Decoupling Applications: SQS, SNS, Kinesis, Active MQ

Question 1

What is the different between synchronous and asynchronous communication?

Answer 1

In synchronous communication, applications talk to each other in real time and are tightly coupled.
In asynchronous communication, applications use a queue or middleware in between themselves to decouple themselves from each other.

Question 2

Why can decoupling be useful?

Answer 2

If all applications are tightly coupled, a sudden spike in traffic can result in the receiving application being overwhelmed and unable to handle all of the requests.

Question 3

What is Amazon SQS?

Answer 3

Simple Queuing Service, used to decouple applications through the use of a queue

Question 4

What is the limit of messages in an SQS queue?

Answer 4

There is no limit

Question 5

What is the maximum and default retention of messages in an SQS queue?

Answer 5

Max = 14 days
Default = 4 days

Question 6

What type of delivery do SQS queues have? What does this mean?

Answer 6

“At least once” delivery - this means that the message is delivered at least once, but since messages are only deleted once they have been processed, the same message can be processed by two different consumers at once

Question 7

What is used to regulate access to the SQS API?

Answer 7

IAM policies

Question 8

What is SQS message visibility timeout?

Answer 8

A set period of time for which a message is made ‘invisible’ to consumers so that it is not polled directly after being polled by a different consumer

Question 9

What can a consumer do if it knows it needs more time to process a message it has polled from the SQS queue?

Answer 9

It can increase the message visibility timeout through the use of an API

Question 10

What are the risks of setting the message visibility timeout to being too short or too long?

Answer 10

Too short: the messages will be consistently processed by more than 1 consumer which is inefficient
Too long: if a consumer crashes the message won’t be processed for a long period of time

Question 11

What is long polling in SQS?

Answer 11

When a consumer increases how long it waits for a message while connected to the queue if there are no messages in it. Once a message comes in, the consumer goes and processes it

Question 12

Why is long polling advantageous?

Answer 12

It reduces the number of API calls made to SQS while increasing the efficiency and reducing latency of the application

Question 13

What does FIFO stand for?

Answer 13

First-in, first-out

Question 14

What is the 1 positive and 1 negative of FIFO queues?

Answer 14

Positive: exactly-once delivery - no need to worry about duplicate processing
Negative: decreased throughput

Question 15

What model does SNS work off? What does this mean?

Answer 15

Publish/Subscription model - multiple subscribers receive messages as and when they are published by a publishing source

Question 16

Do subscribers get all published messages for topic they are subscribed to?

Answer 16

Yes, unless they filter them

Question 17

What is the fan out pattern? What services are included and why is this useful?

Answer 17

The fan out pattern is used if you want to send messages en masse to SQS queues.
You subscribe all the SQS queues to an SNS topic and then they proliferate out from there.
This is much easier than manual integration and is a fully decoupled architecture.

Question 18

What are the 4 Kinesis services?

Answer 18

Data Streams
Data Firehose
Data Analytics
Video Streams

Question 19

What is Kinesis Data Streams made for?

Answer 19

Streaming big data into your systems

Question 20

What are the two capacity modes for Kinesis Data Streams?

Answer 20

Provisioned and on-demand

Question 21

How are Kinesis Data Streams streams scaled?

Answer 21

More shards are added if more capacity is needed

Question 22

What is Kinesis Data Firehose?

Answer 22

A service that batch writes data from producers into destinations both inside and outside of AWS. Near real time. Fully managed.

Question 23

Does Kinesis Data Firehose or Kinesis Data Streams have data storage?

Answer 23

Data Firehose does not
Data Streams does, up to 365 days of replayability

Question 24

Is Kinesis Data Firehose or Kinesis Data Streams real time?

Answer 24

Data Firehose is not
Data Streams is

Question 25

When would you use Kinesis for data ordering over SQS and vice versa?

Answer 25

Kinesis would be used when you have larger amounts of data.
SQS FIFO is better if you want to dynamically adjust the number of consumers you will use.

Question 26

What is Amazon MQ?

Answer 26

A managed service for customers already using RabbitMQ or ActiveMQ - these are open-source message broker services.
MQ has both queue and topic features in one broker.