Note: I’ll be putting up a separate DSSD post later tonight…most info I’ve seen on it so far was in Area 52.
Got here a bit late…during VNXe demo. First thing I heard was Chad Sakac joking about his height…the VNXe is small but powerful. 😉
Summary – this is the most comprehensive Area 52 blog post writeup I’ve seen…lots of good info and screen shots below.
Teaser = XtremIO pushing 260k IOPs at sub-1 ms latency.
More after the jump…
Demo of VNXe 3200 creating a tiered storage pool with FAST Cache enabled – only 2U for 25 drives.
It’s cool to talk about things that are big – but also cool to talk about things that are small and simple.
What if multiple VNX’s & VNXe’s? That’s where we love Unisphere Central – demo showing digging into storage pools, see utilization rates.
Now showing EMC Storage Integrator for x64 for Exchange – does all the steps to provision Exchange very easily.
Project Liberty – virtual VNX – perfect for ROBO environments, test environments.
Now talking platforms – type 2 platforms with tightly coupled scale out and shared memory. This almost feels like the spoken version of Chad’s storage architecture post.
Stephen Manley coming on stage – talking about data protection. Application folks get nervous about data protection. Showing a command prompt to kick off an Oracle RMAN backup. Showing this live – see screenshot. Backup process is nuts – multiple data flows…call it data trombones.
Showing the first public release of ProtectPoint – and how it minimizes storage impact during backups. Clear benefits in CPU, storage impact, network throughput.
Now going into XtremIO – Building an All Flash Array is different than normal disk arrays. XtremIO uses RDMA to share metadata between X-Bricks. XtremIO is almost the only flash array that’s all type 2 – most competitors are type 1.
Pablo coming on stage now – handles the VMWare Hands-On Labs. VMware HoL stats on XtremIO – 150-250k steady state, 500k IOPs boot storm at 1-2 ms latency!
Showing a live performance test using SLOB – driving over 80k+ IOPs at sub-1 ms latency.
Moving into cloning operations off XtremIO – cloning LUNs on XtremIO and pulling into Oracle ASM. Now we can drive multiple databases at 260k IOPs with sub-1 ms latency. Very cool live demo.
Even better, pushing this performance can be done without turning off any features. EMC is releasing the full details on the test harness for anyone (even competitors) to test against.
Matt Cowger coming out for a ScaleIO demo – more good live stuff. Starting with cloud provider “A”having a couple nodes in ScaleIO – then added a couple nodes from vCloud Hybrid Services and ScaleIO rebalanced. Largest that Cowger has deployed is 1,000 nodes for $76 on “A”.
Video testimonial from Xbox Video Services – all built on Isilon. Storage always growing.
Isilon live demo – root user has a single character password. Now showing a multi-site Isilon cluster across multiple sites with Isilon grids at each site. Next step = multi-site across Isilon grid and at public cloud provider (MS Azure shown.) This is a BETA product. Have options around encrypting the data or compressing the data based on which public cloud you’re using. Announcing Isilon virtual nodes – showing it live as well.
Can get virtual Isilon today – http://www.emc.com/getisilon
Moving into type 4 platforms – Ron Redman from Assure360 coming on stage for a ViPR demo.
Showing uploads into the cloud via ViPR – kicker was that @VirtualSwede used Chad’s AD account, Chad’s VMware lab account, and Chad’s Facebook account.
Now Bill Moore with DSSD on stage with Chad – talking about a “5th phylum” of storage. There are growing “workload innovation” pressures.
DSSD looked at how application developers work – they pick a framework (HANA, etc.) and then write their application with it since it solves their need as a developer. DSSD said…since we’re building our own hardware, let’s scrap all the layers shown below.
Now we’re all CPU’s – or at least thinking in microseconds and nanoseconds. Slides below illustrate how long a CPU has to wait in human terms to get responses.
Bill – “The fastest software in the planet is the software you never have to run.” Chad – “It also has the fewest bugs.”
Inside the mystery box is the world’s largest PCIe fabric – Flash top of rack pool, No need for SCSI, No need for file handles.
This scraps a ton of traditional storage stuff – lets the lowest layers talk directly to the applications. This is really a new phylum of storage. Slides below to illustrate…
So…we now have a ton of platforms, right? Crazy complex, right? This is why ViPR exists – SDS control abstractions – to make it easy to the admin and applications.
Area 52 done – Area 53 tomorrow.