Module 8 Test: 20 MCQs
Ye tumhara progress tracker hai: 20 sawal, 5 Easy, 10 Moderate, 5 Hard . Pehle sawal khud hal karo (code ho to Colab me chalane se pehle output sochna), phir Answer Key se check karo.
| Score | Matlab |
|---|---|
| 18 se 20 | Zabardast! Module 8 tumhari mazboot buniyad hai. |
| 12 se 17 | Acha hai. Jin sawalon me ghalti hui, un chapters ko dobara parho. |
| 12 se kam | Pareshan nahi. Chapters 41 se 45 dobara parho aur projects par amal karo, phir test dobara do. |
EASY
Q1. Big Data ke 5 V me se kaunsa V batata hai ke data kitni tezi se aa raha hai?
- A) Volume
B) Variety
C) Veracity
D) Velocity
Q2. HDFS me default tor par har block ki kitni copies (replicas) rakhi jati hain?
- A) 3
B) 1
C) 2
D) 5
Q3. MapReduce ke teen qadam ka sahi tarteeb kaunsa hai?
-
A) Reduce, Map, Shuffle
-
B) Shuffle, Map, Reduce
-
C) Map, Shuffle, Reduce
D) Map, Reduce, Shuffle
Q4. Neeche me se Spark ka action kaunsa hai?
A) filter()
-
B)
count() -
C)
select()
D) withColumn()
Q5. Cloud me pay-as-you-go ka matlab kya hai?
-
A) Poore saal ka bill pehle dena
-
B) Cloud hamesha free hota hai
-
C) Sirf bari companies istemal kar sakti hain
D) Jitna istemal karo utna hi bill
MODERATE
Q6. 2 TB data, har disk ki raftar 200 MB/second, data 20 disks par barabar bata hai (1 TB = 1024 x 1024 MB). Poora data parhne me taqreeban kitna waqt lagega?
-
A) Taqreeban 52 second
-
B) Taqreeban 524 second
-
C) Taqreeban 5243 second
D) Taqreeban 10486 second
Q7. 1000 MB ki file, block size 128 MB, replication 3. Kitne blocks banenge aur kul kitni block copies rakhi jayengi?
A) 8 blocks, 24 copies B) 7 blocks, 21 copies C) 8 blocks, 8 copies D) 8 blocks, 3 copies
Q8. Is code ka output kya hoga?
from collections import defaultdict
g = defaultdict(list)
for line in ["a b a", "b a"]:
for w in line.split():
g[w].append(1)
print({k: sum(v) for k, v in g.items()})
A) {"a": 2, "b": 3} B) {"a": 5, "b": 5} C) {"a": 1, "b": 1} D) {"a": 3, "b": 2}
Q9. Spark ke explain() plan me Exchange nazar aaye to iska kya matlab hai?
A) Data cache me rakha ja raha hai
B) File Parquet me likhi ja rahi hai
C) Data partitions ke beech bheja ja raha hai (shuffle)
D) Schema badla ja raha hai
Q10. Is Spark code me **
print
tak asli hisab (data ki processing) kitni rows par hua?**
df2 = df.filter(F.col("qty") > 2)
df3 = df2.withColumn("x", F.col("qty") * 2)
print("tayyar")
A) Sab rows par, kyunke filter chal gaya
B) Kisi par nahi, kyunke koi action nahi chala
C) Sirf qty > 2 wali rows par
D) Sirf pehli 20 rows par
Q11. Partition ka usool key % 4 hai. Keys 10 se 19 (10 keys) hon to partition 2 me kitni keys jayengi?
A) 3
B) 2
C) 4
D) 1
Q12. Ek table me 10 lakh rows aur 20 columns hain. Query ko sirf 2 columns chahiye. Row-store ke muqable column-store (Parquet) me taqreeban kitni values kam parhni parti hain?
