Module 5: Adding a compute layer with eC2 Flashcards
Role of cloud architect when designing a compute layer
Analyse key characteristics of the workload to choose AMI, EC2 instance type and storage options that optimize the perf and security
What are the different type of compute options proposed by AWS
1) Virtual Machines
2) Containers
3) Virtual Private Servers
4) Platform as a service
5) Serverless
What AWS service propose virtual Machines
EC2
What AWS Services allow the usage of containers?
Amazon Elastic Container Service (ECS)
Amazon Elastic Kubernetes Service (EKS)
Use both service to run container application in highly available and scalable environment
What AWS service allows for Virtual Private servers (VPS)
Amazon Lightsail
What AWS service aloow computing in a PaaS setting
Beanstalk
What serverless computing services are part of the AWS ecosystem
AWS lambda
AWS fargate
What is EC2?
Elastic Cloud compute.
It provides secure and resizable virtual servers in the cloud that can run workloads of differing sizes.
What is ECS?
Elastic Container Services:
Allows you to use docker containers in the cloud
What is EKS
Elastic Kubernetes services
Allows you to use kubernetes
What is Amazon Lightsail?
Lightsail is a virtual private server that provides devekopers with compute storage and networking capabilities. Light weight with everything. Simple management interface. Use case include running website, devloping apps, blog, e-commerce…..
What is Lambda?
Serverless compute solution allowing you to run code like Java, Go, Powershell, Node.js, C#, Python or Ruby
What is Beanstalk
Solution that runs web apps and services developped in java, .net, php, node.js, ruby, Go and docker. you upload your code and the service automatically handle the devlopment for you.
What languages are supported by beanstalk?
Java
:NET
PHP
Node.js
Python
Ruby
Go
Docker
What is Fargate?
serverless compute for containers
What is the advantage of serverless services?
They enable you to run workloads withou having to provision configure or manage servers
Compute service categories differentiators
Structure of an EC2
The instance runs on top of an Hypervisor. The hypervisor is the platform that provides access to the actual hardware supporting the machine.
What is an instance store ?
Ephemeral storage physically attached to the host computer and provides temporary block storage to the instance
What is a common supporting memory service to EC2?
EBS (Elastic Block Store) used for the boot disk and other storage needs. Provides persistent block storage volumes. The data will be persisted even when the instance is stopped.
How does EBS optimized instance improve the synergy with EC2?
By minimizing the I/O contention between the volume and other traffic from the instance
Why choose EC2?
You need complete control of your computing resources, OS; CPU; STORAGE; Applications… It can be used for running any type of workload
Type of instances for optimizing cost?
On demand instance,
Reserved instances,
Spot instances,
Saving plans
Dedicated hosts
What do you need to configure if you want to access the instance using Secure Shell (SSH)or Remote desktop Protocol (RDP)
You need to setup a key pair. A set of security credentials
AMI or Amazon Machine Image
An AMI is a template for the root volume
AMI benefits
Repeatability, Reusability, Recoverability
5 key characterirstics to choose an AMI
1) The Region: Each AMI exists in a specific Region
2) The OS: Windows o Linux variant
3) Storage for the root device: EBS for persistence or Instance store
4) Architecture: Processor architecure 32bits or 64 bits
5) Virtualization type : Paravirtual (PV) or Hardware Virtual Machine (HVM). (Better perf = HVM)
Sources to obtain AMI :
Quick starts (AWS built AMI)
My AMI (create your own)
AWS Marketplace: AMI from software vendors for a multiple of use case
Community built AMI: created by a community of users but not vetted by AWS. Use at own risk
Instance on EBS vs instance store:
Which boots faster?
EBS
If you want a fast boot what type of storage should you choose?
EBS
Max Size of root device on Instance store
10Gib
Max Size of root device on EBS
16 TiB
Can I stop the instance without an EBS ?
No. Either running or terminated. You need EBS to stop
Can you change the instance type on instance store?
No you can’t because it needs to be stopped to be change and you can’t stop without EBS
What is more expensive ?
EBS or instance store
EBSE because you also are charge for EBS volume usage, storing your AMI as a snapshot and the instance usage
If you want persistent storage. What do you use for storage?
EBS (Elastic Block Store)
Through which interface can you reboot an instance
the EC2 console, CLI or SDK.
A rebooting instance keeps:
the public DNS and public IP address. It stays on the same physical host
Can you recover or reconnect to a terminated instance ?
No
Starting a stop instance migrate the instance to a new host computer, true or false
True
What is the difference between stop and hibernate
It retains the data when hibernated
How do you create an image ?
