Presentation on theme: "1 RAID Overview 91.560. 2 n Computing speeds double every 3 years n Disk speeds cant keep up n Data needs higher MTBF than any component in system n IO."— Presentation transcript:
2 n Computing speeds double every 3 years n Disk speeds cant keep up n Data needs higher MTBF than any component in system n IO Performance and Availability Issues! The Motivation for RAID
3 n PERFORMANCE u Parallelism u Load Balancing n AVAILABILITY u Redundancy: Mirroring, or Striping with Parity n FLEXIBILITY u Selectable Performance/Availability/Cost RAID to the Rescue!
4 n Redundant Array of Independent Disks (a.k.a. Disk Array) u Multiple drives, single host disk unit n Provides opportunity to increase: u Performance via Parallelism u Data Availability via Redundancy n Cheap cost via commodity disks What is RAID Technology?
5 Chunk size tuneable for BW vs Thruput tradeoffs u Large Chunk High Throughput (IO/sec) u Small Chunk High Bandwidth (MB/sec) Chunk 1 Parity Chunk 11Chunk 16Chunk 21Chunk 5Chunk 10Chunk 15Chunk 20Chunk 2Chunk 7Chunk 17Chunk 22Chunk 3Chunk 8Chunk 13Chunk 23 Drive 5 Chunk 12Chunk 0Chunk 18Chunk 9Chunk 14Chunk 19Chunk 24Chunk 4Chunk 6 Drive 4 Drive 3 Drive 2 Drive 1 Performance: Disk Striping
6 Data Stream, OR IO Stream, can be multiplexed across multiple disks, depending on BW vs Thruput. Parity data is also stored on disk. Same data written to both disks. Striping with Parity Single Host Unit > Add XORd parity for increased availability. Mirroring Availability: Redundancy
7 n Parity = XOR of data from every disk in the RAID unit DATA 0 1 0 0 parity - Any single disks data can be recovered by XORing the data of the surviving disks. Parity Redundancy
8 n Many to choose from n Each offers unique tradeoffs u Performance u Availability u Costs n We offer levels 0, 1, 3, 5, 10 RAID Levels
9 Disk Striping with No Redundancy Data Stream, OR IO Stream, can be Striped across multiple disks (BW vs Thruput). High Performance; Low Availability Data Striped on Multiple Disks Multi-threaded Access RAID 0
11 Disk Mirroring n Single-disk Performance; Expensive Availability n Data 100% duplicated across both spindles. n Single-threaded access SAME data written to BOTH disks -- no segmenting. Single Host Unit RAID 1
12 Striped Mirrors Data Stream, OR IO Stream, can be Striped across multiple disks (BW vs Thruput). Highest Performance; Most Expensive Availability Multi-threaded Access Single Host Unit RAID 1/0
13 Disk Striping with dedicated parity drive Data Stream is Striped across N-1 disks for high bandwidth. High BW Performance; Cheap Availability Sector-granular data striping Single-threaded Access XOR Parity Data RAID 3
14 Chunk size = single sector (pure RAID 3 would be single byte) n All parity data on same spindle Chunk 1 Parity Chunk 9Chunk 13Chunk 17Chunk 4Chunk 8Chunk 12Chunk 16Chunk 2Chunk 6Chunk 14Chunk 18Chunk 3Chunk 7Chunk 11Chunk 19 Drive 5 Chunk 10Chunk 0Chunk 15Parity Chunk 5 Drive 4 Drive 3 Drive 2 Drive 1 RAID 3 Striping
15 Disk Striping with rotating parity drive IO Stream is Striped across N-1 disks for high IOs per second (thruput). n High Read Performance, expensive Write performance; Cheap Availability n Tuneable Stripe granularity n Optimized for multi-thread access XOR Parity Data RAID 5
16 Chunk size is tuned such that typical IO aligns on single disk. n Parity rotates amongst disks to avoid write bottleneck Chunk 1 Parity Chunk 9Chunk 16ParityChunk 4Chunk 8Chunk 12ParityChunk 2Chunk 6Chunk 13Chunk 17Chunk 3Chunk 10Chunk 18Parity Drive 5 Chunk 0Chunk 14Chunk 7Chunk 11Chunk 15Chunk 19ParityChunk 5 Drive 4 Drive 3 Drive 2 Drive 1 RAID 5 Striping
17 1. Read old data. OldNew 2. Write new data Old P. 3. XOR old and new data to create Partial Product. 4. Read old parity data. 5. Xor old parity with partial product, writing out result as new parity. P. } XOR } New P. Chunk 1Chunk 2Chunk 3 Drive 5 Chunk 0Parity Drive 4 Drive 3 Drive 2 Drive 1 RAID 5 - Write Operation
18 u RAID 0 - Data striping, Non-redundant. u High Performance, Low Availability u RAID 1 - Mirroring u Moderate Performance, Expensive High Availability u RAID 1/0 - Striping and Mirroring u High Performance, Expensively High Availability u RAID 3 - Striping, single parity disk. u High Bandwidth Performance, Cheap Availability u RAID 5 - Striping, rotating parity disk. u High Thruput Performance, Cheap Availability RAID Level Review
19 n Increasing performance gap between CPU and IO n Data availability a priority n RAID meets the IO challenge: u Performance via parallelism u Data Availability via redundancy u Flexibility via multiple RAID levels, each offer unique performance/availability/cost tradeoffs Summary
Your consent to our cookies if you continue to use this website.