we call “XtremIO Future Directions”, we really wanted to show without getting into too much details where we are heading in the next few years, think of it as a technical vision for the near term future.

We started by giving a quick overview of the business, we think that Dell EMC XtremIO is the FASTEST growing product ever seen in the storage business, with more than 40% of the total AFA market share since our GA in November 2013 is something that can’t be taken lightly. For me personally, I can say that the journey has been amazing so far, as a relatively young father, I think of the acceleration the product had to go through in such a short time, the market demand is absolutely crazy! More than 3,000 unique customers and over $3Bn in cumulative revenue. From a technical perspective, If I try to explain the success of XtremIO over other AFAs, it really boils down to “purpose built architecture” – something which many other products now claim, when we built XtremIO following four pillars were (and in my opinion still are) mandatory building blocks:

Purpose-Built All Flash Array

We had the luxury of writing the code using a clean slate, that meant SSD’s only in our case and many unique features e.g. XDP, that could never have happened in the old world (you can read more about XDP here: http://www.emc.com/collateral/white-paper/h13036-wp-xtremio-data-protection.pdf)
Inline All The Time Data Services

XtremIO is a CAS (Content Aware Storage) architecture, many customers think of dedupe (for example) as a feature, in the case of XtremIO, it really isn’t. In the old days we described deduplication as a “side effect” of the architecture but since “side effect” is normally thought of as a bad thing, we took that terminology out J. But seriously, what I mean is that we examine the data in real time and give each chunk of data a unique signature and by doing so, when we write the data, if the signature already exists, we simply dedupe it, there is no post process hashing like so many products out there, no “throttling” to the CAS engine etc. This is a killer feature not just because of data savings but rather HOW the data is being written and evenly distributed in real time across all the available nodes and drives, apart from the CAS/Deduplication engine, we of course compress all the data in real time, no knobs needed. We also encrypt the data and everything is done while using thin provisioning so you really only store the unique data you are producing, again, all in real time.
Scale-Out IOPS & Bandwidth

Wow, this one is funny – I remember in the old days spending hours explaining why a Scale-Out architecture is needed and had to debunk competitors claims like “no one needs 1M IOPS” and “no one needs more than a dual-controller architecture”. While I would agree that not everyone needs it, if I look at our install base, so many DO. It’s also not just IOPS, the tightly coupled Scale-Out architecture is what gives you the bandwidth and low latency that your applications need.
CDM (or, unique Snapshots capabilities)

If IDC/Gartner are right by saying that more than 60% of the data stored in your datacenter is actually multiple copies of your data and if your storage array can’t cope with these copies in an efficient way (note that “efficient” is not just capacity but also without performance penalty) then, you’re not as agile as you could be and your TCO goes out of the window – read more about it here:

http://wikibon.com/evolution-of-all-flash-array-architectures/

Luckily, XtremIO snapshots have been specifically designed to be ultra-efficient and as a result we see many customers that are using them not just for the primary data itself but for its copies as well, you can read more about it here https://volumes.blog/2014/05/12/xtremio-redefining-snapshots/

https://www.emc.com/collateral/solution-overview/solution-overview-xtremio-icdm-h14591.pdf

Moving on to our longer term vision for XtremIO, what’s interesting is the that the current XtremIO array, (internally known as X1), was never the endgame, it’s rather a STEP on the road to what we believe customers will need in the coming years. The issue we have faced is that to build what we want means that many new technologies need to become available. We are gradually implementing them all, each cycle, with what’s available from a hardware/software perspective.

The current architecture is doing an amazing job for:

Acceleration (the ability to simply accelerate your workload, e.g. your Database) by moving it to XtremIO.
Providing very good copy data management (iCDM)
Consolidation of large environments into the cluster (think virtualized environments with thousands of VMs residing in the large XtremIO cluster configurations, up to 8 X-Bricks, all 16 storage controllers running in an Active/Active fashion with global dedupe and compression)

We aren’t stopping there, we are going to take scalability to different dimensions, providing far denser configurations, very flexible cluster configurations and…new data services. One of them is the NAS add-on that was the highlight of the session we had at Dell EMC World. Note that it is only one of a number of new data services we will be adding. So why did we mention NAS specifically if we are going to introduce other data services as well? It’s very simple really, this is the first “Dell” and “EMC” World, we wanted to highlight a true UNIFIED ENGINEERED
solution coming from both companies which are one now (Dell EMC).

Before moving to read ahead about the NAS part, we also highlighted other elements of the technical roadmap e.g. the ability to really decouple compute (performance) from capacity, ability to leverage some elements of the Software Defined Storage into the solution and to really optimize the way we move data, not just in the way it lands in the array but rather going sideways to the cloud or other products. Here again, the core CAS architecture comes into play, “getting the architecture right” was our slogan in the early days, you now understand why it was so important to make it right the first time..

