DSSD has control module which manage Garbase Collection,Wear Leveling for entire DSSD so that each
NAND element can be last longer than other flash products (They have to manage on disk level which caused bottleneck)
Am I answering your question ?
Thanks for the explanation! I just come up with idea that GPU and DSSD can be used together.
Pivotal had MATLAB kmeans test result that took 5days for process, if you can use DSSD for this
process,it will make faster ....
I agree with you that GPU computing and DSSD will mix well together .I know some customer had tried .
Processing of clustering is continuous co-reading,reasoning,and writing.The more clustering has been done , the more data creates...This is the ideal for DSSD. DSSD is built for processing huge amount of data.
Thanks for the reply ! you made good point ! Ryo.T !
Actually , there is a plan for DSSD and GPU ....Serious GPU user wants to use 3-4 GPU cards and end up
fight over PCI slot with DSSD !
It is interesting that mixing GPGPU and DSSD ...GPU has high-bandwidth, energy-efficient interconnect called NVlink ,
Imagine mixed with DSSD will create massive system !
( Mind you that NVLink only supports IBM power-processor ,,,,not DSSD yet)
OH ,,,there is technology for connecting GPU together ! NVlink looks faster than DSSD PCI gen3 ! ..
Super first Frontend with NVlink and Super first Backend with DSSD .... Gee this will be a beast wouldn't it ?Thank you for the great information !