Small knowledge, big challenge! This article is participating in the creation activity of “Essential Tips for Programmers”.

Fast reconstruction: The CHUNK of each physical disk and the CHUNK of multiple disks form a RAID group. If a single disk fails, more disks participate in the reconstruction than the traditional method, which greatly improves the reconstruction speed. The reconstruction can be completed in 30 minutes per TB. The following uses RAID5 as an example. If disk 1 is damaged, data in CKG0 and CKG1 is damaged. The system selects CHUNK for reconstruction. As shown in the following figure, CHUNK 14 and CHUNK 16 are damaged. Free chunks in the POOL will be randomly selected for reconstruction (the yellow squares in Figure 3-29). The randomly selected chunks will be distributed on different disks as much as possible.

The bottleneck of traditional disk reconstruction lies in the target disk (hot spare disk). After all member disks read all data, the reconstructed data is written to the target disk. Therefore, the write bandwidth determines the reconstruction speed. After RAID2.0+ block virtualization, there are two improvements:

  1. Multiple target disks, such as two target disks, shorten the reconstruction time to 9 hours. When the number of chunks and member disks increases, the target disk can reach the number of member disks, so the reconstruction speed will be greatly improved.
  2. Data reconstruction by CHUNK: When the number of chunks allocated to the faulty disk is small, the data to be reconstructed decreases greatly, and the reconstruction speed increases. RAID2.0+ reconstruction can be completed in 30 minutes per TB. The shorter reconstruction time greatly reduces the probability of dual-disk failure.

Thumb up ah