Examkin

Question

Difficulty: HardHigh-Performing and Scalable Storage Solutions

A financial analytics firm runs a real-time risk simulation workload on a fleet of Amazon EC2 Linux instances deployed in a single Availability Zone. The application requires a shared file storage solution that can be mounted simultaneously by all instances using the NFS protocol. The workload performs highly intensive metadata operations and requires consistent sub-millisecond latencies of less than 1 ms1 \text{ ms}, a minimum throughput of 2,500 MB/s2,500 \text{ MB/s}, and up to 100,000 IOPS100,000 \text{ IOPS}. Which storage solution will meet these requirements with the lowest latency?

  1. A
    Amazon EFS configured with Elastic Throughput
  2. B
    Amazon EBS io2 volumes with Multi-Attach enabled
  3. Amazon FSx for OpenZFS file systemsAnswer
  4. D
    Amazon FSx for Lustre persistent file systems

Answer

Amazon FSx for OpenZFS file systems
The correct file system choice provides fully managed shared file storage using the NFS protocol with sub-millisecond latencies down to 0.1–0.2 ms0.1\text{--}0.2 \text{ ms}. This solution supports up to 350,000350,000 IOPS and 10 GB/s10 \text{ GB/s} throughput, making it ideal for low-latency, high-frequency database and financial analytics workloads in a single Availability Zone.

Step-by-Step Solution

1
Analyze the access requirement and protocol.
The storage must be shared, accessed by multiple Linux EC2 instances concurrently, and support the NFS protocol.
This rules out standard EBS volumes (which are single-instance only) and points towards shared file systems (EFS or FSx).
2
Evaluate latency and performance requirements.
The file system must sustain sub-millisecond latencies (less than 1 ms1 \text{ ms}), up to 100,000 IOPS100,000 \text{ IOPS}, and 2,500 MB/s2,500 \text{ MB/s} throughput.
Amazon EFS has a standard metadata latency profile of 1–3 ms1\text{--}3 \text{ ms}, which is too high for this workload.
3
Compare FSx file system types for NFS low-latency workloads.
Amazon FSx for OpenZFS provides sub-millisecond latencies (around 0.1–0.2 ms0.1\text{--}0.2 \text{ ms}) over standard NFS and meets all IOPS and throughput constraints.
FSx for Lustre is a parallel file system that requires a Lustre client and is optimized for scale-out parallel processing rather than ultra-low latency NFS metadata operations.

Key Concept

Low-latency shared file storage optimization using Amazon FSx for OpenZFS
Rate this question

Topics

High-Performing and Scalable Storage Solutions

More questions from this topic