美文网首页
Database Hardware Discussion

Database Hardware Discussion

作者: KevinWu12 | 来源:发表于2018-06-13 19:47 被阅读0次
    Effective memory access time:

    EA = HC + (1-H)S
    Where H = Hit ratio
    Hit ratio = references satisfied by cache / total references
    C = cache access time; S = memory access time

    Effective disk buffer cache access time:

    EA = HC + (1-H)S
    Where H=Hit ratio C = cache access time; S = memory access time
    Disk access time = seek time (tracks)+ rotational time (section) + transferlength / bandwidth

    Storage System

    • MTBF (Mean Time Between Failure):
      Describes the expected time between two failures for a repairable system

      Time between failures.svg
      • Time between failures = {down time - up time}
      • C12pture.PNG
    • Mean Time to Repair
      (Total down time) / (number of breakdowns)

    • Mean Time to Fail:
      For the product that cannot be repaired.

    RAID (Redundant Array of Independent)

    • RAID 0

    ure.PNG
    • Block Level Striping(4k or 8k bytes of storage)

    • Provides balanced I/O of disk, throughput approximately doubles

    • Any one of the disk failure could cause the MTTF reduces by a factor of 2

    • RAID 1

    Capddure.PNG
    • Block Level Mirroring

    • Provides higher read throughput but lower write throughput

    • MTTF increases substantially (MTTF^2)

    • RAID 2

    ggture.PNG
    • Bit Level Mirroring(4k or 8k bytes of storage)

    • Provides higher transfer rate

    • MTTF increases substantially (MTTF^2)

    • RAID 3

    Cadfure.PNG
    • Byte Level Striping (Parity)

    • Provides higher transfer rate as RAID 0

    • MTTF increases substantially (1/3 of RAID 1 = MTTF^2/3)

    • P0 is parity (exclusive-or) for b0 and b1. Knowing any two of the disk data could recover the third data.

    • RAID 4

    Cadture.PNG
    • Block Level Striping (Parity)

    • Similar to RAID 3

    • Disk 3 has a lot more stress because it has more writes than either disk 1 and disk 2 as the parity has to be updated on every write

    • RAID 5

    Capture.PNG
    • Block Level Striping (Parity)

    • Single disk failure tolerant array

    • No dedicated disk for parity blocks, parity blocks are also striped

    • MTTF is same as RAID 3


      Capture.PNG
    • RAID 6

    • At least 5 disks including two parity blocks used

    • Reliability is of the order of MTTF^3/10

    • p0 and p1 are the parity blocks for blocks A0,A1,A2

    相关文章

      网友评论

          本文标题:Database Hardware Discussion

          本文链接:https://www.haomeiwen.com/subject/ceifeftx.html