Please enable JavaScript.
Coggle requires JavaScript to display documents.
EC2: Instance Storage - Coggle Diagram
EC2: Instance Storage
Elastic Block Store (EBS)
Network drive
There's latency
Mounted on one instance at a time
Locked to AZ
Move to other AZ via snapshot
Has provisioned capacity
Size in GBs
IOPS
Can increased over time
Delete on terminate
On
Default
Root EBS Volume delete when instance terminated
Other EBS volumes not deleted
Off
No volumes deleted on instance terminate
Snapshots
Backup of EBS volume
Recommended to first detach
Can copy across AZs & Regions
Features
Snapshot Archive
Cheaper storage tier
24 - 72 hours to restore
Recycle bin
1 day to 1 year
Fast Snapshot Restore
Force full init of snapshot on restore
Expensive
No latency on first use
Amazon Machine Image (AMI)
OS & software already installed
Faster config time
Locked to region
Can be copied between regions
Types
Public AMI
Custom AMI
Marketplace AMI
Creation process
Start instance and configure
Stop instance
Build AMI
Launch other instances from AMI
Instance Store
Very high IOPS
Ephemeral
Use cases
Buffer
Cache
Scratch
Temporary data
Data loss if hardware fails
Backups & replication is your responsibility
EBS Volume Types
SSD
GP2/GP3
1GB - 16TB
GP2
3 IOPS per GB
Max IOPS = 16000
GP3
IOPS: 3000 - 16000
Throughput: 125MiB/s - 1000MiB/s
IOPS and Throughput can be increased independently
General purpose SSD
IO1/IO2
High performance SSD
If more than 16000 IOPS required
Good for databases
IO2 is more durable
Max IOPS: 64K (EC2 Nitro) / 32K (all other EC2)
Size and IOPS not linked
IO2 Block Express
4GB - 64TB
Sub-ms latency
Max IOPS: 256K
IOPS:GiB ratio: 1000:1
4GB - 16TB
Supports EBS multi-attach
HDD
ST1
Use cases
Big data
Data warehouse
Log processing
Low cost HDD with high throughput
Max throughput: 500MiB/s
Max IOPS: 500
SC1
Lowest cost HDD
Max throughput: 250MiB/s
Max IOPS: 250
Cannot be boot
125GB - 16TB
Multi attach
Only IO1/IO2
Attach same volume to multiple instances
Each instance has full read/write permissions
Use cases
HA in clustered apps (ex. Teradata)
Apps must manage concurrent write
Max 16 instances
FS must be cluster aware
Encryption
Handled transparently
Minimal impact on latency
Uses KMS keys (AES-256)
Encrypted un-encrypted volume
Create EBS snapshot
Encrypt snapshot using copy
Create new volume from encrypted snapshot
Attach encrypted volume to original instance
Snapshots of encrypted volumes are encrypted
Elastic File System (EFS)
Attributes
Works with EC2 in mutli-AZ
Highly available
Scalable
Expensive
Security group to control access
Encryption at rest
Linux only (POSIX file system with NFS 4.1 FS)
Scales automatically, pay per use
Managed Network File System (NFS)
Use cases
Content management
Web serving
Data sharing
Wordpress
Scale
1000s of NFS clients
10 GB/s + throughput
Petabytes
Performance mode
General Purpose
Latency optimized
Max I/O
Higher latency
Higher throughput
Highly parallel
Throughput mode
Bursting
1TB = 50MiB/s + burst of up to 100MiB/s
Provisioned
Set throughput independent of size
Elastic
Autoscales throughput based on workload
Read: 3GiB/s max
Write: 1GiB/s max
Storage tiers
Standard
Frequently accessed files
Infrequent Access (EFS-IA)
Cost to retrieve files
Lower price to store
Enabled with Life-cycle Policy
Availability
Standard
Multi-AZ
Good for prod
One Zone
Good for dev
Backups enabled by default
Compatible with IA