File this: XtremIO to fling forth filer functionality

NASty: Gen-two gear will add transactional file operations on block base

Analysis XtremIO's rush to revenue glory is going to get accelerated with coming file data services added to its block base.

The story starts when Israeli startup XtremIO started developing an all-flash array, block-access product in 2009. The product's capabilities included global in-line deduplication, snapshotting, and InfiniBand support.

Step 2 was in 2010, when Dell spent $12m and acquired the assets of Exanet, an Israeli startup developing scale-out NAS technology.

It set up an Exanet tech-based R&D center in Israel and out of that came its Fluid FS (filesystem), which was then implemented on its PS (EqualLogic) and SC (Compellent Storage) arrays. This gave Dell unified file and block access and obviated the need for it to have a dedicated filer or scale-out file product.

We move forward two years to when EMC bought XtremIO for $430m in May, 2012. The first generation array was shipped in 2014 and, by late 2016, has achieved astounding success. While that was going on, the XtremIO engineers started working on generation 2, the X2 product generation, which, it was revealed at Dell EMC World, will have file services functionality and provide unified file and block access.

This on top of basic hardware development such as denser, higher-capacity configurations. The File (NAS) part will include NFS and SMB access protocols.

It turns out that this had been developed with help from the Exanet team in Dell's Fluid FS center in Israel and, with Dell's purchase of EMC, that team became an in-house resource for XtremIO instead of a quasi skunkworks-like project, Project-F.

The file services addition will provide, as it were, a file section on the X-Bricks, and takes advantage of existing XtremIO features such as metadata generation and snapshotting. There is a copy (file) data management service coming and that uses the snapshotting. There will be an ability to examine big data sets and analyse them quickly.

In fact, the system will be fast, with 0.2 to 0.5 millisecond latencies, and will be targeted at transactional file workloads. That provides the positioning versus Isilon AF (all-flash) which is for bulk, non-transactional file storage. We're told creating a volume is instantaneous, snapshotting at the file level is instantaneous, and 1,000 virtual machines could be snapshotted in a couple of minutes.

XtremIO CTO Itzik Reich said: "XtremIO is transactional workloads first and other use cases second. We will enter leading platform for mixed block/file workloads that need scale-out. Isilon is HPC first; different horses for different courses."

Project-F will also have a bunch of other features and gen 2 XtremIO could be shipping in mid-to-late 2017.

We also understand a third generation is in the XtremIO folks' eyes, perhaps with extensibility to the cloud.

What we can see is that, as flash becomes cheaper, transaction-focussed filers will all move to flash. Deduplication and compression will enlarge their effective capacity and existing, disk-based assumptions about how and where filers should be used will be overturned.

We may well, for example, see NVMe over fabrics block array access being looked at to see if the same approach could be used to speed network filer access.

What's apparent is that XtremIO has become a flash-powered platform, potentially a long-lived platform, for the development of integrated data services. The rumors of it being parked in the end-of-life product care home earlier this year were baseless. ®