EC2 Placement Groups

Question 1

What are the three types of Placement Groups?

Answer 1

Cluster, Spread and Partition

Question 2

What’s the primary goal of Cluster placement groups?

Answer 2

To ensure that EC2 instances are physically close together

Question 3

What are the two performance benefits of Cluster placement groups?

Answer 3

Higher speeds and lower latency

Question 4

Can you configure which AZ a Cluster placement group is deployed into?

Answer 4

No.

AWS decides the AZ at launch time, based on the first instance that’s launched.

Question 5

How do you change the AZ that a Cluster placement group is in?

Answer 5

You can’t. Once the first VM is deployed, its placement group is locked to that single AZ.

Question 6

How many AZs can a Cluster placement group span?

Answer 6

One.

Question 7

What is the maximum data transfer rate between hosts in a Cluster placement group, and how is this limit expressed?

Answer 7

It is 10 Gbps per 5-tuple (also known as single-stream). This is double the normal limit of 5 Gbps.

Question 8

Is it possible to span a Cluster placement group across VPC peers, and if so, what impact does it have?

Answer 8

Yes. But doing so impacts performance.

Question 9

Cluster placement groups must use ………… instance types

Answer 9

Supported

Question 10

Can you use mixed EC2 instance types within in a Cluster placement group.

Answer 10

Yes, but it is recommended to use the same instance type for all members.

Question 11

What is the recommended way to launch instances in a Cluster placement group? Why is this?

Answer 11

It is strongly recommended to launch all instances at the same time.

This allows AWS to choose hardware that has room for all of your instances.

Question 12

What’s the primary goal of Spread placement groups?

Answer 12

To provide infrastructure isolation, spreading your instances across distinct underlying hardware to reduce correlated failures.

Question 13

How do Spread placement groups achieve their goal?

Answer 13

By running each instance from a different rack, with each rack having its own network and power source.

Question 14

With Spread placement groups, there is a max limit of …… instances per ……….

Answer 14

Max of 7 instances per AZ.

Question 15

Spread placement groups <can|cannot> span multiple AZs?

Answer 15

They can.

Question 16

What class of instances are NOT supported in Spread placement groups?

Answer 16

Dedicated instances and Dedicated hosts are not supported. Only shared instances are supported.

Question 17

What is the typical use case for Spread placement groups?

Answer 17

Critical workloads that require high availability and reliability.

Question 18

What is the ideal placement group type for critical workloads that require high availability and reliability?

Answer 18

Spread placement groups.

Question 19

How do Partition placement groups achieve their goal?

Answer 19

Instances are placed into partitions, which have their own power and network connectivity. If there’s a failure in one partition, only the instances within that partition will be affected.

Question 20

In a Partition placement group, what determines which partition an instance is placed into?

Answer 20

The user can choose, or AWS can auto-assign.

Question 21

What is the typical use case for Partition placement groups?

Answer 21

Large distributed and replicated workloads, such as big data applications or large-scale distributed databases. Think HDFS, HBase, Cassandra.

Question 22

What is the ideal placement group type for large distributed and replicated workloads, such as big data applications or large-scale distributed databases?

Answer 22

Partition placement groups.

Question 23

What is the ideal placement group type for HDFS, HBase or Cassandra?

Answer 23

Partition placement groups.