Ok, so back to the NAS Support! We gave the first peek into something internally called “Project-F and I must say, the response has been overwhelming so we thought we should share it with you as well – please note that a lot of this is still under wraps and as always, when you deal with roadmap items, the usual disclaimers apply – roadmaps can change without notice etc.

Ok, so what is it?

During 2017, Dell EMC will release an XtremIO-based unified block and file array. By delivering XtremIO NAS and block protocol support on a unified all-flash array, we plan to deliver the transformative power and agility of flash in modern applications to NAS. XtremIO is the most widely adopted All-Flash platform for block workloads. However, we recognized an opportunity to extend our predictable performance, inline data services, and simple scalability to file based workloads. This is the first unified, content-aware, tightly-coupled, scale-out, all flash storage with inline data services to provide consistent and predictable performance for file and block.
The main objective of the Dell/EMC merger is to create an entity that is greater than the sum of its parts. Not just operational synergies, but technical synergies that enable new compelling solutions. The new XtremIO file support feature set is the first of many synergies to come. XtremIO’s NAS capabilities are based on a proven NAS file system (FluidFS) from DELL Technologies and offers:

Full NAS feature-set
Supports Multiple NAS protocols (NFS, SMB/CIFS, HDFS and NDMP)
Over 1M NAS IOPS, predictable sub-millisecond latency
Enterprise-grade scalability, and reliability

A common question that I get is “don’t you already have other NAS solutions?” to me, this question is silly. EMC and now, Dell EMC has always been about a portfolio approach. Lets ignore NAS for a second, wouldn’t this question be applicable to Block based protocols (and products) as well? Of course it will, and as in Block, different products are serving different use cases, for the XtremIO NAS solution, we were looking for a platform that can scale out in a tightly-coupled manner, where metadata is distributed, one that can fit the use cases below. Again, there is nothing wrong with the other approaches, each has their cons/pros for different use cases which is the beauty of the portfolio, we don’t force the use case to one product, we tailor the best product to the specific customer use case. Regardless of the problem you are trying to solve, Dell EMC has a best-of-breed platform that can help. If you want to learn more about the storage categories, I highly encourage you to read chad’s post here http://virtualgeek.typepad.com/virtual_geek/2014/01/understanding-storage-architectures.html

Target Use Cases

This solution targets workloads that require low latency, high performance, and multi-dimensional scalability: Transactional NAS applications are well-suited use cases for the XtremIO NAS capabilities. A few examples are VSI, OLTP and VDI and mixed-workloads including TEST & DEV, and DEVOPS. These workloads will also leverage XtremIO’s data services such as compression and deduplication.

A good rule of thumb is, if you are familiar with the current XtremIO use cases and want them to be applied over NAS/SMB, that is a perfect match.

File Capabilities & Benefits

With its addition of NAS protocol support, XtremIO can deliver all this with its unique scale-out architecture and inline data services for Transactional NAS and block workloads. Storage is no longer the bottleneck, but enables database lifecycle acceleration, workload consolidation, private/hybrid clouds adoption, and Transactional NAS workload optimization. The key features and benefits include:

Unified All-Flash scale-out storage for block and file
• Multi-protocol support of FC, iSCSI, NFS, SMB, FTP, HDFS and NDMP
• Predictable and consistent high performance
• In memory metadata architecture w/ inline data service

Scalable
• Over 1M NAS IOPS w/ sub-millisecond latency
• Billions of files
• Petabytes scalability
• Elasticity – scale-out for block and file, grow when needed, the NAS part can scale from one appliance (2 Active/Active controllers) to 6 appliances (12 Active/Active controllers!)

Simple
• Single unified management (scroll down to see a demo of it)

Resilience
• Built-in replication for file
• Enterprise grade availability
• Proven technology from DELL and EMC

Efficiency
• Inline deduplication and compression
• Efficient virtual copy technology for block and file
• Thin provision support

Both file and block will be using the XtremIO inline data services such as encryption, inline compression and deduplication. In addition, for file workloads native array based replication is available.

Other technical capabilities includes:
• ICAP for antivirus scan
• Virtual copy technology
• Remote replication
• Quotas (on name, directories or users)

Below, you can see a recorded demo on the upcoming HTML5 UI, note that it is different than the web UI tech preview that we introduced in july 2016 ( https://volumes.blog/2016/06/30/xios-4-2-0-is-here-heres-whats-new/ ), yes, as the name tends to suggest, during the “tech preview”, we have been gathering a lot of customers feedback about what they want to see in the future UI of XtremIO (hence the name, tech preview)