Amazon FSx for OpenZFS customers

Bronto uses Amazon FSx for OpenZFS to query petabytes of telemetry in seconds.

Bronto operates an intelligent data platform for observability, including AI-powered services for root cause analysis that run on Amazon Bedrock and Amazon Bedrock AgentCore.

Challenge
Bronto's multi-tenant SaaS platform ingests petabytes of telemetry, enriches it at ingestion, and makes it searchable across 12 months of history by default. The design of the storage system is a key factor in its high performance: ingestion is continuous and write-heavy, and query speed depends on low-latency index reads. Bronto needed a highly reliable and available storage layer that could absorb streaming appends, serve indexes fast, and keep client configuration and management to a minimum.

Solution
Bronto built its platform around Amazon FSx for OpenZFS as the high-performance layer in front of long-term Amazon S3 storage. Telemetry is written to FSx for OpenZFS as it arrives, using append-capable POSIX files, keeping write latency low while data is still streaming in. FSx for OpenZFS consolidates small writes, and raw data moves to Amazon S3 once it reaches a target file size or age, while index files stay on FSx for OpenZFS much longer. Many queries use these indexes to prune across terabytes without reading raw data, or return pre-computed aggregates held in the index. Queries that do need raw data fan out to tens of thousands of AWS Lambda functions reading from Amazon S3 in parallel. Accessed over standard NFS with no client-side kernel dependency, FSx for OpenZFS delivers consistent low latency on the small, random index reads on the critical path of every query.

Outcome
Bronto answers questions across petabytes of data in seconds. Because FSx for OpenZFS is fully managed, Bronto runs stateless service instances instead of hundreds of instances with large disks attached.

• Performance: Most searches complete in under a second, with FSx for OpenZFS carrying the latency-sensitive index path that sets how fast the platform feels.
  
• Managed simplicity: Index and recent data live on FSx for OpenZFS rather than local disks, so Bronto's service instances hold no state.
  
• Operations: With stateless instances, every service re-rolls weekly onto the latest AWS golden images.

"Our entire architecture rests on being able to read an index quickly. We write telemetry into FSx for OpenZFS as it arrives and we keep our index files there, and that is what lets us answer questions across hundreds of terabytes of data in seconds. The fact that it is fully managed matters just as much. It means we don't have to run a large number of instances with very large disks attached."

David Tracey, CTO, Bronto

Amdocs achieves efficient production Oracle data-warehouse operations and high-performance storage with Amazon FSx for OpenZFS.

Amdocs is a leading global provider of software and services for communications, media, and financial services organizations, helping them deliver digital transformation and enhanced customer experiences.

Challenge: Amdocs needed to meet strict SLA requirements for their daily production Oracle data-warehouse extracts, which had to be completed before 6:00 AM each morning. They required efficient database infrastructure management to maintain high-quality data delivery through nightly database refreshes and support real-time decision-making through recurring extraction tasks.

Solution: By implementing Amazon FSx for OpenZFS, Amdocs gained access to high-performance storage with up to 10GB/s throughput and 400K IOPS from disk. The solution provided zero-copy snapshots, efficient LZ4 compression, and block-based replication capabilities. This enabled continuous database replication and efficient point-in-time image creation for their data extraction processes.

“Amazon FSx for OpenZFS delivers exceptional reliability, blazing performance, and powerful data reduction capabilities. Its efficient snapshot features streamline data management. For Amdocs, this proved to be a game changer—enabling the seamless implementation of our production data-warehouse processes in the cloud.”

Octavian Paul Rotaru, Principal Technical Architect, Amdocs

Amdocs uses Amazon FSx for OpenZFS to create high-performance cloud-scale database copies while reducing capacity utilization by 60%.

Amdocs is a leading software & services provider to communications & media companies, providing innovative solutions, intelligent operations & delivery .

