NoSQL

Question 1

Strengths of RDBMS (4)

Answer 1

1. Consistency (ACID)
2. Integration of data with schema normalization
3. SQL language well known
4. Robust (40+ years in organizations)

Question 2

Weaknesses of RDBMS (4)

Answer 2

1. Bad scaling
2. Prioritizes consistency over latency
3. Schema rigidity (no evolution)

Question 3

No SQL common features (4)

Answer 3

1. No just rows and tables
2. Freedom from joins
3. Schemaless or soft-schema
4. Distributed architecture (cluster)

Question 4

Can NoSQL systems be used for OLAP?

Answer 4

Possibly, but through analytical tools like Spark

Question 5

4 main data models seen in NoSQL

Answer 5

1. Key-value
2. Document
3. Wide column
4. Graph

Question 6

Graph data model

Answer 6

Focuses on relationship between data elements with vertices representing the entity, arcs representing relationship between entities, and properties descriping the vertices

Question 7

2 examples of graph database queries

Answer 7

1. Find friends of friends
   (user)-[:KNOWS]-(friend)-[:KNOWS]-(foaf)
2. Find shortest path between A and B
   shortestPath(:KNOWS*…5]-(userB))

Question 8

What is the opposite of Graph Modeling?

Answer 8

Aggregate modeling
(key-value, document, wide-column)

Question 9

What do we call tables in the document data model?

Answer 9

Collections, which hold a list of documents (often JSON format)

Question 10

What does each document need to contain in the document data model? (2)

Answer 10

A set of fields corresponding to Key-value pair and mandatory ID

Question 11

{
“_id”: 1,
“name”: “Martin”,
“adrs”: [
{“street”:”Adam”, “city”:”Chicago”, “state”:”illinois”, “code”:60007},
{“street”:”9th”, “city”:”NewYork”, “state”:”NewYork”, “code”:10001}
],
“orders”: [ {
“orderpayments”:[
{“card”:477, “billadrs”: {“street”:”Adam”, “city”:”Chicago”, “state”:”illinois”, “code”:60007}},
{“card”:457, “billadrs”: {“street”:”9th”, “city”:”NewYork”, “state”:”NewYork”, “code”:10001}}
],
“products”:[
{“id”:1, “name”:”Cola”, “price”:12.4},
{“id”:2, “name”:”Fanta”, “price”:14.4}
],
“shipAdrs”: {“street”:”9th”, “city”:”NewYork”, “state”:”NewYork”, “code”:10001}
}]

Answer 11

We can query this into different sections, such as Product Collection and Order Collection to return document with just those relevant fields

Question 12

What does a key contain and what does a value contain in document data model

Answer 12

unique string (path, queries, REST calls, ID)
BLOB (binary large object) - HTML, pdf etc

Question 13

Why is the value considered a black box when querying key-values in NoSQL?

Answer 13

There are no indexes on the values, no “where” clauses allowed. Schema information is often indicated in the key

Ex:
Key Value
user: 1234: name Enrico

Question 14

Wide-column data model

Answer 14

in a RDBMS, data is stored in tables with rows that span a certain number of columns. If a particular record/row needs another column, you have to add it to the entire table. In Wide-column, you don’t

Question 15

Key-value vs wide-column

Answer 15

Key-Value databases are the simplest model and can be thought of as a configuration file or a two-column table of keys with an associated value. Wide-column databases expand that key-value store concept across multiple columns, but only the columns that are needed for that record.

Question 16

How to query wide-column data

Answer 16

SQL-like language works since its similar to relational model

Question 17

Is it easier to scale aggregate data or graph?

Answer 17

Aggregate, because splitting graph data across cluster often means arcs are “cut” meaning several cross-machine links

Question 18

sharding

Answer 18

distributing data across different nodes

Question 19

replication

Answer 19

creating copies of the data on several nodes

Question 20

3 good practices for sharding

Answer 20

1. Data locality (italian customer data in european data center)
2. Balance (same amount of data on each node)ù
3. Related data accessed together (orders for each client stored on same node)

Question 21

Hash data partition strategy

Answer 21

equal distribution across nodes, but range queries become inefficient

Question 22

range data partition strategy

Answer 22

distribute based off value ranges, can lead to heavy data redistribution

Question 23

Master job in NoSQL

Answer 23

manage data and handle write operations

Question 24

Slaves job in NoSQL

Answer 24

Enable read operations and become master if the master fails

Question 25

peer-to-peer replication

Answer 25

different from master-slave model, each node has same importance and can handle write operations, but two users may update the same value from different replicas…

Question 26

Write conflict mitigation methods (3)

Answer 26

Last write wins
conflict prevention - verify that value hasn’t changed since last read
conflict detection - preserve history, merge results and let user decide

Question 27

Consistency in RDBMS (ACID)

Answer 27

Atomicity (no partial transactions)
Consistency
Isolation
Durability

Question 28

Consistency in NoSQL (CAP)

Answer 28

A distributed system can only guarantee 2 out of the following 3 properties:

1. Consistency (C)
2. Availability (A)
3. Partition Tolerance (P)

Question 29

P.AC E.LC model

Answer 29

Refinement of CAP theorem adding another dimension

PA EL= “Prioritize Availability”
PA EC= “Sacrifice consistency only for partitioning”
PC EL = “Enforce consistency during partitioning”
PC EC = “Strong consistency at all times”