4.2.39 · D5 · HinglishOperating Systems
Question bank — RAID — levels 0, 1, 5, 6, 10 — trade-offs
4.2.39 · D5· Coding › Operating Systems › RAID — levels 0, 1, 5, 6, 10 — trade-offs
Shuru karne se pehle, ek one-line reminder un vocabulary words ki jo har trap mein kaam aati hain:
Recall Woh words jo pehle pakke hone chahiye
Striping ::: ek file ko blocks mein kaatna aur unhe disks par scatter karna taaki parallel mein read/write ho sake. Mirroring ::: kisi disk ki byte-for-byte identical copy doosri disk par rakhna. Parity ::: ek computed check-block jo kisi bhi ek missing data block ko rebuild karne deta hai. XOR () ::: "bits-differ" operation jisme magic property hai aur , isliye yeh khud ko undo kar leta hai. Stripe ::: blocks ka woh set (data + parity) jo saari disks par same position par hote hain.
True or false — justify karo
Yahan har item ek claim hai. Pehle true ya false decide karo, phir reason do — reveal mein hamesha kyun hoga, sirf verdict nahi.
RAID 0 reliability improve karta hai kyunki data kaafi saari disks par spread hota hai.
False. Data ko disks par spread karna failure ko zyada likely banata hai — koi bhi ek disk khone se poora logical volume destroy ho jaata hai, isliye mean-time-to-failure worse hoti hai, better nahi. Dekho Reliability and MTBF.
RAID 1 mirror nightly backups ka valid replacement hai.
False. Mirror har write instantly copy karta hai, jisme accidental
rm, ransomware encrypt, ya filesystem corruption bhi shaamil hai — galti dono disks par jaati hai. RAID availability deta hai, history nahi.RAID 5 do disks ke ek saath fail hone se bachaa sakta hai.
False. Ek parity block = ek equation = ek recoverable unknown. Do simultaneous failures mein do unknowns aur sirf ek equation hoti hai, isliye data unrecoverable hai.
RAID 6 do failures se is liye bachta hai kyunki parity do baar store karta hai.
False. XOR parity do baar store karne se same equation do baar milti hai — phir bhi sirf ek usable equation. RAID 6 ka doosra parity $GF(2^8)$ par compute hota hai isliye woh independent hota hai, ek genuine doosri equation deta hai.
Same number of disks ke liye, RAID 5 mein RAID 6 se zyada usable capacity hoti hai.
True. RAID 5 mein ek disk ki space sacrifice hoti hai (), RAID 6 mein do (), isliye capacity mein RAID 5 jeeeta hai — lekin doosri failure tolerate karne mein haarta hai.
RAID 10 aur RAID 0+1 same usable capacity dete hain.
Capacity mein True (dono hain), lekin yahi ek cheez hai jo share karte hain — unki failure tolerance aur rebuild behaviour alag hai kyunki mirroring aur striping ka order alag hai.
RAID 1 par reads single disk se faster ho sakti hain.
True. Read ke liye sirf ek copy chahiye, isliye do mirrored disks ek saath do independent reads serve kar sakti hain, roughly read throughput double ho jaati hai.
RAID 1 par writes single disk jitni cheap hoti hain.
False. Unhe identical rakhne ke liye har write dono copies par land karni chahiye, isliye mirror double write work karta hai (haalaanki dono writes parallel hoti hain).
Zyada parity disks add karna hamesha array ko faster banata hai.
False. Parity write cost (read-modify-write) add karta hai, isliye RAID 6, RAID 5 se small writes par slower hai, jo RAID 0/10 se slower hai jisme parity bilkul nahi hai.
Kyunki parity "sirf XOR" hai, ise compute karna essentially free hai.
Half-true. compute karna sasta hai; mehenga part yeh hai ki small write par ise update karna extra disk reads aur writes force karta hai — cost I/O ki hai, arithmetic ki nahi.
Error dhundho
Har line mein ek flawed statement hai. Reveal uss specific broken step ka naam batata hai.
"RAID 5 mein disk 2 rebuild karne ke liye, sirf surviving data disks ko aapas mein XOR karo."
Error: parity block ko bhi XOR karna zaroori hai. Rule yeh hai — drop karo aur surviving data akela missing block reconstruct nahi kar sakta.
"Ek small RAID 5 write 2 I/Os hai: naya data likho, nayi parity likho."
Error: pehle read karna padta hai old data aur old parity ko taaki compute ho sake. Yeh hai 2 reads + 2 writes = 4 I/Os.
"RAID 6 ka write penalty 4 I/Os hai, same as RAID 5."
Error: RAID 6 do parity blocks ( aur ) update karta hai, isliye ek small write hai old data read + old read + old read + data write + write + write = 6 I/Os.
"8 disks of 4 TB ke saath, RAID 6 mein 28 TB usable milti hai."
Error: yeh RAID 5 ka answer hai. RAID 6 mein do disks parity ko jaati hain, isliye usable TB.
"RAID 5 rebuild ke dauran array fully protected rehta hai."
