AWS EBS & EFS Elastic Block Storage & Elastic File System

EBS is a network drive you can attach to ONLY ONE INSTANCE at any one time

 

 

important: EBS is not a network file system!

is bound to an AZ

 

think of it like a network USB stick

 

30GB per month free of type SSD or magnetic GP2 or GP3 volume

 

EBS can have some latency on the AWS network

 

can be detached from one EC2 and attached to another, eg for failovers, big advantage!

 

but cannot be moved to other AZs, unless you do a snapshot and then move.

 

have to provision in advance and the level of IOPS you want

 

you pay according to this after the first 30GB.

 

 

BUT you can have 2 EBS volumes attached to an EC2 – that is not limited.

 

however they are bound to an AZ

 

they can be attached on demand, do not have to be actively attached.

 

 

EBS Delete on Termination attribute – is enabled by default for EBS root volume

 

but not for other EBS volumes as it is disabled for the latter by default

 

you can change this…
advantage: to preserve root volume when EC2 instance is terminated.

 

You can transfer an EBS volume to another region or az by means of using snapshots.

 

You can move snapshots to archive which is much cheaper.

 

Snapshots are usually gone once deleted, but if you have recycle bin enabled then you can retrieve them for a limited time period according to retention rule you set.

 

 

EBS Volume Types

 

6 types:

 

gp2 /gp3 SSD general purpose balances price/performance

 

io1/io2 SSD high performance for mission critical low latency/high throughput workloads

 

st1 HDD low cost HDD for frequently accessed throughput intensive workloads

 

sc1 HDD lowest cost HDD for less frequently accessed workloads

 

EBS volumes are by size, throughput, iops

 

NOTE: only gp2/gp3 and io1/op2 can be used as boot volumes

 

 

General Purpose SSD

 

cost-effective storage

 

system boot vols, virtual desktops, dev and test env

 

1GiB – 16TiB

 

gp3

 

3000 iops and 125 miB/s

 

can go up to 16k iops and 1000 MiB/s

 

gp2

 

small vols can do burst iops to 3000

 

size of volume and iops are linked max iops is 16k

 

3 iops per GB means at 5,334 GB we have max iops

 

 

Provisioned iops ssd

 

for critical biz apps with sustained iops performance

or apps that need more than 16k iops

 

good for db workloads

 

io1/io2 4GiB – 16TiB

 

max piops 64k for nitro ec2 and 32k for others

 

can increase piops indep of storage size

 

io2 gives more durability and more iops per gib

 

io2 block express 4gib – 64tib

 

very low latency
max piops 256k

 

 

with hdd:

 

cannot be a boot vol

125 mib to 16tib

 

throughput optimized hdd st1:

 

big data, data warehousing and log processing

 

max througput is 500 mib/s max iops 500

 

cold hdd sc1

 

for infreq accessed data
where low cost is important
max throughput is 250 mib/s max iops 250

 

 

 

you *dont* need to know these details for the exam, but be aware of the main difference in the variations

 

 

 

EBS Multi-Attach – for io1/io2 family

 

attaches same ebs volume to multiple instances in same az at same time

 

each instance has full r/w permissions to the vol

 

use case:

 

for high app availability in clustered linux apps eg teradata

 

apps must be able to manage concurrent write ops

 

only thus for very specific workloads,

 

you must use a cluster aware file system ie NOT ext4 xfs etc.

 

 

EFS Elastic File System

 

is a managed NFS-type system that can be mounted on multiple EC2s in multiple AZs.

 

highly available, scalable, but expensive, you pay per usage per gigabyte

 

use cases:

 

content management,

 

web and data sharing, wordpress

 

uses nfs4.1

 

uses security groups to control access

 

important!

compatible only with linuxbased ami’s and not windows!

 

can enable enctryption at rest using kms

 

is a posix linux system with standard file api

 

scales automatically by itself!

 

exam question:

 

efs performance and storage classes:

 

efs scaling:

 

can support 1000s of concurrent nfs clients 10gb throughput

 

can grow to petabyle size automatically

 

performance mode

 

– set at efs creation time
general purpose default – latency sensive use cases such as webserver, cms etc

 

to maximum i/o — gives higher latency, highly paralleluse for this io1 : this is best for big data applications, eg media processing etc

 

throughput mode
— bursting 1tb = 50mib/s and up to 100 mib/s

 

provisioned: set your throughput regardless of storage size eg 1 1gib/s per 1tb storage

 

 

EFS Storage Classes

 

you can set up storage tiers for lifecycle management

 

eg move to another tier after N days…

 

 

 

– standard tier – used for frequently accessed files
– infrequent access tier (IA) -efs-ia: costs to retrieve files

 

but lower price to store, enable efs-ia by means of a lifecycle policy

 

 

Availability and Durability of EFS

 

2 options:

 

standard: can set up EFS to be multi AZ
one-zone: you use one AZ only, backups are default enabled, compatible with IA

 

90% cost saving

important:
exam will ask you which tier /storage class you should use for which use case, and you need to be aware of the cost implications!

 

 

 

Differences Between EBS & EFS

 

must know for exam!

 

EBS

 

can only be attached to only one instance at a time
are locked into an AZ

 

gp2. io increases if disk size increase
io1: can increase the io independently.

 

to migrate an EBS across AZs

 

– take a snapshot
– restore the snapshot to the other desired AZ

 

note that ebs backups use up io and so you should not run them when your app has heavy traffic overhead

 

also, root ebs volumes get terminated by default if or when the ec2 instance gets terminated! very important to be aware of this
– but you can disable this

 

 

EFS by comparison:

 

can mount on 100
s of instances across AZs!

 

it is multi-AZ, multi-client/instance

 

can be used to share data

 

only available for linux posix, not windows!

 

 

efs more expensive than EBS but can use efs-ia to reduce costs

 

 

so: efs is more for multi instances

 

ebs is more for one instance

 

 

and

 

Instance Store

 

instance store: is an ephemeral local instance drive just for an instance – you lose it with the instance when the instance is deleted.