Thursday 17 March 2011

Special RAID Type

Recently, I’ve seen two very special types of RAID. They are neither RAID5 nor RAID6. But they have the same tolerance of drive failure as RAID5/RAID6, where the data is still recoverable when two drives are failed.



1. RAID from Adaptec 3805
It has a data stripe size of 512 sectors and two rows as one data unit (5 data stripes) cycle. The data structure is:



P1= D5 XOR D2
P2= D1 XOR D3
P3= D2 XOR D4
P4= D3 XOR D5
P5= D4 XOR D1

For example, if Drive 1 and Drive 3 are failed, data on them can be regenerated by:
D1= D4 XOR P5;
D3= D5 XOR P4;




Due to the parity bits are not on the same row as the data bits, there is no software available in the market supports the regeneration of the missing data strips from crossed-row data and parity. The parity has to be on the same row as the data to regenerate the missing data by R-studio. Special script needs to be written to regenerate the content of the failed drives, so as to use conventional software to retrieve the data in the final step.
To regenerate the D1 on the first row of each data unit cycle, as in the example above, the script will calculate the XOR result between D4 on the first row and P5 on the second row and store the result on the first row of the missing drive. After running the script, the data structure will look like:



Now, we can use the R-Studio to retrieve the data. Build a virtual RAID using all the working drives and the drives regenerated. Set the RAID type as customized:




2. The second special type of RAID allows maximum two failed drives. It uses the redundancy of the original data, playing the same role as the parity. The data structure is:



Within a data unit cycle, it also has two rows as the first type. There is no parity at all. The green colored set of data is the redundancy. As can be seen from the data structure, even though it supports up to two failed drives, these two failed drives can NOT be 1+3, 1+4, 2+4, 2+5, 3+5. Any other failed drives combinations will be tolerated. For example, Drive 1 and Drive 2 are failed.




Using R-Studio, we can construct a customized virtual RIAD to retrieve the data as:


Written by: Zijian Xie (R&D Manager, MSc,BEng)

Thursday 24 February 2011

Script to attach MDF without LDF under SQL2000

USE MASTER
GO
SP_CONFIGURE 'ALLOW UPDATES',1 RECONFIGURE WITH OVERRIDE
GO
UPDATE SYSDATABASES SET STATUS =32768 WHERE NAME='db_NAME'
Go
sp_dboption 'db_NAME', 'single user', 'true'
Go
DBCC CHECKDB('db_NAME')
Go
update sysdatabases set status =28 where name='db_NAME'
Go
sp_configure 'allow updates', 0 reconfigure with override
Go
sp_dboption 'db_NAME', 'single user', 'false'
Go

Seagate 7200.12 ‘SIM ERROR’ Firmware Fault

The symptoms of ‘SIM Error’ firmware fault from hard drives belong to Seagate 7200.12 family are:

1. Hard drive status keeps BUSY forever after switching on;

2. In the serial command line, it outputs (when PCB is attached to hard drive):

SIM Error 3005
No HOST FIS-ReadyStatusFlags 2002A1A5
ASCII Diag mode

3. In the serial command line, it outputs (when PCB is detached from hard drive):

No HOST FIS-ReadyStatusFlags 2002A1A5

To use the conventional method of fixing a ‘BUSY and LBA=0’ 7200.11 hard drive, it outputs:

No HOST FIS-ReadyStatusFlags 2002A1A5
ASCII Diag mode
F3 T>/2
F3 2>Z
Spin Down Complete
Elapsed Time 6.015 secs
F3 2>
ASCII Diag mode
F3 T>/2
F3 2>U
HighPowerMode
ExecuteSpinRequest
Spin Up Complete
Elapsed Time 3.624 secs
F3 2>/1
F3 1>N1
Init SMART Fail

The initialization of SMART module is failed even though the hard drive passes the Spin up and Spin down operation. To solve this problem, instead of doing N1 to clear the SMART after comand U, we use command m0 first.

(P) SATA Reset

SIM Error 3005
No HOST FIS-ReadyStatusFlags 2002A1A5
ASCII Diag mode

F3 T>/2

F3 2>Z

Spin Down Complete
Elapsed Time 6.035 secs
F3 2>U

HighPowerMode
ExecuteSpinRequest

Spin Up Complete
Elapsed Time 3.573 secs
F3 2>/

F3 T>m0,2,2,,,,,22
Max Wr Retries = 00, Max Rd Retries = 00, Max ECC T-Level = 16, Max Certify Rewr
ite Retries = not supported

User Partition Format Successful - Elapsed Time 0 mins 00 secs

F3 T>/1

F3 1>N1

Clear SMART is completed.
F3 1>
Rst 0x08M
(P) SATA Reset


Written by: Zijian Xie (R&D Manager, MSc,BEng)

Thursday 13 January 2011

Types of RAID6

1. P+Q RAID6

Each data stripe of a P+Q double parity RAID6 has a standard RAID5 parity and a ‘Q’ parity which is generated by Reed-Solomon coding algorithm. It allows up to two drives failure. The minimum number of disks is 4.

The HP ADG RAID6 is a special type of P+Q RAID6. Within each standard data stripe, there are an integer number of smaller data stripes where the P and Q parities remain on the same disk. This is also called delay parity. Generally speaking, this type of RAID6 has a delay parity value of 16.


2. NetApp Double XOR RAID 6

It is also called RAID-DP and has a similar data structure of RAID4. P and Q parities are generated and stored on the same drives. It allows up to two drives failure. The minimum number of disks is 4.




3. X-Code RAID6

It is also called vertical XOR RAID6. The minimum number of disks is 4. It allows up to two drives failure. Each data stripe group has N (disk number) of data stripes. For example, if it has 5 disks, each data stripe group has 5 rows of data stripes as shown in the figure below.




4. ZZS Code RAID6
The minimum physical number of disks of ZZS code RAID6 is 5 and it has to be a prime number (5, 7, 11, 13, 19...). But the ZZS coding algorithm also allows one of the drives contains zeros. In another word, the minimum logical number of disk is 4. In ZZS coding algorithm, it also defines the calculation unit (one cycle) is equal to (n-1)/2 if there is n physical disks in the RAID6. For example showing below, it has 7 disks, thus, each cycle has three data stripes.



5. Park Code RAID6
Park is an engineer of IBM. He designed a new coding algorithm to implement a type of RAID6, supports disk number of 3 to 38. Park code RAID6 is always using three data stripes as a calculation unit.




Currently, I don't find any software application support the ZZS and Park RAID6. But, with all the disks intact, the data are still retrievable by customised RAID of R-studio. Leave your comments if you know any other types of RAID6 or solutions.


Written by: Zijian Xie (R&D Manager, MSc,BEng)