Discuss EFS performance modes and storage classes.
EFS offers two performance modes (General Purpose and Max I/O), two throughput modes (Elastic and Provisioned), and four storage classes (Standard, Standard-IA, One Zone, One Zone-IA) that allow you to balance performance, availability, and cost based on your workload's access patterns.
Amazon EFS is designed to support a wide variety of workloads from small, latency-sensitive applications to large-scale parallel processing jobs. Performance and cost can be fine-tuned by selecting the appropriate performance mode, throughput mode, and storage class. These settings can be changed after creation, giving you the flexibility to optimize over time.
- 1
General Purpose (default and recommended) — lowest per-operation latency (sub-millisecond for metadata operations). Suitable for latency-sensitive use cases like web serving, content management, home directories, and general application workloads. Supports up to 35,000 IOPS.
- 2
Max I/O (legacy) — optimized for applications that require the highest possible aggregate throughput and IOPS, with a trade-off of slightly higher latency per operation. Designed for massively parallel workloads like big data analytics, media processing, and genomics. Note: AWS recommends using Elastic Throughput mode with General Purpose instead of Max I/O for new file systems.
- 1
Elastic Throughput (recommended) — automatically scales throughput up or down based on your workload. No configuration needed. Burst to 3 GiB/s for reads and 1 GiB/s for writes. Best for unpredictable or spiky workloads. You pay only for the throughput you use.
- 2
Provisioned Throughput — you specify a fixed throughput level (in MiB/s) independent of the amount of data stored. Use when your workload requires consistently high throughput that exceeds what Elastic mode delivers at your storage size. Billed for the provisioned amount regardless of actual usage.
- 3
Bursting Throughput (legacy) — throughput scales with the size of the file system. Baseline of 50 MiB/s per TiB stored, with burst credits that allow bursting to 100 MiB/s per TiB (minimum 100 MiB/s). Not recommended for new file systems.
- 1
EFS Standard — default storage class. Data is stored redundantly across multiple AZs. Highest availability and durability. Ideal for frequently accessed files. Higher cost per GB.
- 2
EFS Standard-IA (Infrequent Access) — same multi-AZ durability as Standard but at up to 92% lower storage cost. There is a per-request retrieval fee when files are accessed. Ideal for files not accessed every day.
- 3
EFS One Zone — data is stored in a single AZ only. Lower cost than Standard (about 47% cheaper). Risk of data loss if the AZ fails. Best for dev/test environments or non-critical data that can be rebuilt.
- 4
EFS One Zone-IA (Infrequent Access) — single AZ storage for infrequently accessed files. The lowest cost EFS storage class (up to 95% cheaper than Standard). Suitable for backups or secondary copies of data in non-critical environments.
Scenario Questions
0-2 years experience
- 1You need to mount an EFS file system for a web server handling a few hundred requests per second. Which performance mode would you pick and why?
- 2If you store backup files that are accessed rarely, which EFS storage class would you select and what cost impact does it have?
- 3What happens if you use the Max I/O performance mode for a low‑latency application?
2-5 years experience
- 1Your team notices high latency spikes when many EC2 instances read from the same EFS. How would you diagnose whether the performance mode is the cause, and what changes could you make?
- 2You have a mixed workload where some files are hot and others are cold. How would you configure EFS storage classes and performance mode to balance cost and throughput?
- 3During a deployment the EFS mount fails with 'access point limit exceeded'. Explain how performance mode and throughput mode interact and what you would adjust.
5-8 years experience
- 1Design a data‑processing pipeline where multiple EC2 workers write large files to EFS. Explain your choice of performance mode, throughput provisioning, and storage class, and how you would monitor and scale it.
- 2A legacy application expects POSIX semantics but is hitting I/O throttling on a General Purpose EFS. Propose a migration plan to Max I/O while minimizing downtime and ensuring data consistency.
- 3Explain edge cases where switching from General Purpose to Max I/O could degrade performance, and how you would mitigate them in a high‑concurrency environment.
8+ years experience
- 1Your organization is consolidating several regional file stores into a single EFS for a global analytics platform. Discuss the architectural trade‑offs of performance modes, storage classes, and cross‑AZ latency, and outline a migration strategy that addresses cost, performance, and data residency.
- 2When planning a multi‑tenant SaaS platform on AWS, how would you leverage different EFS performance modes and storage classes to isolate tenant workloads, control costs, and meet SLAs?
- 3What long‑term governance and monitoring practices would you put in place to ensure that EFS performance mode and storage class choices continue to meet evolving workload patterns across many teams?
Follow-up Questions
- How does bursting throughput differ between General Purpose and Max I/O?
- What CloudWatch metrics would you monitor after switching performance modes?
- Is it possible to use both Standard and Infrequent Access storage classes in the same file system, and how?