Learning & Operating Presto | Caching in Presto (Sept 28, 2023)
Dive into the new O’Reilly ebook, Learning & Operating Presto. Hear from Alluxio on different ways to get caching in Presto.
Presto at Adobe: How Adobe Advertising uses Presto for Adhoc Query, Custom Reporting, and Internal Pipelines
Rajmani Arya, Varun Senthilnathan & Manoj Kumar Dhakad, Adobe Advertising: We are from the Product Engineering team in Adobe Advertising (https://business.adobe.com/in/product…. Adobe Advertising is a digital advertisement platform. We take care of accumulating all data, providing platform intelligence, building and maintaining machine leaning capabilities, building and maintaining internal pipelines that form derived data to be used by other teams. The volume of total incoming raw data ranges between 8 to 10 tb/ day spread across 7 regions. The total data in the system currently is about 7pb. This data is largely stored in Hive tables with a central metastore. We use Presto in three ways: 1. Data studio – an internal tool to enable data analysts, sales, marketing and other teams to do adhoc querying. This is also used by data engineers to do adhoc querying for engineering tasks. 2. Custom Reports – We create reports for customers to get performance insights on their campaigns. We have 100s of reports that are run on a daily basis. 3. Internal Pipelines – Presto is used to retrieve data to power 100s of pipelines run daily to generate derived data.
Speeding Up Presto in ByteDance – Shengxuan Liu, Bytedance & Beinan Wang, Alluxio
Shengxuan Liu from ByteDance and Beinan Wang from Alluxio will present the practical problems and interesting findings during the launch of Presto Router and Alluxio Local Cache. Their talk covers how ByteDance’s Presto team implements the cache invalidation and dashboard for Alluxio’s Local Cache. Shengxuan will also share his experience using a customized cache strategy to improve the cache efficiency and system reliability.
Customer-Facing Presto at Rippling – Andy Li, Rippling
Presto is used for a variety of cases, but tends to be used for larger scale analytical queries. We have been transitioning to using Presto to power our data platform and customer-facing scripting language, RQL (Rippling Query Language) to run arbitrary customer queries to power core products. Presto helps enable diverse, federated querying at scale. In this talk, Andy will cover where Presto sits in Rippling’s ecosystem as a core query layer, our collaboration and contributions for closer integration with Apache Pinot, and learnings on using Presto to handle a large variety of query patterns.
Scaling Cache for Presto Iceberg Connector – Beinan Wang, Alluxio & Chunxu Tang
While using the Presto Iceberg connector, the in-heap cache in Presto is likely overloaded. In this talk, Beinan and Chunxu will share the design, implementation, and optimization of the off-heap cache to address the scalability challenges. You will learn how to cache Iceberg data and metadata for the Presto Iceberg connector, followed by future work on improving table scans using Apache Arrow.
Predicting Resource Usages of Future Queries Based on 10M Presto Queries at Twitter
Here, Chunxu and Beinan would like to share what they have learned in developing a highly-scalable query predictor service through applying machine learning algorithms to ~10 million historical Presto queries to classify queries based on their CPU times and peak memory bytes. At Twitter, this service is helping to improve the performance of Presto clusters and provide expected execution statistics on Business Intelligence dashboards.
Architecting Your Data Platform with Presto and Alluxio in Heterogeneous Environment – Adit Madan
As the cloud is evolving and the adoption of a hybrid-cloud or multi-cloud approach grows, the data architecture must adapt to heterogeneous environments. In this talk, Adit Madan shares insights on how to architect a data platform with Presto and Alluxio that provides agility and simplicity to your data team.
Speed Up Presto at Uber with Alluxio Caching – Chen Liang, Uber & Beinan Wang, Alluxio
At Uber, Presto is heavily used as one of the primary data analytics tools, and Presto’s query performance has profound production impact at Uber. As part of the Presto optimization effort, we turned to explore Alluxio as a caching solution. Alluxio is an open source data orchestration platform often used by many compute frameworks as the caching layer. Alluxio caching is currently enabled on ~2000 nodes across 6 clusters at Uber. In this presentation, we will talk about our journey at Uber of integrating Alluxio cache into Presto. We will discuss the Uber specific challenges we encountered and how we addressed them. We will also present the performance improvements we have seen. Besides, we will also discuss our plan and next steps, and potential future collaboration opportunities with the community.
Presto Query Analysis for Data Layout Formatting and Query Result Caching – Gurmeet Singh, Uber
In this talk, I will be talking about a microservice that we have built at Uber to be able to analyze Presto queries. The Presto Query Engine does not provide endpoints for query analysis purposes. One has to either execute the query or gather insights from the query explain plan. In this talk, I will talk about 1. The work that we had to do to do the query analysis in a microservice using Presto as a library. 2. Doing predicate analysis on the queries to come up with data formatting recommendations in order to improve query performance. 3. Using the analysis service for query result cache invalidation. The analysis figures out whether the results from a previous run of the query are still valid and can be reused.
Self-serve Data Architecture Evolution with Presto and Alluxio Across Clouds – Adit Madan, Alluxio
In this presentation, Adit Madan shares insights to help architect a data platform ready to minimize the impact of change and evolution. He will co-relate industry trends for a multi-tenant environment with how the Presto & Alluxio stack drives agility for hundreds of users in the cloud, across multiple datacenters and a hybrid cloud.
After RaptorX: Improve Performance Understanding and Workload Analysis in Presto – Ke Wang & Bin Fan
RaptorX, an umbrella project presented in PrestoCon Day in March, enabled the Presto interactive fleet in Facebook to reduce latency by 10x, based on a set of architectural improvements and optimizations with hierarchical caching. This presentation provides an update on the follow-up enhancement. Bin Fan from Alluxio will talk about the exploration of a probabilistic algorithm in Alluxio caching to estimate cache working set and the implementation of shadow cache Ke Wang from Facebook will talk about how shadow cache is used to understand the system bottleneck for better resource allocation and query routing decisions. She will also cover a recent improvement in collecting and aggregating per-query runtime statistics on the Presto engine to better understand the time breakdown, resource usage breakdown and cache hit rate on a per-query basis, which can help identify areas of improvement.
RaptorX: Building a 10X Faster Presto – James Sun, Facebook, Inc
RaptorX is an internal project name aiming to boost query latency significantly beyond what vanilla Presto is capable of. For this session, we introduce the hierarchical cache work including Alluxio data cache, fragment result cache, etc. Cache is the key building block for RaptorX. With the support of the cache, we are able to boost query performance by 10X. This new architecture can beat performance oriented connectors like Raptor with the added benefit of continuing to work with disaggregated storage.
How Carbon uses PrestoDB in the Cloud with Ahana to Power its Real-time Customer Dashboards
Carbon is a real-time revenue management platform that consolidates revenue and audience analytics, data management, and yield operations into a single solution. Real-time analytics is super critical – their customers rely on real-time data to make revenue decisions. After facing issues around performance, visibility & ease of use, and serverless pricing model with AWS Athena, the team moved to a managed service for PrestoDB in the cloud – Ahana Cloud – to power their customer-facing dashboards. In this session, Jordan will discuss some of the reasons the team moved from AWS Athena to a managed PrestoDB on Intel-optimized AWS instances. He will also dive into their current architecture that includes an Ahana-managed Hive Metastore along with Apache ORC file format and an S3-based data lake. Last, he’ll share some performance benchmarks and talk about what’s next for PrestoDB at Carbon.