Elastic File System (EFS)

EFS Architecture

🔧 EFS = AWS implementation of NFSv4
- 💡 NW File System (NFS) is a fairly common storage standard for shared files
- Provides NW-based FSs which can be mounted within 1+ Linux EC2 instance(s)
  - Allows storing media outside of EC2 instances → stateless instances
  - Shared → One FS can be used by multiple Linux instances at once
- Private service, FS is accessed via mount targets inside VPC subnets
  - Can be accessed via hybrid NWing from on-premises: VPC peering, VPN or DX
  - To ensure HA in a region, put mount targets in multiple AZs (same as with NATGWs)
Files in EFS exist separately from EC2 instances
- 💡 Just like EBS volumes exist separately from EC2
  - EFS is file storage and EBS is block storage
- Data is not lost if instances are deleted and recreated
  - This makes EC2 instances stateless, which is usually desired
‼️ EFS is ONLY for Linux instances!!
Linux uses a tree structure for its FS
- Devices can be mounted into directories/folders in the hierarchy
- A FS could e.g. be mounted into a directory called /nfs/media
FSs use POSIX permissions
- 💡 POSIX = a standard for interoperability which is used in Linux
  - A POSIX permissions FS can be understood by any Linux distribution
  - Additional info: https://en.wikipedia.org/wiki/File_system_permissions
EFS Architecture Diagram

Throughput modes:
1. Bursting
  - Has a burst pool, but throughput scales with FS size
    - 💡 Akin to GP2 volumes in EBS
      - The more data you store in the FS, the better performance you get
  - Default (provides basic performance requirements)
2. Enhanced
  1. Elastic
    - Throughput scales automatically, only pay for what you use
    - For workloads with unpredictable I/O
  2. Provisioned
    - Throughput requirements can be specified separately from FS size
      - 💡 Akin to IO1 volumes in EBS
    - More flexible, but not used by default
- Throughput Modes - UI Screenshot
Performance modes:
1. General Purpose
  - 💡 Default for 99.99% of use cases
  - Ideal for latency-sensitive use cases, web servers, CMSs, home directories or even general file serving
2. Max I/O
  - Scales to higher levels of aggregate throughput and IOPS…
    - …with trade-off of increased latency
  - Ideal for highly parallel apps
    - e.g. big data media processing, scientific analysis…
Ref (extra info and numbers): https://docs.aws.amazon.com/efs/latest/ug/performance.html
- 💡 No need to remember raw numbers for exam