AMAZON

EC2
EC2 – Elastic Computing Cloud

Root Volume: Where we install OS

By default, when we create EC2 account with Amazon, our account is limited to maximum 20
instances per EC2 region with 2 default High I/O instances.

Total 7 Types Instances.

1. General Purpose - Balanced Memory & CPU

2. Compute Optimised – More CPU than RAM
3. Memory Optimised – More RAM
4. Accelerated Category / GPU – Graphics Optimised
5. Storage Optimised – Low latency
6. High Memory Optimised – High RAM & Nitro System
7. Previous Generation

Computer Optimized Instances

This C series

Three Type – C4, C5, C5n

Previous Generation – C3

Computer Optimization Instances are ideal for compute bound application that benefit from high
performance processor.

C4 (Web server, Batch Processing, MMO Gaming, Video Encoding)

RAM: 3.75 to 60 GB

VCPU: 2 to 36

Storage: EBS only

Network Band Width: 10 GBps

C5 (Video Encoding, Gaming, Web server, High performance)

RAM: 4 to 192 GB
VCPU: 2 to 72

Storage: EBS only & NVMe SSD

Network Band Width: 25 GBps

1. C5 support maximum 25 EBS volume per instance. Where 1 EBS is Root volume.
2. C5 use Elastic Network Adaptor (ENA) which improve and enhance network activity.
3. C5 uses EC2 Hypervisor, we also called AWS Nitro System. (The Nitro Hypervisor is a
lightweight hypervisor that manages memory and CPU allocation and delivers
performance that is indistinguishable from bare metal. A hypervisor is a type of computer
software, firmware or hardware that creates and runs virtual machines. A computer on
which a hypervisor runs one or more virtual machines is called a host machine, and each
virtual machine is called a guest machine)

General Purpose Instances

General Purpose Instances provide balanced of compute Memory and Networking recourses and can
be used for variety of workload.

There have 3 types of series.

A series (Medium & Large) – A1

M series (Large) – M4, M5 (M5a, M5ad, M5d)

T series (Nano, Small, Medium, & Large) – T2 (Micro), T3, T3a

A1 – It is suitable for scale out workload. (Scale out is a type of capacity expansion concentrating on
the addition of new hardware resources instead of increasing the capacity of already available
hardware resources such as storage or processing silos). That are supported by ARM eco system.

ARM eco system of software partners provide customers a wide range of products to gets to market
faster than the competition. ARM development boards are the ideal platform for accelerating the
development and reducing the risk of new SoC designs.

Microservice is a distinct method to developing software systems that tries to focus on building
single function modules with well-defined interface and operations.

Web Server, Containerised Micro Severs, Caching Fleet, Distributed Data Stores, Application that
requires ARM Instruction set.

M4 –

VCPU: 2 – 40

RAM: 8- 160 GB

Storage: EBS Only

M5 –

VCPU: 2 – 96

RAM: 8- 384 GB

Storage: EBS & NVMe SSD

T2 – (Basic & Low performance)

VCPU: 0.5 – 35

RAM: 2- 8 GB

Storage: EBS Only

Memory Optimized Instances

R Series – R4, R5, R5ad, R5d (In Memory Caching)

VCPU: 2 – 96

RAM: 16 - 768 GB

Storage: EBS & NVMe SSD

X series – X1, X1e (SAP)

VCPU: 4 – 198

RAM: 122- 3904 GB

Storage: SSD

Z series – Z1d (Electric design, Automation) AWS nitro system. XEON processor, Up to 1.8 TB of
instances

VCPU: 2 – 48

RAM: 16- 384 GB

Storage: NVMe SSD

Storage Optimized Instances

Sequential read & Write access to very large data set on local storage. Low latency & Random I/O
operation per second. (IOPS) (30k)

I Series – I3 & I3en (Online Transaction)

VCPU: 2 – 96

RAM: 16- 768 GB

Storage: NVMe SSD

Networking Performance: 25 GBps to 100 GBps

Sequential Throughput:

16 GB / S (Read)

6.4 GB/S (I3 Write) & 8 GB/S (I3en)

D series – D2 (Parallel Processing, Map & Hadoop Computing, Log or data Processing App)

VCPU: 4 – 36

RAM: 30- 244 GB

Storage: SSD

H series – H1 (Medium) Family Feature up to 16 TB of HDD based local storage, High throughput,

VCPU: 8 – 64

RAM: 32- 256 GB

Storage: HDD

Accelerated Computing Instances

Accelerated computing instance families use Hardware accelerators, or co-processors to perform

some functions such as floating-point number calculation, graphics, or data pattern matching more
efficiently than is possible in software running CPU.