By using EC2 Image Builder
What elements are impacted by my choice in AMI ?
CPU
Memory
Storage
Network performance
Why should I know the workload and my cost requirements before launching an instance?
Helps choose the right instance as it will configure the performance settings and cost of the instance
What type of instance do not support enhanced Networking?
T2
What is enhanced networking?
It means that the Session Initiation Protocol (SIP) workloads have access to consistent bandwidth and lower aggregate latency
Naming convention of instance.
What does c7gn.xlarge means
C: family
7: Generation
g: Processor family
n: additional capabilities
X-Large: size
What are the 6 types of instance that we have
1) General Purpose
2) Compute optimized instance
3)Storage Optimized instance
4)Memory optimized
5)Accelerated computed instance
6)High performance computing
What are the instance families for general purpose instances types ?
M;T
Web app server
Enterprise application
Gaming servers
What are isntance families for compute optimized ?
C (compute)
Batch processing
Distributed analytics
What are isntance families for storage optimized ?
I;D;H
High perf databases
Real time analytics
What are isntance families for memory optimized ?
R;X;Z
High perf databases
Big data analytics
What are isntance families for accelerated computing ?
P;DL;VT;G;F;Trn;Inf
Machine learning and AI
What are isntance families forHigh performance computing ?
Hpc (High performance Computing)
Deep learning workloads
When planning for my instance should I underplan and resize or oversize and resize ?
Undersize because you will adjust later on. If you over-specs, then it will hide performance issues and it will cost too much for what is really needed
What is teh advantage of choosing the latest AMI in the family
It generally has the best cost/performance ratio.
There is a service providing recommendations on how to optimize instance type. How is it called ?
AWS Compute Optimizer.
It analyses the run time and advice for improved performance and cost reduction
What techology hides behind Cosmputeoptimizer?
Amazon ML (Machine learning). It provides recommendation for M, C, R, T and X families
What are the recommendations given by compute optimizer ?
Under-provisioned, over-provisionned, optimized or none.
In what circumstances does the compute optimizer returns none
If:
1) The compute optimizer runs since less than 12 hours.
2) The instance has been running for less than 30 hours
3) The instance family is not supported
Match the workload to the instance family:
Transactional databases
C / M / I / P / Inf2 / R / T
I because I family is Storage optimized
Match the workload to the instance family:
Small devlopment environment
C / M / I / P / Inf2 / R / T
T
Because they are general purpose instances
Match the workload to the instance family:
Gaming servers
C / M / I / P / Inf2 / R / T
M
General purpose
Match the workload to the instance family:
In-memory caches
C / M / I / P / Inf2 / R / T
R because R instances are memory optimized
Match the workload to the instance family:
Image and video generation
C / M / I / P / Inf2 / R / T
Inf2 because inf2 is accelerated computing
Match the workload to the instance family:
Machine learning
C / M / I / P / Inf2 / R / T
P because inf2 is accelerated computing
Match the workload to the instance family:
Batch processing
C / M / I / P / Inf2 / R / T
C because C is Compute optimized
Where can I find more info on the instance type
Instance type page in the management console
What are the four storage options for an instance ?
1) Amazon EBS (Elastic Block Store)
2) Amazon EFS (Elastic File System)
3) Amazon FSx for windows file server
4) Instance store
What is a root volume for ?
It stores the OS and initial file systems for the instance
What are the only two storage type can be used to store a root volume ?
EBS and Instance store
When to use an instance store?
If you need fas compute time and data doesn’t need to be stored persistently on the volume.
If the instance is stopped, what happens to the data on the instance store
It gets deleted
If I want the data to persist, what shoud I do ?
Use EBS.
What is the issue common to EBS and instance store?
They can be used by a single volume at a time. However, EBS can be detached and reassigned while instance store is fixed.
What storage type to choose if I want several instances to collaborate on the single volume ?
EFS for linux instances and FSx for Microsoft Windows instances
Instance store technology ?
SSD or HDD on block storage. The disk is physically attached to the computer running the instance.
Good for buffers, caches, scratch data
If I want an instant store with high I/O performance what should I choose ?
SSD with NVMe (non volatile memory express)
If the instance is rebooted, do I keep the data in the instance store ?
YES, but not when terminated
EBS technology
Persistent block storage in SSD or HDD
EBS advantages
Cana attach to any other instance in teh same AZ.
Can be encrypted
Support snapshot that are persisted to S3
Data persists
What type of instance storage can be compared to an external hard disk ?
Amazon EBS (Elastic Block Store)
Why do EBS instance have lw access latency ?
Because they are mounted directly on the instance
Why choose an EBS in SSD?
