设备级分析卸载的实用方法.pdf

1、1|2024 SNIA.All Rights Reserved.A Practical Approach to Device-Level Analytical OffloadsPresented byDonpaul StephensFounder&CEO,AirMettle2|2024 SNIA.All Rights Reserved.“Our scientific large-scale simulations can generate hundreds of petabytes of highly dimensional floating-point data.But the data a

2、ssociated with a scientific feature of interest can be orders of magnitude smaller than the written data,so a key challenge is quickly and efficiently finding whats relevant in this sea of data.To optimize this process,weve been drawn towards computational storage processing data in-place and near s

3、torage to eliminate unnecessary data movement while maintaining parallelism and adequate data protection.”Gary Grider,High Performance Computing division leader3|2024 SNIA.All Rights Reserved.Computational Storage?Erik RiedelsPh.D.Dissertation 1999:Active Disks Remote Execution for Network-Attached

4、StorageWhy didnt it take off?How we can put it to work.4|2024 SNIA.All Rights Reserved.Why is Big Data so SLOW?5|2024 SNIA.All Rights Reserved.Why is Big Data so SLOW?Erasure Coding:In part:How storage protects dataIn part:How storage protects data6|2024 SNIA.All Rights Reserved.Data reliably placed

5、 in storage:First 4 devices shownSimple Table:#2#4Supports data protection algorithms designed for HDD!Supports data protection algorithms designed for HDD!Bytes of data divided evenly across SSDs!Data protection and streaming performance!#1#37|2024 SNIA.All Rights Reserved.Byte-level partitions pre

6、cludes analyticsSimple Table:Bytes of data divided evenly across SSDs!Data protection and streaming performance!#1#3#2#4HDDHDD-centric RAID/Erasure Coding prevent incentric RAID/Erasure Coding prevent in-storage analyticsstorage analytics8|2024 SNIA.All Rights Reserved.Challenge:Array based data for