1. CS5226 2002 Hardware Tuning Xiaofang Zhou School of Computing, NUS Office: S16-08-20 Email: zhouxf@comp.nus.edu.sg URL: www.itee.uq.edu.au/~zxf

2. 2 Outline Part 1: Tuning the storage subsystem RAID storage system Choosing a proper RAID level Part 2: Enhancing the hardware configuration

3. 3 Modern Storage Subsystem More than just a disk Disks, or disk arrays Connections between disks and processors Software to manage and config. devices A logical volume for multiple devices A file system to manage data layout

4. 4 RAID Storage System Redundant Array of Inexpensive Disks Combine multiple small, inexpensive disk drives into a group to yield performance exceeding that of one large, more expensive drive Appear to the computer as a single virtual drive Support fault-tolerant by redundantly storing information in various ways

5. 5 Data Striping Disk 2Disk 3Disk 4Disk 5Disk 6 Disk 1 1234567891011… File blocks (e.g., 8KB per block) Stripe unit: blocks 1-6, 7-12, …

6. 6 Parity Check - Classical An extra bit added to a byte to reveal errors in storage or transmission Even (odd) parity means that the parity bit is set so that there are an even (odd) number of one bits in the word, including the parity bit A single parity bit can only reveal single bit errors since if an even number of bits are wrong then the parity bit will not change It is not possible to tell which bit is wrong

7. 7 Parity Check - Checksum A computed value based on the content of a block of data Transmitted or stored along with the data to detect data corruption Recomputed at the receiver end to compare with the one received Detects all errors with old bits of errors, and most errors with event number of bits It is computed by summing the bytes of the data block ignoring overflow Other parity check methods, such as Hamming Code, corrects errors

8. 8 RAID Types Five types of array architectures, RAID 1 ~ 5 Different disk fault-tolerance Different trade-offs in features and performance A non-redundant array of disk drives if often referred to RAID 0 Only RAID 1, 3 and 5 are commonly used RAID 2 and 4 do not offer any significant advantages over these other types Certain combination is possible (10, 35 etc) RAID 10 = RAID 1 + RAID 0

9. 9 RAID 0 - Striping No redundancy No fault tolerance High I/O performance Parallel I/O

10. 10 RAID 1 – Mirroring Provide good fault tolerance Works ok if one disk in a pair is down One write = a physical write on each disk One read = either read both or read the less busy one Could double the read rate

11. 11 RAID 3 - Parallel Array with Parity Fast read/write

12. 12 RAID 5 – Parity Checking For error correction, rather than full redundancy Each stripe unit has an extra parity stripe Parity stripes are distributed

13. 13 RAID 5 Read/Write Read: parallel stripes read from multiple disks Good performance Write: 2 reads + 2 writes Read old data stripe; read parity stripe (2 reads) XOR old data stripe with replacing one. Take result of XOR and XOR with parity stripe. Write new data stripe and new parity stripe (2 writes).

14. 14 RAID 10 – Striped Mirroring RAID 10 = Striping + mirroring An striped array of RAID 1 arrays High performance of RAID 0, and high tolerance of RAID 1 (at the cots of doubling disks).. More information about RAID disks at http://www.acnc.com/04_01_05.html

15. 15 Comparing RAID Levels RAID 0RAID 1RAID 5RADI 10 ReadHigh2XHigh WriteHigh1XMediumHigh Fault tolerance NoYes Disk utilisation HighLowHighLow Key problems All data lost when one disk fails Use twice disk space Lower throughput with disk failure Very expensive, not scalable Key advantages High I/O performance Very high I/O performance A good overall balance High reliability with good performance

16. 16 What RAID Provides Fault tolerance It does not prevent disk drive failures It enables real-time data recovery High I/O performance Mass data capacity Configuration flexibility Lower protected storage costs Easy maintenance

17. 17 Hardware vs. Software RAID Software RAID Software RAID: run on the server’s CPU Directly dependent on server CPU performance and load Occupies host system memory and CPU operation, degrading server performance Hardware RAID Hardware RAID: run on the RAID controller’s CPU Does not occupy any host system memory. Is not operating system dependent Host CPU can execute applications while the array adapter's processor simultaneously executes array functions: true hardware multi-tasking

18. 18 RAID Levels - Data Settings: accounts( number, branchnum, balance); create clustered index c on accounts(number); 100000 rows Cold Buffer Dual Xeon (550MHz,512Kb), 1Gb RAM, Internal RAID controller from Adaptec (80Mb), 4x18Gb drives (10000RPM), Windows 2000.

