Flash storage impact on predictive analytics in under 3 mins. Mike and Dave quickly discuss the growth of these complimentary technologies and why they fit hand in glove. Dave Littman: Hi Dave Littman Truth in IT. Joined today by Mike Matchett. Mike we...
flash, storage, analytics, predictive, matchett
Flash storage impact on predictive analytics in under 3 mins. Mike and Dave quickly discuss the growth of these complimentary technologies and why they fit hand in glove.
Dave Littman: Hi Dave Littman Truth in IT. Joined today by Mike Matchett. Mike welcome.
Mike Matchett: Thanks Dave.
Dave Littman: Mike is Contributing Analyst here at Truth in IT. Today we're talking about flash storage and how it can enhance predictive analytics. Mike these are two both popular topics. How do you think one compliments the other?
Mike Matchett: Well, clearly predictive analytics places a big stress and demand on a storage system. You want to read lots of data, you want to crunch it fast, and you want to make some operational speed decisions. You definitely want a storage system that can deliver, and flash can deliver high speeds. It's not however, the only storage architecture that can deliver massive read IO. For example, if you're running a new workload it might make just as economic sense to have large clusters with lots of spinning discs dollar wise, but there's other workloads like Spark Machine Learning for example. Spark does a lot of work in memory so you want massive amounts of memory, but you also have to stage that data and get it in and out of disc fast to switch between workloads and maybe load different data sets. And there's a place where flash makes a lot of sense. It can depend.
We've also seen a lot of workloads by the way, in this predictive analytic space looking at VGPU's and running and crunching numbers on those. And those are going to need associated fast storage sets, those MBME and flash that are going to be associated with those kinds of architecture and appliances.
Dave Littman: How about the performance of flash affecting or benefiting things like data mining or machine learning?
Mike Matchett: If you're going to look through an awful lot of data, you definitely have to get that data off of its persistent storage into the processing unit, either through a read stream or loaded into memory if you're going to do this in memory processing. And that can be a bottleneck if you're relying on spinning disc, where flash can really help. At the same time, we're not necessarily saying you have to have flash in order to do machine learning. It's really more about, "Can you create a fast enough architecture to support that workloads IO demands?"
Dave Littman: Fabulous, fabulous. All right Mike. Well, thanks for taking the time to speak with us today and help us understand some of the benefits of flash as it pertains to predictive analytics and machine learning. Really appreciate it.
Mike Matchett: You're welcome Dave. Take care.
Dave Littman: Okay, thanks for watching, and check out our other videos here at TruthinIT.com.