Article count:640 Read by:1001592

Account Entry

Hybrid Memory NVDIMM - Keep Data on the Chip

Latest update time:2021-09-05 16:03
    Reads:

As we all know, there are three factors driving the development of big data: capacity, variability, and velocity. With the advent of persistent storage, some trendy users are starting to "copy" high-performance computing (HPC) systems to increase speed: burst buffering is back!

"How to store data quickly" is an important challenge faced by big data and HPC systems, because these systems usually need to spend a considerable amount of time writing data to disk or SSD (solid state drive) to save the data.

In the past, HPC systems used massive parallel storage systems. Although they can handle a large number of read and write operations in parallel, it takes a certain amount of time to write any data. Similarly, the way big data systems are accelerated is the same, that is, dividing data blocks across many disks, and it still takes a certain amount of time to write.

The burst buffer is a low-latency persistent storage layer inserted between the computing system and the huge parallel backend storage system. Its working principle is very simple: the data is first stored in the burst buffer, which is equivalent to a temporary storage area, and then the data is transparently transferred to the slower backend storage subsystem.

The earliest burst buffers were built using SSDs. Today, with the advent of NVDIMM-N, system designers are boldly adopting NVDIMM-N as a standard component supported by major OEMs. However, a problem that comes with it is capacity: for applications such as storing data transmitted at a high rate, the existing 8GB and 16GB capacities are fully sufficient because this is just a temporary storage area.

Experienced users may have heard of NVDIMMs. Micron Technology's NVDIMMs are based on the -N standard, which means that they can perform DRAM chip writes. Only when the power is off will the controller store the data to the NAND chip (used in SSDs). So Micron Technology's solution has two major advantages:

First, performance has always improved with increasing DRAM speed.

Second, with DRAM, the time required to read and write is equal.

Other solutions use NAND chips. The ratio of write to read time of NAND chips is close to 20:1. For our application, we want to minimize the time taken to write data, so the above advantages of Micron NVDIMM are crucial.

NVDIMMs can be used in storage or memory. System designers can use them as block storage devices without changing them for specific applications. To use them as memory devices for further acceleration, experienced users need to modify their applications using the pmem library.

The birth of NADIVM combines the advantages of both DRAM and NAND, allowing the storage to have the same read and write speed as DRAM, while protecting data security during power outages, thus keeping the data firmly in the "core".

Featured Posts


Latest articlesabout

 
EEWorld WeChat Subscription

 
EEWorld WeChat Service Number

 
AutoDevelopers

About Us About Us Service Contact us Device Index Site Map Latest Updates Mobile Version

Site Related: TI Training

Room 1530, Zhongguancun MOOC Times Building,Block B, 18 Zhongguancun Street, Haidian District,Beijing, China Tel:(010)82350740 Postcode:100190

EEWORLD all rights reserved 京B2-20211791 京ICP备10001474号-1 电信业务审批[2006]字第258号函 京公网安备 11010802033920号 Copyright © 2005-2021 EEWORLD.com.cn, Inc. All rights reserved