Challenge: Amdocs faced operational, cost, and management challenges when creating database copies.Amdocs needed instantaneous, high-performance (>10GB/s of throughput) copies of these databases for creating reports, test/dev, and data warehouse processes. In addition, they needed database copies for production support, performance testing, and other operational tasks.

Solution: Amdocs reduced costs and operational complexity with the simplicity and speed of snapshots/cloning of Amazon FSx for OpenZFS. By using Amazon FSx for OpenZFS for databases for downstream reporting, data warehouses, and testing, Amdocs was able to eliminate the need to implement database-level replication for the downstream database copies. Amazon FSx for OpenZFS supports Multi-AZ, delivering the maximum performance even when the data access is coming from a single client, as is the case for relation databases. Amdocs has the ability to configure capacity, throughput, and IOPS independently to optimize cost and pay only for what they use with FSx for OpenZFS. The built-in compression reduces cost for high-performance databases. Combined with the efficiency of OpenZFS thin snapshots and clones, Amdocs was able do more with less storage and reduce their overall storage cost.

"Amazon FSx for OpenZFS provides us an excellent platform for our cloud deployments, with no compromise on availability and is a cost-effective shared storage solution. The advanced compression features of Amazon FSx for OpenZFS reduced the storage footprint of our relational databases by 60%. Using Amazon FSx for OpenZFS snapshots to create downstream copies of the databases together with compression resulted in an overall data reduction ratio of more than 5x. The in-memory cache and NVMe cache of Amazon FSx for OpenZFS enhances the performance, allowing us to achieve file system read throughputs that exceeded the provisioned value up to 2.25x depending on the workload. Nothing is more difficult, than to be able to choose the right technology for a project. We tested many options before we decided, but once we started to use Amazon FSx for OpenZFS we immediately understood that it is the most suitable solution for our needs."

Octavian Paul Rotaru, Principal Architect – Amdocs Technology

Astera Labs eliminates complexity and experiences instantaneous high performance storage in their semiconductor environment with Amazon FSx for OpenZFS.

Astera Labs, Inc. is a leader in purpose-built connectivity solutions that remove performance bottlenecks throughout the data center.

Challenge: Astera Labs’ was using a self-managed OPENZFS NFS server, and was limited by the total amount of storage that its deployed architecture could support. Capacity expansion was a challenge as it hit per instance EBS limits. At the same time, Astera Labs’ resources were needed to manage the storage solution, including backup/restore, which took up time and resources that could be better spent on its core competencies.

Solution: With Amazon FSx for OpenZFS, Astera Labs experienced easy setup, consistent high-performance, and access to virtually unlimited storage capacity. The company is now able to launch filesystems of 100s of terabytes in a matter of seconds, while also gaining increased throughput resulting in faster simulation times, easier backup/restore, and overall ease of management.

"The simplicity, scale, and durability of Amazon FSx for OpenZFS allows us to expand our semiconductor verification and simulation environment. We have eliminated manual configuration, tuning and troubleshooting of self-managed NAS by moving to a fully managed FSx solution on AWS. We are confident that FSx for OpenZFS will help reduce our time to tapeout."

 Kalyan Mulam, SVP of ASIC Engineering - Astera Labs

Life is Tech! leverages Amazon FSx for OpenZFS to improve performance by 90% to provide an effective learning environment for over 500,000 middle and high school students.

Life is Tech! is a developer of education curriculum and a learning platform intended for elementary and junior high school students.

Challenge: Life is Tech! has many small files optimized for each student's individual learning which requires a storage system with low latency, high performance, and redundancy. The performance and operational burden of file storage had become an issue to support the ever-increasing number of users and learning content.

Solution: By using Amazon FSx for OpenZFS, a fully managed service, it was possible to achieve both stable performance and low operational load even when a lot of students are accessing their own small files.

"Storage latency due to rapid increase of users was a major issue for us from the self-managed storage on Amazon EC2. Using the fully managed Amazon FSx for OpenZFS we improved performance by 90% and reduced cost by 75%. Our operational load has significantly reduced helping us focus on service improvement tasks. The simplicity of scaling performance and capacity with Amazon FSx for OpenZFS are the foundations for providing an effective learning environment for over 500,000 middle and high school students across Japan."

