PURPLELEC BLOG
Let Technology And Passion Collide To Generate More Inspiration. For questions about the blog, please contact the administrators or leave a message.
By PURPLELEC | 01 November 2023 | 0 Comments

How to fully utilize the performance of chip-level SSD

  First, let’s take a look at how traditional HDDs are used:
  1. The protocols generally use SAS and SATA interfaces;
  2. Linux IO scheduling requires the elevator algorithm to rearrange IO to optimize the path of the magnetic head;
  3. Enterprise-level storage usually uses Raid cards for data protection.
  In terms of interface protocols, with the invention of chip-level SSD, the NVMe protocol came into being. Compared with the single queue mechanism of SAS and SATA, NVMe can have up to 65535 queues and directly uses the PCIe interface, eliminating link and protocol bottlenecks.
NVMe SSD
  In terms of control card ecology, major manufacturers have also launched their own NVMe control card chips, and the technology is also very mature.
  Corresponding optimizations have also been made in the Linux driver and IO protocol stack. The NVMe driver can directly bypass the traditional scheduling layer specially designed for HDD, greatly shortening the processing path.
  So far, in order to give full play to the performance of chip-level SSDs, the first two of the three traditional HDD problems mentioned above have been solved. However, in the enterprise market, there has never been a good solution for NVMe-based Raid.
  The most widely used Raid5/Raid6 data protection mechanism (N+1, N+2) in traditional enterprises usually stripes and shards the data, then calculates the redundant ParityCode (parity check code), and stores the data in For multiple hard drives, writing new data is usually a "read-modify" mechanism.
  This mechanism itself has become a performance bottleneck, and "reading, modifying, and writing" has a great loss on the service life of chip-level SSD. In addition, because the NVMe protocol places the control card inside the NVMe disk, IO is completed by the DMA module inside the NVMe disk, which brings greater difficulties to the design of Raid cards based on NVMe.
  Currently, there are very few solutions for this type of Raid control card on the market, and they cannot take advantage of NVMe in terms of performance, so they are not widely used.
  Based on the current situation, many enterprise-level storage solutions still use SAS/SATA chip-level SSDs plus traditional Raid cards. This method will cause the two problems that have been solved previously, and the performance of chip-level SSDs will not be fully achieved. play.

Leave a Reply

Your email address will not be published.Required fields are marked. *
Name
E-mail
Content
Verification code
ver_code
TAGS