19. 19 RAID Levels - Transactions No Concurrent Transactions: Read Intensive: select avg(balance) from accounts; Write Intensive, e.g. typical insert: insert into accounts values (690466,6840,2272.76); Writes are uniformly distributed.

20. 20 RAID Levels SQL Server7 on Windows 2000 (SoftRAID means striping/parity at host) Read-Intensive: Using multiple disks (RAID0, RAID 10, RAID5) increases throughput significantly. Write-Intensive: Without cache, RAID 5 suffers. With cache, it is ok.

21. 21 Which RAID Level to Use? Log File RAID 1 is appropriate Fault tolerance with high write throughput. Writes are synchronous and sequential. No benefits in striping. Temporary Files RAID 0 is appropriate. No fault tolerance. High throughput. Data and Index Files RAID 5 is best suited for read intensive apps or if the RAID controller cache is effective enough. RAID 10 is best suited for write intensive apps.

22. 22 Controller Prefecthing No, Write-back Yes Read-ahead: Prefetching at the disk controller level. No information on access pattern. Better to let database management system do it. Write-back vs. write through: Write back: transfer terminated as soon as data is written to cache. Batteries to guarantee write back in case of power failure Write through: transfer terminated as soon as data is written to disk.

23. 23 SCSI Controller Cache - Data Settings: employees(ssnum, name, lat, long, hundreds1, hundreds2); create clustered index c on employees(hundreds2); Employees table partitioned over two disks; Log on a separate disk; same controller (same channel). 200 000 rows per table Database buffer size limited to 400 Mb. Dual Xeon (550MHz,512Kb), 1Gb RAM, Internal RAID controller from Adaptec (80Mb), 4x18Gb drives (10000RPM), Windows 2000.

24. 24 SCSI (not disk) Controller Cache - Transactions No Concurrent Transactions: update employees set lat = long, long = lat where hundreds2 = ?; cache friendly: update of 20,000 rows (~90Mb) cache unfriendly: update of 200,000 rows (~900Mb)

25. 25 SCSI Controller Cache SQL Server 7 on Windows 2000. Adaptec ServerRaid controller: 80 Mb RAM Write-back mode Updates Controller cache increases throughput whether operation is cache friendly or not. Efficient replacement policy!

26. 26 Enhancing Hardware Config. Add memory Cheapest option to get a better performance Can be used to enlarge DB buffer pool Better hit ratio If used for enlarge OS buffer (as disk cache), it benefits but to other apps as well Add disks Add processors

27. 27 Add Disks Larger disk ≠better performance Bottleneck is disk bandwidth Add disks for A dedicated disk for the log Switch RAID5 to RAID10 for update-intensive apps Move secondary indexes to another disk for write- intensive apps Partition read-intensive tables across many disks Consider intelligent disk systems Automatics replication and load balancing

28. 28 Add Processors Function parallelism Use different processors for different tasks GUI, Query Optimisation, TT&CC, different types of apps, different users Operation pipelines: E.g., scan, sort, select, join… Easy for RO apps, hard for update apps Data partition parallelism Partition data, thus the operation on the data

29. 29 Parallel Join Processing Algorithm: decompose and processing in parallel T = R S Let f: A  (1..n) (a hash function) R =  i=1..n R i, R i = {r  R | f(r.A) = i} S =  i=1..n S i, S i = {s  S | f(s.A) = i} T =  i=1..n R i S i Issues However, data distribution, task decomposition and load balancing are non-trivial AA

30. 30 Parallelism Some tasks are easier to be parallelised E.g., scan, join, sum, min Some tasks are not so easy E.g., sorting, avg, nested-queries

31. 31 Parallel DB Architectures Shared memory Tightly coupled, easy-to-use, but not scalable (bottlenecks when accessing shared memory and disks) Shared nothing A distributed with message-passing as the only communication mechanism Highly scalable Difficult for load distribution and balancing Shared disk A trade-off, but towards the shared-memory end

32. 32 Summary In this module, we have covered: The storage subsystem RAID: what are they and which one to use? Memory, disks and processors When to add what?