F series – F1

F1 instances offers customizable Hardware acceleration with Field Programmable Gate Arrays.
(FPGA). Each FPGA contains 2.5 million logic elements and 6800 DSP engines. Use for highly parallel
operation and data flow.

VCPU: 8 – 64

RAM:122- 976 GB

Storage: NVMe SSD

FPGA – 1 to 8

Used in – Genomics, Research, Financial Analytics, Real Time Video & Big data search

P Series

P2 & P3 – It use NVIDIA Tesla GPU

Up to 32 GB of memory per GPU which is ideal for deep learning & fluid dynamics.

P2 –

VCPU: 4 – 64

RAM: 61- 732

GPU Ram – 12-192 GB

Network BW – 25 GBps

GPU – 1-16

P3 –

VCPU: 8 – 96

RAM: 61- 768

Storage: SSD & EBS

GPU – 1-18

Note: P3 support CUDA9 and Open CL & P2 support CUD8 and Open CL 1.2

G series – G2 & G3

Optimised for Graphics application. Well suited for 3D visualisation. G3 instances use NVIDIA Tesla
M60 GPU and provide a Cost-Effective high-performance platform for graphics application.

VCPU: 4 – 64

RAM: 305 - 488

GPU Ram – 8-32 GB

Network BW – 25 GBps

GPU – 1-4

Used In – Video Creation, 3D visualisation, streaming, Graphics intense

High Memory Instances

High Memory Instances are purpose built to run large in memory database, including production
development of SAP HANA in the cloud.

U series – U6, U9, U12

High Memory Instances are Bare Metal Instances and do not run on a Hypervisor.

Only available under dedicated host purchasing category (For 3 years term)

OS directly on Hardware.

Latest Generation Intel Xeon Pentium 8176M Processor.

6,9,12 TB of Instance Memory, the largest of any EC2 Instance.

Powered by the AWS Nitro System, a combination of dedicated Hardware & Lightweight hypervisor.

Base Metal performance with direct access to host Hardware.

EBS optimised by default at no additional Cost

Network Performance: 25 GBps

Dedicated EBS Bandwidth: 14 GBps

Each instance offers 448 Logical processor.

Previous Generation Instances:

T1, M1, C1, CC2, M2, CR1, I2, HS1, M3, C3, R3

Pay per Second Option is available for Ubuntu & Linux. For Window still it is billing for hourly basis.

Need to pay bill from when we start to boot it to when it we terminate it. If we shut down the
instance the we need pay for EBS storage.

EC2 Instances Purchasing

1. On Demand (*)

For New Company, testing the business, pay per hour, short term irregular workload
that cannot be interrupted, scale up & down, testing & development of application

2. Dedicated Instances
Dedicated instances are run in a VPC on Hardware that is dedicated to a single
customer. Dedicated instances are physically isolated at the host hardware level from
instances that belong to other AWS account. It may share Hardware with other instances
from the same AWS account that are not dedicated. Pay for dedicated instances on demand
save up to 70% by purchasing reserved instances or save up to 90% by purchasing spot
instances.
3. Dedicated Host
Each dedicated host supports a single instance size and type. IT is a Physical server with
EC2 Instance. Only BYOL, Amazon Linux and AWS marketplace AMIs can be launched onto
dedicated hosts. Dedicated host can help you address compliance requirement and reduce
costs by allowing you to use your existing server bound software licenses. Pay for physical
host that is fully dedicated to running your instances and bring your existing per socket, per
core, per VM software license to rude cost. We can not migrate dedicated host to other
region and cannot change to other instances.
4. Spot Instances (*)
Spot instances are spare EC2 capacity that can save you up to 90% off of on demand
prices that AWS can interrupt with a 2 Minute notification. Spot uses the same underlying
EC2 instances as on demand and Reserved instances and is best suited for flexible workload.
Use for testing & development workload. You also have the option to hibernate stop or
terminate your spot instances when EC2 reclaims the capacity back with 2 minutes of notice.
In the case of Hibernate, your instances get hibernated and RAM data persisted. In case of
stop, your instances data shutdown and RAM is cleared.
5. Schedule Instances
Schedule Reserve Instances enable you to purchase capacity reservation that recur on
daily, weekly or monthly basis with a specified short time and duration for a One-year term.
You reserve the capacity in advance so that you know it is available when you need it. You
 pay for it time that the instances are schedule even if you don’t use that. It is good for
workload that do not run cautiously but run on a regular basis.
6. Reserved Instances (*)
It provides a signification discount (up to 75%) compare on-demand. If you want to take
advantage of the capacity reservation, then you should buy RI in a specific availability Zone.
We can reserve that for 1 to 3 years. 3 types (standard (75%), convertible (54%), schedule)