Error: jab ek disk rebuild ho rahi hoti hai tab array zero redundancy ke saath chal raha hota hai — rebuild ke dauran doosri failure se saara data chala jaata hai. Isliye bade arrays RAID 6 prefer karte hain.
"RAID 10 pehle stripe karta hai, phir stripes ko mirror karta hai."
Error: yeh RAID 0+1 describe karta hai. RAID 10 pehle mirror karta hai (pairs mein), phir pairs par stripe karta hai — order matter karta hai ki kitne failure patterns survive karte hain.
"Kyunki parity data ka XOR hai, parity disk zyaatar time idle rehti hai."
Error: RAID 5 mein parity saari disks par rotate hoti hai, isliye koi dedicated parity disk nahi hoti — har disk data aur parity dono blocks rakhti hai, write load spread ho jaata hai.
Why questions
Underlying reason answer karo, sirf fact nahi.
XOR (ordinary addition ki jagah) single-erasure recovery ke liye kyun kaam karta hai?
XOR apna khud ka inverse hai (), isliye surviving blocks ko dobara XOR karna ke andar unki copies cancel kar deta hai aur exactly missing block reh jaata hai — ordinary addition mein borrows/carries hote aur bits par self-cancel cleanly nahi hota. Dekho XOR and Boolean Algebra.
RAID 6 do ordinary XOR parities use karke do failures survive kyun nahi kar sakta?
Do XOR parities bits par same linear equation produce karti hain, isliye milkar bhi sirf ek unknown solve karte hain. Independence ke liye alag coefficient chahiye har disk ke liye, jo multiplication provide karta hai.
Small write ke liye read-modify-write penalty unavoidable kyun hai?
Parity valid rakhne ke liye old data aur old parity chahiye taaki old term cancel ho sake (); unhe read kiye bina yeh nahi jaana ja sakta ki se kya remove karna hai.
RAID 5 parity ko disks par rotate kyun karta hai, ek dedicated parity disk rakhne ki jagah?
Dedicated parity disk har update par write hoti, bottleneck ban jaati; rotating parity write traffic evenly spread karta hai, isliye koi single disk hot nahi hoti. Relates to Disk Scheduling load balancing.
Write-heavy databases ke liye RAID 5 ki jagah RAID 10 kyun prefer kiya jaata hai?
RAID 10 writes plain mirrored writes hain (koi parity recomputation nahi), isliye koi read-modify-write penalty nahi — small random writes, jo databases mein common hain, fast rehti hain.
Zyada disks par striping throughput kyun improve karta hai?
Har disk ka apna head hota hai jo independently seek aur transfer kar sakta hai, isliye disks woh time mein blocks move kar sakti hain jab ek disk ek block move karta hai — parallelism hai, faster individual disks nahi. Isliye File Systems stripe alignment ki parwah karte hain.
RAID 0 array mein disks add karna total failure zyada likely kyun banata hai?
Koi redundancy nahi hai, isliye array tabhi survive karta hai jab saari disks survive karein; jitni zyada disks, utni zyada chance ki kam se kam ek mar jaaye, isliye combined reliability girti hai. Dekho Reliability and MTBF.
Edge cases
Boundary aur degenerate inputs — woh scenarios jo sloppy mental models tod dete hain.
RAID 5 ko minimum kitni disks chahiye, aur exactly 3 disks ke saath kya hota hai?
RAID 5 ko kam se kam 3 disks chahiye; 3 ke saath, do data rakhti hain aur ek (rotated) parity rakhti hai, usable capacity 2 disks ki milti hai aur 1 failure tolerate hota hai.
Smallest valid RAID 6 array kaun sa hai, aur uski usable capacity?
Minimum 4 disks; 4 ke saath, usable capacity disks ki hoti hai, aur yeh phir bhi 2 failures tolerate karta hai — extreme case jahan redundancy capacity ke barabar hoti hai.
Agar RAID 1 pair ki dono disks kho jaayein, toh kya recoverable hai?
Kuch nahi — mirroring sirf ek failure per pair tolerate karta hai. Usse aage, redundancy exhaust ho jaati hai aur data chala jaata hai.
RAID 10 mein, kya array do disk failures survive kar sakta hai?
Kabhi kabhi. Yeh do failures tab survive karta hai jab dono alag mirror pairs mein hon; agar dono failures same pair par hoon, to woh pair (aur array) lost ho jaata hai.
Parity compute karte waqt all-zeros data block ka kya hota hai?
Kuch contribute nahi karta, kyunki — parity zero block se unchanged rehti hai, aur recovery phir bhi kaam karti hai kyunki zero ek perfectly valid recoverable value hai.
Agar kisi stripe ke saare data blocks identical hoon, toh parity kya hogi?
Even number of identical blocks ke liye XOR parity all zeros hogi (pairs mein cancel ho jaate hain); odd number ke liye woh block ke barabar hogi — parity phir bhi kisi bhi single missing block ko uniquely determine karti hai.
Single-disk "array" ki kya redundancy hoti hai?
Koi nahi — koi cheez nahi hai jiske against stripe karein ya mirror karein. RAID levels tabhi value add karte hain jab (mirroring) ya (single parity).