Masayoshi Asakura, Life is Tech, Inc. Infrastructure/SRE Group Manager

Rev uses Amazon FSx for OpenZFS to train speech recognition machine learning (ML) models faster, while reducing operational costs by up to 30%.

Rev is a leader in automatic speech recognition intelligence, providing transcription, captioning, and subtitle services to over 170,000+ customers across industries. Rev processes terabytes of audio data using a powerful GPU compute cluster, which relies on shared file storage.

Challenge: Rev was looking for a fully managed high-performance file storage solution to reduce the operational complexity of managing and maintaining its own NFS file servers.

Solution: Using Amazon FSx for OpenZFS, Rev deploys fully managed, cost-effective NFS-accessible storage that seamlessly replaces its self-managed NFS servers. Amazon FSx for OpenZFS provides consistent low-latency storage to accelerate its ML training workloads, providing over 70% more throughput than its self-managed storage while improving visibility with integrated Amazon CloudWatch metrics.

"With our self-managed NFS storage, our data scientists would encounter performance bottlenecks that slowed down their ability to develop and train models, and we often had to manually reconfigure our file servers to adapt to our changing needs. Amazon FSx for OpenZFS addresses these bottlenecks and we can now scale our file system performance with the click of a button. We’ve eliminated the manual work of maintaining our own file servers, while reducing costs by nearly 30%."

Victor Yap, Senior ML Ops Engineer – Rev

Hatch uses Amazon FSx for OpenZFS to deliver interactive high-resolution 4K gaming experiences to players and spectators around the world.

Hatch Entertainment is a cloud-native gaming company offering mobile-optimized game streaming technology that enables players to launch high-quality games on demand – no downloads, updates, or installations required.

Challenge: To deliver this user experience, Hatch needed a high-performance, cost-effective shared storage solution to power its 4K cloud gaming technology for players and spectators around the world.

Solution: Using Amazon FSx for OpenZFS allows Hatch to deliver simple Network File System (NFS) storage for games running on a variety of platforms. FSx for OpenZFS delivers the fast, low-latency access needed to launch these experiences for Hatch gamers on demand. Using AWS DataSync, Hatch replicates their FSx for OpenZFS storage across different AWS Regions to scale its platform to gamers globally.

"Amazon FSx for OpenZFS provides the speed and scalability we need to power our cloud-native global hosting solution, while offering the simplicity and reliability of a fully managed service on AWS. Using this service, we reduced the operational burden of managing our own storage infrastructure, enabling us to focus on delivering more innovative gaming technology for our customers."

Karl Granström, Chief Operating Officer – Hatch

Vela Games reduces build times for Game Developers by 60% with Amazon FSx for OpenZFS.

Vela Games is a fully independent game studio on a mission to unite gamers around the world through play. They are making engaging, cooperative games that put players first.

Challenge: Vela Games uses BuildGraph, an Unreal Engine tool, to coordinate build jobs and accelerate the build times that were bottlenecking their game development process. To support this process, they needed a simple, high performance shared storage solution to share and persist data assets across a variety of Windows- and Linux-based build jobs.

Solution: With FSx for OpenZFS, Vela Games can deliver high performance, low latency, and fully managed NFS storage to accelerate its mission-critical build jobs. The cross-platform access (Windows/Linux) that Vela needed, while supporting advanced data management capabilities (like point-in-time snapshots and cloning) to enable Vela to further parallelize and accelerate its builds in the future.

"The speed and simplicity of Amazon FSx unlock our ability to use BuildGraph. Using this solution, we can parallelize and reuse assets quickly, which cuts down build time. By using FSx for OpenZFS and BuildGraph, Vela has reduced its total build time from an average of 56 minutes to 24 minutes."

Adam Comerford, Technical Director - Infrastructure & Operations, Vela Games