Real-time OLAP databases are optimized for speed and often rely on tightly coupled storage-compute architectures using disks or SSDs. Decoupled architectures, which use cloud object storage, introduce an unavoidable tradeoff: cost efficiency at the expense of performance. This makes them unsuitable for databases that need to provide low-latency, real-time analytics, especially the new wave of LLM-powered dashboards, retrieval-augmented generation (RAG), and vector-embedding searches that thrive only when fresh data is milliseconds away. Can we achieve both cost efficiency and performance?
In this talk, we’ll explore the engineering challenges of extending Apache Pinot—a real-time OLAP system—onto cloud object storage while still maintaining sub-second P99 latencies.
We’ll dive into how we built an abstraction in Apache Pinot to make it agnostic to the location of data. We’ll explain how we can query data directly from the cloud (without needing to download the entire dataset, as with lazy-loading) while achieving sub-second latencies. We’ll cover the data fetch and optimization strategies we implemented, such as pipelining fetch and compute, prefetching, selective block fetches, index pinning, and more. We'll also share our latest work about integration with open table formats like iceberg, and how we will continue to achieve fast analytics directly on parquet files by implementing all the same techniques that apply to tiered storage.
Real-time OLAP databases are optimized for speed and often rely on tightly coupled storage-compute architectures using disks or SSDs. Decoupled architectures, which use cloud object storage, introduce an unavoidable tradeoff: cost efficiency at the expense of performance. This makes them unsuitable for databases that need to provide low-latency, real-time analytics, especially the new wave of LLM-powered dashboards, retrieval-augmented generation (RAG), and vector-embedding searches that thrive only when fresh data is milliseconds away. Can we achieve both cost efficiency and performance?
In this talk, we’ll explore the engineering challenges of extending Apache Pinot—a real-time OLAP system—onto cloud object storage while still maintaining sub-second P99 latencies.
We’ll dive into how we built an abstraction in Apache Pinot to make it agnostic to the location of data. We’ll explain how we can query data directly from the cloud (without needing to download the entire dataset, as with lazy-loading) while achieving sub-second latencies. We’ll cover the data fetch and optimization strategies we implemented, such as pipelining fetch and compute, prefetching, selective block fetches, index pinning, and more. We'll also share our latest work about integration with open table formats like iceberg, and how we will continue to achieve fast analytics directly on parquet files by implementing all the same techniques that apply to tiered storage.
Videos:
Presentation Slides:
Real-time OLAP databases are optimized for speed and often rely on tightly coupled storage-compute architectures using disks or SSDs. Decoupled architectures, which use cloud object storage, introduce an unavoidable tradeoff: cost efficiency at the expense of performance. This makes them unsuitable for databases that need to provide low-latency, real-time analytics, especially the new wave of LLM-powered dashboards, retrieval-augmented generation (RAG), and vector-embedding searches that thrive only when fresh data is milliseconds away. Can we achieve both cost efficiency and performance?
In this talk, we’ll explore the engineering challenges of extending Apache Pinot—a real-time OLAP system—onto cloud object storage while still maintaining sub-second P99 latencies.
We’ll dive into how we built an abstraction in Apache Pinot to make it agnostic to the location of data. We’ll explain how we can query data directly from the cloud (without needing to download the entire dataset, as with lazy-loading) while achieving sub-second latencies. We’ll cover the data fetch and optimization strategies we implemented, such as pipelining fetch and compute, prefetching, selective block fetches, index pinning, and more. We'll also share our latest work about integration with open table formats like iceberg, and how we will continue to achieve fast analytics directly on parquet files by implementing all the same techniques that apply to tiered storage.
Complete the form below to access the full overview:
.png)
Videos
Amazon S3 and other cloud object stores have become the de facto storage system for organizations large and small. And it’s no wonder why. Cloud object stores deliver unprecedented flexibility with unlimited capacity that scales on demand and ensures data durability out-of-the-box at unbeatable prices.
Yet as workloads shift toward real-time AI, inference, feature stores, and agentic memory systems, S3’s latency and limited semantics begin to show their limits. In this webinar, you’ll learn how to augment — rather than replace — S3 with a tiered architecture that restores sub-millisecond performance, richer semantics, and high throughput — all while preserving S3’s advantages of low-cost capacity, durability, and operational simplicity.
We’ll walk through:
- The key challenges posed by latency-sensitive, semantically rich workloads (e.g. feature stores, RAG pipelines, write-ahead logs)
- Why “just upgrading storage” isn’t sufficient — the bottlenecks in metadata, object access latency, and write semantics
- How Alluxio transparently layers on top of S3 to provide ultra-low latency caching, append semantics, and zero data migration with both FSx-style POSIX access and S3 API access
- Real-world results: achieving sub-ms TTFB, 90%+ GPU utilization in ML training, 80X faster feature store query response times, and dramatic cost savings from reduced S3 operations
- Trade-offs, deployment patterns, and best practices for integrating this tiered approach in your AI/analytics stack
Pratik Mishra delivered insights on architecting scalable, deployable, and resilient AI infrastructure at scale. His discussion on fault tolerance, checkpoint optimization, and the democratization of AI compute through AMD's open ecosystem resonated strongly with the challenges teams face in production ML deployments.
In this talk, Bin Fan, VP of Technology at Alluxio, presents on building tiered architectures that bring sub-millisecond latency to S3-based workloads. The comparison showing Alluxio's 45x performance improvement over S3 Standard and 5x over S3 Express One Zone demonstrated the critical role the performance & caching layer plays in modern AI infrastructure.