Welcome to this exciting Ask the Expert event.
On March 5 we announce the exciting new server flash products. EMC executives, customers, and analysts will discuss these game-changing technologies and EMC’s vision for software-defined server flash on a Live web cast.
Following that this Ask the Expert event will be open and you can continue this conversation with our hosts about the details of the FLASH.NEXT announcement.
This discussion begins on Tuesday, March 5th and concludes on March 18th. Get ready by bookmarking this page or signing up for e-mail notifications.
Avi (Avishek Kumar) is a Senior Product Manager in EMC’s Flash Products Division, focusing on EMC XtremSF and EMC XtremSW Cache. Prior to joining this team, Avishek was a Corporate Systems Engineer in the CLARiiON Hardware Engineering and Performance Engineering groups. Avishek holds an MS in Computer Engineering from the University of Massachusetts Dartmouth.
Elad_Horn is a Senior Product Manager in EMC’s Flash Products Division, responsible for EMC XtremSW Cache. Prior to joining EMC, Elad managed database security products at Imperva and held various engineering and positions at Marvell and Intel. Elad holds a B.Sc. in Electrical Engineering from the Technion-Israel Institute of Technology and an MBA from IDC Israel.
Ok guys, although I was in the virtual announcement yesterday I had to leave the screen frequently, so in the end all I noticed were commercial words like "1 million IOps", "linear scalable" and stuff, but I need to dig into pdfs to make up my mind and come up with questions.
Can you provide me with some MUST HAVE notes, ppt or white papers so I can actually ask some questions? I won't bother you with questions about pricing, although some sort of a direction or hint about cost per GB is interesting.
I have a question about the "inline data reduction" in the XtremeIO: is this 100% true dedupe? Or a sort of dedupe (best effort) that works well if workload is low and less dedupe when the workload is high? I almost can't imagine that the inline dedupe really can keep up with very extreme workloads, or can it?
RRR - inline deduplication in XtremIO is "truly" inline and always ON. What this means is that this feature (along with other features like thin provisioning) are always ON and you do not have an option to turn it OFF. In fact, the more deduplication the array does, the more performance benefit you see from an application point of view.
This happens because deduplication has been baked in the software design of the array and is not a "add-on" feature that would potentially impact the array's performance.
Avishek already responded but I thought I'd add some more. XtremIO is the ONLY array in the market that has truly inline data reduction. Other vendors who claim to have it can only perform some operations inline (as you pointed out) when the array is not very busy, and will switch off data reduction entirely when the array is processing moderate to heavy I/O loads. Others don't have any kind of data reduction and others have post-processing. This leads to very unpredictable performance. Why are you buying a flash array? For high levels of predictable performance!
Furthermore, inline data reduction in combination with other capabilities (for example, VMware VAAI integration and memory-based metadata management) allows the XtremIO array to do some very cool things that nobody else can do. For example, if you clone a virtual machine in vCenter, no data needs to be copied. It's already on the XtremIO array, so a few pointer references created in memory are all that's needed. Thus VMware provisioning tasks are extraordinarily accelerated - cloning VMs takes only a few seconds and doesn't consume and back-end IOPS on the SSDs in the XtremIO array.
Here's a demo of this amazing capability in action: XtremIO Data Center Scale Virtualization Demonstration - YouTube
Other facts about data reduction on XtremIO:
Keep the great questions coming!