July 8th 2014. EMC MegaLaunch 4. Theme: Redefine Possible (or #RedefinePossible on Twitter). What previously was impossible, now is possible! Catchy theme and something that we’ve seen in IT for a number of times now. For example: In the 1990’s, who would have thought it was possible to migrate a server from one datacenter to another, possibly a couple of miles away, without downtime, in a couple of seconds?! Doing things fundamentally different, better: that’s the goal we’re always trying to achieve. So how can we apply this to Isilon?
What’s the reason behind this change?
Over the last couple of years the amount and percentage of unstructured data has been growing rapidly. 2013 saw 37EB of data of which 67% was unstructured data (file as opposed to database). According to IDC, in 2017 this will be close to 133EB of data with 80% being unstructured data. Think pictures being taken by the general public (unaware of the demand this places on IT backbone systems), music and movies being created, surveillance video, airplane engine statistics… data is growing and it’s growing rapidly!
If you upload all this data to a system, you might want to run some analytics on it. You don’t want to have to copy all that data to a different location just so you can run some kind of Hadoop workload on it. You want that data accessible by all kind of protocols, as fast as you can create it. So instead of creating (in effect) silo’s of data, you want one big “data lake” (as EMC puts it), accessible by all protocols. EMC Isilon will help you accomplish this: a single, scalable, high performance file system accessible through any protocol you like: SMB, NFS, FTP, NDMP, HDFS…
It’s all about performance…
EMC promises 2x performance (in testing closer to 2,3x performance but marketing doesn’t like comma’s!) for 50% of the $/IOp. It does this with new nodes running on OneFS 7.1.1: the S210 for high transaction workloads and X410 for high throughput workloads. Take the S210 for example: from 12.000 IOps in a single S200 node, to over 26.000 IOps in a S210.
No doubt a big part of this performance is coming from a new feature called SmartFlash. SmartFlash has been created because EMC was running into a bit of a problem. With a ton of spindles and TB’s of HDD data and only a finite amount of DRAM (up to 37TB in a grid) there’s quite a big gap between extreme performance and high capacity disks which currently scales up to 20PB. If only there was a way to bridge that gap…
Enter flash. SmartFlash allows you to add up to 1PB of globally coherent flash cache to the cluster. 30x more flash compared to DRAM, at 1/10th of the price of DRAM. Ideal to bridge the gap between DRAM performance and spindle capacity. And 100% of it gets used; no wasting precious SSD resources here! Pretty much like the VNX family does with FastCache…
Multi-Protocol support and vBlocks
Accessing that data lake is already possible using SMB, NFS, FTP, HDFS, NDMP… you name it. With the newest iteration of OneFS, SMB 3.0 (multichannel!), HDFS 2.3 and OpenStack SWIFT join the party. SMB 3.0 multichannel allows your server to use two NICs at the same time in sending and receiving data, increasing performance and availability at the same time. HDFS 2.3 offers in-place analytics for faster time to results. Swift in turn is a native, open source object storage protocol for connecting your Isilon to the cloud.
Now, we already know vBlocks. Compute, networking and storage, pre-packaged in one or multiple racks, ready for deployment at the customer site. Add power and cooling and external network and you’re pretty much ready to run. Available with VMAX and VNX. And now also with Isilon and XtremIO! XtremIO is the all flash, block array. Isilon is a scale-out NAS. Package the two together and you get the ideal combo: XtremIO for VDI, Isilon for files in homefolders etc.
I can’t wait to get my hands on these new features and see how they work out in the real world. Some of these features may indeed redefine possible for a specific customer: just that additional bit of performance or that protocol that they’ve been waiting for to fully optimize their Isilon usage. And more is to come…