If I have high transactional workloads with frequent read/write operations and small I/O sizes
If I want to use an instance to support a datawharehouse with a lot of transactions what type of storage is the best suited ?
EBS with IOPS SSD (in out per second)
Why should I choose and EBS HHD backed?
For large streaming download. Where throughput in MIb/s is a more critica component.
Two version of HHD for EBS and advantage of each?
Throughput optimized instances:
Streaming workload, big data
Cold HDD:
Lowest cost HDD volume but designed for less frequent access
I can select an EBS to match an instance but can I choose an optimized instance for my EBS ?
YES
What if you have multiple instances that must use the same storage ?
Best solution is to use EFS for Linux or FSx for windows.
Why is S3 not prefered to EFS or FSx when sharing storage accross instance ?
Because S3 is object based so if you have a high througuput then because you need to overwrite entire file for each modification, this is not at all optimal compared to a file system like EFS or FSx.
Which file storage system service is Linux based
EFS
Advantages of NFS
Scales automatically,
Support NFS,
Petabytes of capacity
Works accross several instance
Uses NFS protocol
Are mutlitple AZ supported by EFS ?
YES
Each AZ has a mount with an IP that EFS connects to.
EFS in One zone storage is cheaper by up to how much compared to multi AZ?
up to 47% cheaper. So consider the use case well before spreading your EFS.
Of course you gain both durability and availability with multiple AZ.
If I am One zoning an EFS can I still connect an instance from another AZ to it ?
Yes. The EFS will just be hosted on a single AZ but still made accessible through mount in the AZ it is hosted on
What is instance user data
It is a script used to run shell commands or cloud init directives. It runs with root user priviledges after the instance start but before it gets accesible to the network.
You can use it to patch and update software from the AMI.
What is instance metadata and what is it used for ?
It is information about the instance.
1) id
2) mac (media access control)
3) hostname
4) public and local ip
It is used because you might want to have this information to launch or access your instance. You can access the meta data of the instance with specific url:
http://169.254.169.254/latest/meta-data/
Can you edit the user data on a running instance
No you need to stop it before, remove the old ones and put in the new before relaunching
Why use user data and not run commands when the instance as started ?
For reproducibility purposes.
What are the different AMI deployment models?
Basic
Silver
Gold
Goes from basic configuration to some configuration to fully configured.
Gold AMI have shorter boot times but increased build times.
What do you get with a basic AMI
Only the OS
What is a placement group used for?
This is when you want to influence the placement of your instances in an availability zone.
To increase performance or reduce correlated failures for example
What are the three modes of placement groups for EC2 instances
Cluster: close together to reduce latency
Partition: Spread accross logcial partition to not share the same risk in the partition
Spread: Spread accross distinct harware to reduce correlated failures.
Purchasing strategies type on EC2
Purchase models => emphasize on saving
Capacity reserved models => emphasize on availability
Dedicated models => emphasize on compliance, regulation and data privacy
What is the on demand purchase model
Pay for compute capacity by the second or the hour. No commitment.
Good for spiky workloads on experimental workloads
What is the reserved purchase model
Commit for 1 or 3 years. Receive a discount. Committed workloads
What are EC2 saving plans?
A model by which you have similar disconts than a reserve instance but you commit for a $/hour in exchange.
Compute saving plans are very flexible and lets you save up to 66%
Instance Saving plans are less flexible (can’t change region) But offer bigger discounts
What is EC2 Spot?
You bid on the unused capacity of EC2 instances and get a share of the compute time when available. Can be stopped anytime by AWS with a 2 minutes notice. Billed by the second.
Good for workloads that can be stopped
What is capacity reservation for the EC2 cost model?
It lets you reserve compute capacity in a specific AZ
What are the different types of capacity reservation?
1)On demand capacity reservation:
You always have access to capacity for as long as you need it.
2)Ec2 Capacity Blocks for ML
Reserve GPU for a future date to run any ML workload
Example of strategy for cost optimization using EC2
1) Use reserved instance or saving plans for known steady state workload.
2) Use on demand instance for new or spiky workloads
3) Scale using spot instance for fault tolerance and flexible workloads
Well architected pillars for using EC2
Security
Performance efficiency
Cost optimisation
Sustainability
Security & EC2 : Best practices
1)Automate compute protection:
EC2 image builder; user data scripts; Silver or Gold AMIs to lock down secrutiy configurations
2)Control traffic at all layers:
Use VPS and security groups. Separate workloads.
Best practice for Performance efficiency and EC2:
Scale the best compute options and right size
Best practice for cost optimization with EC2:
Select the correct ressource type and the best pricing model
