EC2 Flashcards
What pricing models are available for EC2
On-Demand
Reserved (contract 1-3 yrs)
Spot
Dedicated Hosts
EC2 Family F is for
FPGA (Genomics, Financial Analytics, Big Data)
EC2 Family I is for
IOPS (NoSQL, DBMS, Data warehousing)
EC2 Family G is for
Graphics (Video encoding, 3d Streaming)
EC2 Family H is for
High Disk Throughput (MapReduce, File System)
EC2 Family T is for
Cheap General Purpose (think t2.Micro)
EC2 Family D is for
Density (Fileservers, data warehouses, Hadoop)
EC2 Family R is for
RAM (Memory intensive apps)
EC2 Family M is for
Main choice (Applications)
EC2 Family C is for
Compute (CPU intensive)
EC2 Family P is for
Graphics (Think Pics; Machine Learning, BitCoin mining)
EC2 Family X is for
Xtreme Memory (Hana, Apache Spark)
EC2 Family Z is for
Xtreme Memory AND CPU (Electronic Design Automation)
EC2 Family A is for
ARM-based workloads
EC2 Family U is for
Bare-metal (to eliminate virtualization overhead)
True or False: Termination protection is off by default for new instances?
True; you must enable it yourself
True or False: By default, your root EBS volume will be deleted when you terminate your instance
True
True or False: You can blacklist specific ports or IP addresses using security groups.
False; You can specify allowed traffic only
What is the maximum number of security groups you can have per region?
2500
What are your cheap EBS volume options?
The Hard Disk ones, st1 and sc1
What are the SSD options for EBS volumes?
gp2 (general purpose) and io1 (for provisioned IOPs)
When would you use an Instance Store volume as opposed to an EBS?
When you only need ephemeral storage; instance store volumes are deleted when the instance is terminated
What are the components of an AMI?
The EBS snapshot or instance store templates
Launch permissions that control which AWS accounts can use the AMI
Block device mapping that specifies the volumes to attach when an instance is launched
You can select an AMI based on the following characteristics
Region OS Architecture Launch Permissions Storage for root device
True or False: You can modify the instance type, kernal, RAM, and volume usage on an instance-store backed instance?
False; you can’t stop (you can only run or terminate) an instance store backed instance so you can’t modify these parameters once you start.
What scenarios would you attach additional ENIs to an instance?
To create separate networks for your workloads, i.e. a management workload and an application workload. Adding ENIs won’t increase your throughput.
What scenarios would you use Enhanced Networking (ENA or Intel 82599 VF)?
You would use theses for SR-IOV device virtualization for higher I/o and lower CPU utilization to achieve speeds of 10 Gbps to 100 Gbps at no additional charge.
When would you use an Elastic Fabric Adapter (EFA)
To accelerate HPC for machine learning applications and to achieve OS-bypass on Linux machines.
What CloudWatch metrics are available for EC2 instances by default?
CPU, network, disk, and status checks (you’d have to create a custom one for memory utilization for example)
What is the default interval CloudWatch polls your EC2?
Every 5 minutes
What is the main difference between CloudWatch and CloudTrail?
CloudWatch is all about monitoring performance; CloudTrail is about auditing AWS calls and actions and resource activity
What information is available at 169.254.169.254/latest/user-data?
The bootstrap script
What information is available at 169.254.169.254/latest/meta-data?
Lots of information, including your public and private IP addresses, your tags, security groups, etc.
How do you attach an EFS to EC2s?
- Create the EFS
- Create an inbound rule on your instance security groups to allow NFS port 2049
- Mount EFS to the instances
What protocol is used for EFS?
NFSv4
How large can an EFS scale?
Petabytes
How many NFS connections can EFS support?
1000s
What kind of consistency does EFS provide?
Read after Write
What kind of availability does EFS provide?
Storage across multi-AZ
What type of file system would you use for SMB file storage?
Windows SFx
What high-performance file system is available for Linux machines and provides hundreds of Gbps throughput, millions of IOPs, and sub-millisecond latency?
SFx for Lustre
What is unique to a Spread Placement Group?
You can only have up to 7 running instances per availability zone per group
What are the features of a Partitioned Placement Group?
It’s like a Spread Placement Group, but allows multiple instances per partition. Each partition is within its own set of racks. This is ideal for workloads for HDFS, Hbase, Cassandra.
You can have a maximum of 7 partitions per Availability Zone.
What is the use case for a Clustered Placement Group?
Workloads that require high-throughput, low latency network for instances running logically clustered together within a single Availability Zone.
True or False: You can merge placement groups?
False
The name of a placement group must be unique within your AWS account within a
Region
True or False: You can launch host tenancy instances within a placement group
False
How can you specify your spot instance/fleet to run for a specified duration?
By using a Spot Block that reserves for a specified amount of time and will only be interrupted for capacity needs
What is the use case for spot instances?
Stateless, fault-tolerant, flexible applications like big data, containerized workloads, ci/cd, stateless web servers, HPC, and rendering workloads
What is specified in a spot instance request?
The persistence (one-time or persistent) Maximum spot price
What is specified in a spot fleet request?
Set of instances (which can include on-demand instances)
Target capacity
Maximum spot price
What AWS service allows you to run 100s of thousands of scheduled, parallel jobs spanned crossed instances
AWS Batch
Which AWS service provides an easy way to model an HPC cluster using a text file to define instances, resources, VPCs, subnets, cluster types, and scaling?
AWS ParallelCluster
