Chapter 5 checkpoint

Chapter 5 checkpoint: NoSQL and storage engines

Every question from this chapter’s lessons. Your first answer to each is the one that counts; afterward you can practice again as often as you like. Passing needs 75% right on the first attempt.

0 of 17 answered0 right on the first attempt

Question 1 of 17

Match each NoSQL data model to the job the lesson says it suits.

Key-valueDocumentWide columnGraph

Question 2 of 17

How does designing tables for Cassandra differ from relational normalization?

Schema design does not matter because it is schemalessYou put everything in one table with foreign keysYou start from the queries and design a table for each, duplicating data as neededYou normalize around entities and write joins later

Question 3 of 17

What does BASE stand for?

Basically available, soft state, eventual consistencyBalanced, available, sharded, elasticBackup, archive, snapshot, exportBatched, atomic, serializable, enforced

Question 4 of 17

Which of these does the primer list as data that suits a NoSQL store?

Rapid ingest of clickstream and log dataReports built from complex joinsA bank ledger that needs transactionsStrictly structured data with a fixed schema

Question 5 of 17

Why do write-heavy stores such as Cassandra and RocksDB use LSM-trees rather than B-trees?

They make point reads faster than a B-treeThey avoid write amplification entirelyThey keep each key in exactly one placeThey batch writes into large sequential runs instead of updating pages in place

Question 6 of 17

What does a Bloom filter do for an LSM-tree read?

It tells the read exactly which SSTable holds the keyIt keeps the memtable sortedIt compresses SSTablesIt lets the read skip SSTables that definitely do not hold the key

Question 7 of 17

A B-tree changes a 50-byte record by writing back its whole 8 KiB (8,192-byte) page. Ignoring the WAL record, what is the write amplification for that change?

Your answer×

Answers within 2% count.

Question 8 of 17

Put the steps of an LSM-tree write in order.

Compact SSTables into fewer sorted runs
Insert it into the in-memory memtable
Flush the full memtable to an immutable SSTable
Acknowledge the write to the client
Append the write to the write-ahead log

Question 9 of 17

A store keeps

N = 5

replicas and reads with

R = 2

. What is the smallest write quorum

W

that makes every read overlap every write?

Your answerreplicas

Answers within 1% count.

Question 10 of 17

A Cassandra table has a replication factor of 5. How many replicas must answer a read at consistency level QUORUM?

Your answerreplicas

Answers within 1% count.

Question 11 of 17

With

N = 3

replicas, match each setting to what it gives you.

W = 1

R = 3

W = 3

R = 1

W = 2

R = 2

W = 1

R = 1

Question 12 of 17

Match each repair mechanism to how it works.

Read repairHinted handoffMerkle-tree anti-entropy

Question 13 of 17

Two clients update the same Cassandra cell a moment apart, and the earlier update ends up as the stored value. What is the likely cause?

The read quorum was too smallVector clocks chose between them at randomClock skew: last-write-wins picked the update with the later timestamp, not the later updateHinted handoff lost the later write

Question 14 of 17

An IoT service stores readings in Cassandra with the reading’s date, such as 2026-11-13, as the only partition key. What happens under heavy ingestion?

Each write scatters across the whole clusterAll of today’s writes land on the same few nodes: a hot partitionNothing: Cassandra rebalances hot partitions automaticallyWrites spread evenly, because dates hash well

Question 15 of 17

You spread a hot daily partition over 200 partitions with a suffix, and you also need to read single orders by id. Which suffix should you use?

The full timestamp instead of the dateA suffix calculated from the order id, such as its hash modulo 200No suffix; add more nodes insteadA random suffix from 1 to 200

Question 16 of 17

A service writes each order to PostgreSQL and then to a search index from application code. Why does the lesson recommend change data capture instead?

CDC delivers each change exactly once, so no duplicates need handlingCDC makes the PostgreSQL write itself fasterA crash between the two writes leaves the stores silently disagreeing; CDC follows the database’s own logCDC gives one transaction across both stores

Question 17 of 17

In a polyglot design, match each kind of data to the store the lesson pairs it with.

Payments and ordersHigh-volume time-series eventsSessions and countersA social graph