No need pays extra charge to convert instances. Can change instances through EC2
management console or get reserved instances management quota API.

EC2 Access

1. To access instances, you need a key and key pair name.

2. You can download private key only once.
3. The public key is saved by AWS to match it to the key pair name, and private key when
you try to log in to the EC2 instances.
4. Without key pair you cannot access instances via RDP or SSH (Linux)
5. There is 20 ECS instances soft limited per account, you can submit request to aws to
increase it.

EC2 Status Check

1. By default, AWS EC2 service perform automated status check every one minute.
2. This is done on every running EC2 instances identify any H/W and S/W issue.
3. Status check is built into the AWS EC2 instance.
4. They cannot be configured, deleted or disable.
5. EC2 services can send its metric data to AWS cloud watch every 5 minutes enabled by
default.
6. Enable detailed monitoring is chargeable and sends metric in every 1 minute.
7. You are not charged for EC2 instances if they are stopped, however attached EBS volume
incur charge.

EBS Backed EC2 Instances: Root Volume EBS type where we have OS

Instances Store Backed EC2: Root volume Instances Store Type

When You Stop an EBS Backed ECS Instance

1. Instances perform a shutdown.

2. State changes from Running -> Stopping
3. EBS volume remain attached to the instance
4. Any Data cached in RAM or instances store volume is gone
5. Instances retain its IPv4 or any IPv6 address
6. Instances releases its public IPv4 address back to AWS pool
7. Instances return its elastic IP address

When You terminate a running instance

1. State changes Running -> Shutting Down -> Terminated

 2. During shutting down and terminated states you do not incur charges
3. By default, EBS Root devices volumes are deleted automatically when EC2 instances are
terminated
4. Any additional (Non boot/boot) volumes attached to the instances by default, persist the
instances in terminated
5. You can modify both behaviours by modifying the ‘Delete on Termination attribute of
any EBS volumes during instances launch or while running.
6. Enable EC2 Termination protection against accidental termination.

EC2 Metadata

1. This is instances data that you can use to configure or manage the instance.
E.g. IPv4 address, IPv6 address, DNS hostname, AMI-id, Instance ID, Instance Type, Local
host name, Public keys, Security Group.
2. Metadata can be only viewed from within the instances itself i.e. you can have to log in
to the instances
3. Metadata is not protected by encryption, anyone that has access to the instances can
view the data
4. To view instances Metadata
Get http://169.254.169.254/latest/Metadata

Instances User Data

1. Data supplied by the user at instance launch in the form of a script to be executed
during the instances boot.
2. User data is limited to 16 KB
3. You can change user data by stopping EC2 first
4. User data is not encrypted

EC2 Bare Metal Instances

1. Non-Virtualised Environment
2. OS runs directly on Hardware
3. Suitable for Licensing Restricted Tier-1 business critical application

EBS Volume

1. Most Common Replicate with A-Z

2. EBS volumes attached at launch are deleted when instances terminate
3. EBS volumes attached to a running instance are not deleted when instances are
terminated but are detached with data interact

Instances Storage
1. Physically attach to the host server
2. Data not lost when OS is rebooted
3. Data lost when underlying drive fails, instances is stop or terminated
4. You cannot detach or attach to another instance
5. Do not Relay on for valuable long-term data