Professional Cloud ArchitectManage implementationHard
A research institution is processing massive datasets (petabytes) for scientific simulations using custom-built HPC applications. These applications are highly parallelized and require extremely low-latency, high-bandwidth storage shared across hundreds of compute nodes. The data is accessed frequently during simulations and needs to be durable. Which Google Cloud storage solution is best suited for this demanding workload?
- ACloud Storage (Standard Storage class)
- BCloud SQL
- CPersistent Disk (SSD)
- DFilestore (Enterprise tier)
Show answer & explanationAnswer & explanation
Correct answer: D. Filestore (Enterprise tier)
Filestore Enterprise provides fully managed, shared NFS file storage with high performance and availability, specifically designed for demanding workloads like HPC that require shared file systems across many compute nodes. Its Enterprise tier offers higher IOPS and throughput suitable for petabyte-scale, low-latency access.
Why the other options are wrong
- A. Cloud Storage is object storage, not a shared file system, and while durable and scalable, it doesn't offer the POSIX-compliant, low-latency, high-bandwidth shared file system access required by HPC applications.
- B. Cloud SQL is a relational database service, entirely unsuitable for storing and sharing petabytes of raw simulation data for HPC applications.
- C. Persistent Disk (SSD) offers high performance but is block storage attached to a single VM (or shared between a very limited number of VMs in specific scenarios), not a shared file system across hundreds of compute nodes.
Google Cloud Filestore
A fully managed, high-performance file storage service for Google Cloud, offering Network File System (NFS) protocol support, ideal for workloads requiring shared filesystems.
- Provides POSIX-compliant shared file systems.
- Offers different service tiers (Basic, High Scale, Enterprise) for varying performance needs.
- Suitable for HPC, GKE Persistent Volumes, media rendering, and general file sharing.
- Enterprise tier provides multi-zone availability and higher performance.
Memory trick: Files for fast, shared science.