一篇文章读懂Spark AQE SkewedJoin该如何使用。

本文基于开源 spline 方案的调研，对如何丰富 Spark APP 的血缘解析， 提供了方案和深入的原理剖析。

要彻底搞懂一篇论文，最好的方式就是动手复现它，复现的过程你会遇到各种各样的疑惑、理论细节。

本文介绍了Spark离线开发框架的设计与实现，让开发变得简单、易上手，同时也解决了日常工作中数据回溯的痛点问题。

We can control the way Spark partitions our data and us it to parallelize computations on our dataset.

Shuttle：一个高可用 高性能的Spark Remote Shuffle Service。支持AQE功能，为Spark引擎提供更稳定，更高效的计算保障。

字段血缘可以很好的帮助我们了解数据生成的处理过程，在探索中我们发现了可以通过Spark的扩展来优雅的实现这一功能。

In a previous blog post, we discussed LyftLearn’s infrastructure built on top of Kubernetes. In this post, we will focus on the compute layer of LyftLearn, and will discuss how LyftLearn solves some of the major pain points faced by Lyft’s machine learning practitioners.

本文总结了工作中Spark在供应链核算中的应用。

Hudi、Iceberg等数据湖引擎目前使用的越来越广泛，很多B端客户在使用Spark SQL的时候也存在需要使用数据湖引擎的需求，因此字节EMR产品需要将数据湖引擎集成到Spark SQL中，在这个过程碰到非常多的问题。

本文讨论京东Spark计算引擎研发团队基于Bloom Filter算法的Runtime Filter Join优化机制，助力京东大促场景的探索和实践。

Spark, Hive and HDFS (Hadoop Distributed File Systems) ecosystems are online analytical processing (OLAP)-oriented technologies. They are designed to process huge amounts of data with full scans. From time to time, users want to use the same data for more ad-hoc oriented tasks:

• Multi-row load— explore small sets (typically 1%) of the data by specific IDs (not random).
• Single-row fetch — for example, building a serving layer to fetch a specific row upon a REST-API request.

These kinds of tasks are traditionally solved using dedicated storage and technology stacks (HBase, Cassandra, etc.) which require data duplication and add significant operational costs.

In this post, we describe our journey for solving this challenge by using only Spark and HDFS. We will start by introducing an example use case, generalize and define the requirements, suggest some optional solutions, and finally dive into our final solution.

To achieve our mission of bringing everyone inspiration through our visual discovery engine, Pinterest relies heavily on making data-driven decisions to improve the Pinner experience for over 475 million monthly active users. Reliable, fast, and scalable interactive querying is essential to make those data-driven decisions possible. In the past, we published how Presto at Pinterest serves this function. Here, we’ll share how we built a scalable, reliable, and efficient interactive querying platform that processes hundreds of petabytes of data daily with Apache Spark SQL. Through an elaborate discussion on various architecture choices, challenges along the way, and our solutions for those challenges, we share how we made interactive querying with Spark SQL a success.

Qunar 智能风控场景中，风控研发团队经常会应用一些算法模型，来解决复杂场景问题。典型的如神经网络模型，决策树模型等等。而要完成模型从训练到部署预测的全过程，除了模型算法之外，离不开技术框架的支撑。本篇文章将和大家分享一下，在预测服务部署阶段，基于 Tensorflow for Java 和 Spark-Scala 构建分布式机器学习计算框架的实践经验。

随着近几年业务快速发展与迭代，大数据的成本也水涨船高，如何优化成本，建设低成本高效率的底层服务成为了有赞数据基础平台2020年的主旋律。本文主要介绍了随着云原生时代的到来，经历7年发展的有赞离线计算平台如何拥抱云原生，通过容器化改造、弹性伸缩、大数据组件的错峰混部，做到业务成倍增长的情况下成本负增长。

在滴滴SQL任务从Hive迁移到Spark后，Spark SQL任务占比提升至85%，任务运行时间节省40%，运行任务需要的计算资源节省21%，内存资源节省49%。在迁移过程中我们沉淀出一套迁移流程， 并且发现并解决了两个引擎在语法，UDF，性能和功能方面的差异。