EBS Block Storage

EBS provides block storage volumes that attach to EC2 instances as boot or data volumes.

gp3 (General Purpose SSD) is the default—3,000 IOPS and 125 MB/s baseline, scalable to 16,000 IOPS and 1,000 MB/s independently. Best price-performance for most workloads. gp2 is previous generation—3 IOPS per GB (min 100, max 16,000). Migrate to gp3 for better performance and cost.

io2 Block Express delivers sub-millisecond latency and up to 256,000 IOPS per volume—designed for mission-critical databases (Oracle, SAP). io1 provides up to 64,000 IOPS for high-performance needs.

st1 (Throughput Optimized HDD) for large, sequential I/O—data warehousing, log processing, MapReduce. sc1 (Cold HDD) for infrequent access—lowest cost, ideal for archival. Previous gen: Magnetic (standard) for boot volumes of older instance types.

EBS Snapshots are incremental backups stored in S3—only changed blocks are copied after initial snapshot. Create snapshots for disaster recovery, migration, and compliance. Snapshots are regional but can be copied across regions.

Snapshot Lifecycle Policies automate retention—define policies to retain daily/weekly/monthly snapshots and delete old ones. Create policies per volume or tag-based. Lifecycle Manager also creates EBS-backed AMIs automatically.

EBS Encryption uses AWS KMS to encrypt volumes and snapshots. Enable default encryption on accounts to encrypt all new volumes. Encrypted volumes and snapshots maintain encryption chain—snapshots of encrypted volumes are encrypted. Encryption is free but requires KMS key management.

Fast Snapshot Restore (FSR) initializes snapshots instantly—no lazy loading of data. Enable FSR for critical volumes where immediate full performance is required after restore.

Multi-Attach allows a single io1/io2 volume to be attached to multiple EC2 instances in the same AZ. Requires instances that support NVMe and clustered applications handling concurrent writes (e.g., Oracle RAC). Use with shared-nothing architectures.

Elastic Volumes dynamically modify volume type, size, and performance without detaching. Increase size online, upgrade gp2→io1, or change IOPS—no downtime. Modify through console, CLI, or SDK. Performance changes take effect in minutes.

IOPS and throughput are independent: IOPS = operations per second, throughput = MB/s data transfer. For databases: provision high IOPS. For data processing: provision high throughput. Monitor VolumeReadOps, VolumeWriteOps, VolumeThroughputPercentage in CloudWatch.

Multi-Attach considerations: only supported on io1/io2, NVMe instances, same AZ. Application must handle concurrent access—EBS doesn't provide cluster coordination. Use with caution for write-heavy workloads.

Right-size volumes: monitor utilization and resize over-provisioned volumes. Use gp3 instead of gp2—gp3 provides 3,000 baseline IOPS and 125 MB/s throughput regardless of size, and is 20% cheaper.

Snapshot management: delete unused snapshots, use lifecycle policies to automate retention. Snapshots of deleted volumes are retained—set up cleanup automation. Copy snapshots to another region for DR.

Delete on termination: set DeleteOnTermination=true for non-persistent data (dev/test). Set false for data volumes that survive instance termination.

Performance monitoring: use CloudWatch EBS metrics to identify bottlenecks. BurstBalance (gp2/st1/sc1) indicates remaining burst credit. VolumeQueueLength shows pending I/O. Scale up before queue depth impacts performance.

EBS-optimized instances: use EBS-optimized instances for consistent IOPS/throughput to EBS volumes. Most modern instance types are EBS-optimized by default.