A) 2 guna kam B) 20 guna kam C) 10 guna kam (20M ke muqable 2M)
D) Koi farq nahi
Q13. Ek cluster ki farzi qeemat $0.50 per node per ghanta hai. 3 nodes, roz 10 ghante, 20 din chalte hain. Kul compute kharcha? A) $30 B) $150 C) $3,000 D) $300
Q14. Ek user ki IAM policy sirf s3:GetObject ki ijazat deti hai (ek bucket ke objects par). Wo user kya kar sakta hai?
A) File parh bhi sakta hai aur upload bhi
B) Us bucket ki file parh sakta hai, magar nayi file upload ya delete nahi kar sakta
C) Sirf delete kar sakta hai
D) Kuch bhi nahi kar sakta
Q15. sales me 5000 rows hain. regions table me 5 cities me se Quetta maujood nahi (sirf 4 cities). sales.join(regions, "city", "left") ke baad kitni rows hongi aur Quetta wali rows ka region kya hoga? (Quetta ki 975 rows hain.)
A) 4025 rows, Quetta ki rows hat jayengi
B) 5001 rows
C) 5000 rows, Quetta ka region null
D) Error aayega
HARD
Q16. Ek join bohat sust hai kyunke ek customer_id ke 90% records hain aur baqi partitions farigh baithe hain. Sab se munasib hal kaunsa hai?
A) Salting ya skew wali key ko alag sambhalna (ya broadcast agar ek table chhoti ho)
B) Cluster ke sab executors ki RAM barha dena
C) Result ko cache() kar dena
D) coalesce(1) kar ke sab ek partition me rakhna
Q17. Streaming events ke waqt (second) neeche hain aur 5 second ki tumbling window hai. Sab se zyada events kis window me hain aur kitne?
events = [1, 3, 4, 7, 8, 11, 12, 13, 14, 14] # (second)
window = 5
A) [0s, 5s): 3 events
B) [5s, 10s): 5 events C) [10s, 15s): 4 events
D) [10s, 15s): 5 events
Q18. Data lake me Raw zone par overwrite/badlav karna kyun ghalat hai?
A) Kyunke Raw zone ki files Parquet me honi chahiye
B) Kyunke asli data kho jata hai aur safai ki ghalti hone par dobara process nahi kar sakte
C) Kyunke Raw zone me sirf aggregates hote hain
D) Kyunke overwrite se bill kam ho jata hai
Q19. Ek Spark job har raat sirf 2 ghante chalta hai, magar cluster 24x7 on rehta hai aur bill bohat aata hai. Sab se behtar tabdeeli kaunsi hai?
A) Cluster me aur nodes jod dena
B) Data ko CSV me badal dena
C) Transient/auto-terminate cluster ya serverless: kaam ke waqt hi chale
D) Bucket ko public kar dena
Q20. Is Spark code ka nateeja (k, v) kya hoga? ( row_number har k ke andar v ko ghatay (descending) tarteeb me ginta hai.)
data = [("A", 10), ("A", 30), ("B", 5), ("B", 50), ("B", 20)]
df = spark.createDataFrame(data, ["k", "v"])
w = Window.partitionBy("k").orderBy(F.desc("v"))
(df.withColumn("r", F.row_number().over(w))
.filter("r = 2").select("k", "v").show())
A) ("A", 10) aur ("B", 20) B) ("A", 30) aur ("B", 50) C) ("A", 10) aur ("B", 5) D) ("A", 30) aur ("B", 20)
Module 8 Answer Key (Jawabat aur wazahat)
| # | Level | Jawab | Wazahat |
|---|---|---|---|
| 1 | Easy | D) Velocity | Velocity data ki raftar hai (har second aane wale events). Volume miqdar, Variety qisme, Veracity bharose ki baat hai. |
| 2 | Easy | A) 3 | Default replication factor 3 hai. Isi liye ek ya do machines kharab hone par bhi data bach jata hai. |
| 3 | Easy | C) Map, Shuffle, Reduce | Pehle Map (key, value banana), phir Shuffle (ek jaisi key ek jagah), aakhir me Reduce (jama/ginti). |
