This post is also available in: Italian
The first on the second day was Exablox, and probably their presentation was one of the best, both for the contents and the type of presentation.
Exablox is a recent company founded on 2010 by lot of veteran people from different storage (but on only) reality like Data domain, Veritas, EMC, NetApp, Virident, FusionIO, …
Lot of experience to build something different for a specific market share: their idea and mission was to build an enterprise class NAS storage for the mid-size market. A simple and clear idea with a pragmatic approach and some unique idea (there are 15 patent pending).
Was really interesting see how Exablox develops enterprise class storage product but validates the model with mid-market segment. Building something with enterprise concepts and then adapt to the mid-size is possible (not necessary simple, but possible), adapt a mid-size solution to become an enterprise one, could be not necessary possible (probably you need to re-engineering the product).
And the enterprise approach is based on the cloud scale model (or web scale): a powerful scale-out model with focus on the simplicity (easy to configure in only 5 minutes and easy to expand with something similar to a plug and play).
Scalability can start from one OneBlox node (without node availability) or using a ring (like a cluster) from 3 node, scaling from a few TBs to nearly 300TBs. By Adding Oneblox node the ring scale in capacity, performance and availability. But each single node can scale in capacity by adding more disks (or replace exising disks with more bigger).
Each node is a 2U rack block with 8 slots for hard disks (3.5″ size)… and here another unique (in this segment) approach: you buy the node without the spindles! Sounds really strange but they have this “Bring Your Own Disk” approach, to free the customer on which type of disk he prefer, included consumers hard disks. This is pretty uncommon in storage world where several vendors for to use “validated” hard disks that may costs much more than the same disk acquired directly by Seagate or Western Digital (the two main productor of spindles).
But also the hardware is quite different: instead of using “commodity” hardware they have build a specific box with 4 Ethernet (1 Gbps) interfaces, a single power supply (on this choice maybe I don’t totally agree, considering that several customers starts with a single node), 8 HD slot (really uncommon in how HD could be loaded without any rail that must be manually mounted) and a big LCD panel with several information (that also cover the HD slots when this panel is closed).
All is based on a MIPS architecture, so not the common x86 platform, but with a customized Linux operating system that could be easy adapted to different platforms. Considering that one of the reasons of this choice was the power-efficiency could be interesting see if they will considering the ARM platform for the future (also in order to build a smallest box for the SMB).
As written the disks slots are simple 3.5 format, in order to use the larges HD that are available on the market (actually 6TB)… this also mean that this product position is not for production storage (like virtualization or business critical application), but for secondary storage (like for example archiving, backup target, …) and this kind of position explain why there isn’t (yet) a node version with full-flash or 2.5″ slots for some kind of hybrid storage (actually they are building capacity box, but could be interesting see if and when they will build also performance box).
Also the choice of the type of protocols that are exposed is quite pragmatic: in order to keep it simple and avoid LUN, RAID, SAN concepts this is just (and only) a NAS storage, but note that internally is an object storage (with the capability to handle 1,3 Billion objects per node). There is a distributed filesystem that is exposed with SMB (1.0, 2.0, 2.1, 3.0) or also NFS (although is not mentioned on the tech specs).
Another unique aspect is the OneSystem management interface: it’s a (public) cloud management platform (provided still by Exablox) and is the only way to manage their storage. Seems a strange choice but in this was you have an high-availability management service that is web-oriented, geographical and multi-tenant (really interesting for example for the resellers to manage, if needed, the storage of different customers).
Of couse there are also some disadvantages in this kind of choice: not all customers really like the idea to have storage that are “connected” to the Internet (but for example for upgrade this could be useful, in this case upgrade is just a click on a web interface) and this kind of apprach limit the integration with other management tools (like System Center) or the integration with hypervisors (implementing vSphere VASA will be not easy). From the other side there are also a lot of advantages, like in the telemetry or in the alarms management.
The approach of data protection and availability is also interesting: with its real-time replication, OneBlox is able to transparently replicate and redistribute the data across multiple drives and OneBlox to protect against two drive or OneBlox failures. Additionally, with its advanced architecture, information is protected against an entire ring failure when replication over wide-area distances is implemented.
Also OneBlox continuous data protection by taking continuous and unlimited snapshots of information written to OneBlox, every file can be easily recovered in the event of data corruption, deletion, or other errors. In fact, end users can recover their own information, simply by navigating through Mac Finder or Windows Explorer (this is instegrated with Microsoft Shadow Copy for Shared Volumes in order to see previous version directly in the Windows Explorer). Additionally, OneBlox protects against data loss by encrypting the data on each drive with AES256, making data theft from lost or stolen drives a thing of the past.
So several differentiators for this kind of product and for sure a clear vision and a clear target. And will be interesting see which kind of evolution can have in the future (potentially target also the production storage or also the SMB segment).
Considering that a user case is backup archiving, would also be nice see a future integration with some backup vendor, starting with Veeam where already exist (in v8 of their product) some specific “optimized” targets.
For more information see also those other posts/articles:
- Hunt the RAID killer: Exablox builds object storage that isn’t
- Silicon Valley 2014 : Exablox mise sur la simplicité pour le stockage out-scale (French)
- IT Press Tour : Exablox vise l’Europe pour sa solution de stockage Cloud objet (French)
Disclaimer: I’ve been invited to this event by Condor Consulting Group and they have paid for accommodation and travels, but I’m not compensated for my time and I’m not obliged to blog. Furthermore, the content is not reviewed, approved or published by any other person than me.