Mixpanel如何比数据仓库提供高达7倍的快速漏斗分析

Introduction

介绍

Now that Mixpanel can connect to tables from the data warehouse, many are curious how Mixpanel stacks up to SQL, the traditional interface for warehouse data in terms of query performance. Mixpanel not only eliminates the need to write SQL but also saves time with improved performance, thanks to our custom-built database optimized for product analytics.

连接到数据仓库的表，许多人都想知道Mixpanel在查询性能方面与传统的SQL界面相比如何。Mixpanel不仅消除了编写SQL的需要，还通过我们专为产品分析优化的自定义数据库提供了改进的性能，从而节省时间。

Today, we’ll benchmark Mixpanel’s performance against Snowflake, a popular data warehouse. Our focus will be on funnels, a critical type of product analytics query used to determine conversion rates and time-to-convert within defined product and business flows.

今天，我们将对比Mixpanel与流行的数据仓库Snowflake的性能。我们的重点将放在漏斗上，这是一种用于确定定义的产品和业务流程内的转化率和转化时间的关键类型的产品分析查询。

At Mixpanel, we built our own database called Arb. It’s an efficient column store, with a purpose-built query engine, based on the event data model. By modeling client, server, and warehouse data as events, we were able to optimize the performance of many of the common product analytics queries that data warehouses struggle with.

在Mixpanel，我们构建了自己的数据库Arb。它是一个高效的列存储，具有专门构建的查询引擎，基于事件数据模型。通过将客户端、服务器和仓库数据建模为事件，我们能够优化许多常见的产品分析查询，这些查询是数据仓库难以处理的。

To be clear before we get to the comparison— at Mixpanel, we love the data warehouse for many use cases. We store and process almost all of our business data in BigQuery, maintain a large DBT project, and thoroughly dogfood Mixpanel’s data warehouse integrations (Data Pipelines and Warehouse Connectors). However, warehouses struggle with many common product analytics queries; their relational model, while flexible, isn’t optimized for that use case.

在我们进行比较之前，明确一点——在Mixpanel，我们喜欢数据仓库的许多用例。我们将几乎所有的业务数据存储和处理在BigQuery中，维护一个庞大的DBT项目，并且彻底使用Mixpanel的数据仓库集成（数据管道和仓库连接器）。然而，数据仓库在许多常见的产品分析查询中存在问题；它们的关系模型虽然灵活，但并不针对该用例进行优化。

In our benchmarking, we found that Mixpanel is 3–7X faster than a 6XL Snowflake warehouse, the largest ...