| 4 | Easy | B)count() | count()action hai: kaam foran chalata hai. filter, select,withColumn transformations hain jo sirf plan banate hain. |
| 5 | Easy | D) Jitna istemal karo utna hi bill | Pay-as-you-go me tum sirf istemal ki hui cheez (waqt, storage, data) ke paise dete ho. |
| 6 | Moderate | B) Taqreeban 524 second | Kul MB = 2 x 1024 x 1024 = 2,097,152. Raftar = 200 x 20 = 4000 MB/s. Waqt = 2,097,152 / 4000 = 524.3 second (taqreeban 8.7 minute). |
| 7 | Moderate | A) 8 blocks, 24 copies | 1000 / 128 = 7.8, yani 8 blocks (aakhri chhota). Har block ki 3 copies: 8 x 3 = 24. |
| 8 | Moderate | D) {"a": 3, "b": 2} | "a b a" aur "b a" me a teen baar (2 + 1) aur b do baar (1 + 1) aata hai. Shuffle ek jaisi key jama karta hai, Reduce un ka sum. |
| 9 | Moderate | C) Data partitions ke beech bheja ja raha hai (shuffle) | Exchange shuffle ko zahir karta hai (groupBy, join, orderBy jaisi wide transformations). Ye mehnga hota hai, isliye performance me shuffle kam karna maqsad hota hai. |
| 10 | Moderate | B) Kisi par nahi, kyunke koi action nahi chala | filter aur withColumn transformations hain (lazy). Jab tak count/show/write jaisa action na ho Spark sirf plan banata hai. |
| 11 | Moderate | A) 3 | Remainder 2 wali keys: 10, 14, 18, yani 3. Baqi: partition 0: 12, 16 (2); partition 1: 13, 17 (2); partition 3: 11, 15, 19 (3). |
| 12 | Moderate | C) 10 guna kam (20M ke muqable 2M) | Row-store me har row poori parhni parti hai: 20 x 1M = 20M values. Column-store me sirf 2 columns: 2M. Yani 10 guna kam. |
| 13 | Moderate | D) $300 | 0.50 x 3 x 10 x 20 = 300 dollar. |
| 14 | Moderate | B) Us bucket ki file parh sakta hai, magar nayi file upload ya delete nahi kar sakta | GetObject sirf object parhne ki ijazat hai. PutObject (upload) ya DeleteObject alag ijazaten hain jo yahan nahi di gayin (least privilege). |
| 15 | Moderate | C) 5000 rows, Quetta ka region null | Left join left table ki sab rows rakhta hai, isliye 5000. Jin ki key right me nahi milti un ke right wale columns null hote hain. Inner join me 4025 rows bachti. |
| 16 | Hard | A) Salting ya skew wali key ko alag sambhalna (ya broadcast agar ek table chhoti ho) | Masla data skew hai: ek partition par bojh. RAM barhane ya cache se bojh nahi bantta; coalesce(1) ise aur kharab karta hai. Salting key ko bant deti hai. |
| 17 | Hard | D) [10s, 15s): 5 events | Windows: [0,5): 1, 3, 4 = 3 events; [5,10): 7, 8 = 2; [10,15): 11, 12, 13, 14, 14 = 5. Sab se zyada [10s, 15s) me 5 hain. |
| 18 | Hard | B) Kyunke asli data kho jata hai aur safai ki ghalti hone par dobara process nahi kar sakte | Raw jaisa aaya waisa rehna chahiye. Agar safai ya hisab me ghalti nikle to Raw se Clean aur Curated dobara banaye ja sakte hain. Parquet/aggregates Clean/Curated ki baat hai. |
| 19 | Hard | C) Transient/auto-terminate cluster ya serverless: kaam ke waqt hi chale | Kharcha cluster ke chalne ke waqt par hai. Kaam ke waqt hi cluster banane/band karne se (ya serverless se) bill kam hota hai. Nodes barhane se bill barhta hai; public bucket khatarnak hai. |
| 20 | Hard | A) ("A", 10) aur ("B", 20) | Har k me v ghatay tarteeb: A: 30, 10; B: 50, 20, 5. |
r = 2
doosri row
hai: A ki 10 aur B ki 20.|