- 1. EBS
- a. Storage Types
- i. Instance Store
- 1. Ephemeral
- ii. S3
- 1. Standard
- 2. RRS
- iii. EBS
- 1. GP-SSD
- 2. PIOPS
- 3. Magnetic
- iv. EFS
- 1. Network Attached Storage
- 2. Managed file system
- 3. Only supports Linux instances
- i. Instance Store
- b. EBS Characteristics
- i. Billed on storage capacity and I/O
- ii. Does not need to be attached to an instance
- 1. Detach and re-attach
- iii. Transferred between availability zones
- iv. EBS volumes are designed for an annual failure rate (AFR) of 0.1 – 0.2 %
- v. The average failure rate for a hard drive is 5-10%
- vi. EBS volume data is replicated across multiple servers in an availability zone
- vii. SLA: 99.95%
- c. EBS Volume Types
- i. General Purpose (SSD)
- 1. Use cases
- a. System boot volumes
- b. Virtual desktops
- c. Small to medium DBs
- d. Dev and test
- 2. Volume Size: 1GB – 16TB
- 1. Use cases
- ii. Provisioned IOPS (SSD)
- 1. Use cases
- a. I/O intensive
- b. Relational databases
- c. NoSQL databases
- 2. Volume Size: 4GB – 16TB
- 1. Use cases
- iii. Magnetic
- 1. Use cases
- a. Infrequent data access (archiving)
- b. SSD has come down in price
- 2. Volume size: 1GB – 1TB
- 1. Use cases
- iv. Features
- 1. Bursting will usually happen when an instance starts to increase boot time
- 2. 48,000 max IOPS
- 3. 800 Mbps max throughput
- i. General Purpose (SSD)
- d. Increasing IOPS performance
- i. Stripping multiple volumes into a RAID-0 group
- 1. Cheaper alternative to provisioned IOPS
- 2. This works because each disk has its own max IOPS
- 3. When you use RAID-0 your max IOPS is the summation of the max for all your disks
- 4. With RAID-0 you do risk losing a drive if one drive fails
- 5. RAID-5 or RAID-6 might be more reliable
- 6. Function of the guest OS
- 7. EBS is not designed for long term storage, S3 targets that use case
- ii. EBS-Optimized instances
- 1. Dedicated capacity for Amazon EBS I/O
- 2. 500 – 4000 Mbps performance boost
- 3. Supported instance families:
- a. C, D, M, R, G
- b. Varies from type to type, need to look up in AWS documentation to confirm
- 4. GP-SSD within 10% of baseline and burst performance 99.9% of the time
- 5. PIOPS within 10% of provisioned performance 99.9% of the time
- 6. EBS optimized are designed for all EBS volume type
- 7. Additional hourly fee
- i. Stripping multiple volumes into a RAID-0 group
- a. Storage Types
- 2. Snapshots
- a. Characteristics
- i. Point-in-time snapshots
- 1. Taking a snapshot of a running instance: Nothing going on will be taken into account, not a clean cut over.
- 2. Power down and then cut over.
- ii. Support incremental snapshots
- 1. Only the changes between the last snapshot (saves a lot of space)
- 2. Billed only for the changed blocks
- iii. Deleting a snapshot removes only the data not needed by any other snapshot
- iv. Should I delete this one first or not
- 1. AWS keeps track of enough information about the snapshot to know which snapshots are dependent on others
- v. EBS leverages S3 for snapshot storage
- 1. Not necessisarily inside a bucket…its storage in a special location and accessible from you instances through the management portal
- i. Point-in-time snapshots
- b. Features
- i. Resizing EBS volumes
- ii. Sharing EBS snapshots
- iii. Copying EBS snapshots across regions
- iv. Lazy loading
- 1. Very slow due to being stored on S3
- 2. Pre-warming of EBS volumes
- a. Process changes on Windows and Linux
- i. “dd for windows” will read all blocks at the same time (accelerate a recall process from S3)
- a. Process changes on Windows and Linux
- v. Create AMI images from snapshots so you can launch instances from it
- a. Characteristics
Sky Cliffs 
Pingback: AWS Certified SysOps Administrator Exam: Study Guide | Sky Cliffs