字号·· | 护眼
theregister

MongoDB收购Firebolt,将事务与分析纳入同一平台MongoDB snaps up Firebolt to bring transactions and analytics under one roof

点「原文对照」整页切到原文,或双击某段只看那段的原文。

MongoDB 收购了分析型数据库专家 Firebolt,这是其再次尝试将事务性和分析性工作负载整合到一个平台上。这家文档数据库供应商凭借灵活的、类 JSON 的文档,而非僵化的预定义表结构,在数据库市场开辟了自己的领地。开发者将其应用于 Web、移动、电商和内容管理应用。此后,MongoDB 拓展业务,提供一个据称已为 AI 时代和跨事务环境工作的 AI 智能体做好准备的数据平台,承诺提供规模、速度和可靠性。

MongoDB has acquired analytical database specialist Firebolt in another attempt to bring transactional and analytical workloads onto one platform. The document database vendor carved out its patch of the database market with flexible, JSON-like documents rather than rigid, predefined table structures. Developers put it to work in web, mobile, ecommerce, and content management applications. It has since branched out to offer a data platform it says is ready for the age of AI and AI agents working across transactional environments, promising scale, speed and reliability.

Firebolt 于 2019 年在以色列成立。它构建了一个云原生数据仓库,核心理念是在数据摄入时自动且透明地合并和压缩文件。时任 CTO 的 Eldad Farkash 2020 年在接受《The Register》采访时声称,这大幅减少了用户需要加载到系统中的数据量,从而加速查询性能并提升并发性。Firebolt 于 2022 年完成 C 轮融资,筹集 1 亿美元,据报估值达 14 亿美元。MongoDB 未披露收购价格。但这将为 MongoDB 带来什么,对 MongoDB 用户意味着什么?

Firebolt was founded in Israel in 2019. It built a cloud-native data warehouse based on the notion of automatically and transparently merging and compacting files as they are ingested. Speaking to The in 2020, then CTO Eldad Farkash claimed this slashes the amount of data users need to load into the system, accelerating query performance and improving concurrency. Firebolt raised $100 million in Series C funding in 2022 at a reported valuation of $1.4 billion. MongoDB has not disclosed the acquisition price. But what will it get, and what does that mean for MongoDB users?

公司在一篇博客文章中表示:“通过将 Firebolt 集成到 MongoDB 的智能数据平台中,我们将高性能分析引擎带到了您的运营数据所在之处。Firebolt 的引擎将与您的数据库并行运行,我们将自动保持其同步,因此无需构建或管理数据管道。您将获得一个同时支持运营和实时分析工作负载的平台,使团队能更快交付,且需运维和保障的系统更少。”

In a blog post, the company said: "By integrating Firebolt into MongoDB's Intelligent Data Platform, we're bringing a high-performance analytics engine to where your operational data already lives. Firebolt's engine will run right alongside your database, and we'll keep it in sync automatically, so there's no pipeline to build or manage. You'll get one platform for operational and real-time analytical workloads, so teams can ship faster, with fewer systems to operate and secure."

该集成功能尚不可用。MongoDB 表示,客户将在未来几个月内获得访问权限,而现有 Firebolt 客户将保留其当前的服务和支持。MongoDB 的主张是,AI 智能体将需要新鲜的运营数据和快速的分析答案,而无需客户拼凑独立的平台。

The integration isn't available yet. MongoDB says customers will gain access in the coming months, while existing Firebolt customers will retain their current service and support. MongoDB's pitch is that AI agents will need fresh operational data and fast analytical answers, without customers having to stitch separate platforms together.

MongoDB远非首家追逐这一组合的厂商。将在线事务处理(OLTP)与在线分析处理(OLAP)结合起来,有望减少需要管理的基础设施、减少需要维护的数据管道。难的是在不做妥协的情况下同时实现两者。今年早些时候,Databricks声称已攻克这一难题,承诺“一份数据、零妥协、零副本”。它借助新的计算引擎Reyden以及运行在开放对象存储上的无服务器PostgreSQL产品Lakebase来实现。

MongoDB is hardly the first vendor to chase that combination. Bringing online transaction processing (OLTP) and online analytical processing (OLAP) together promises less infrastructure to manage and fewer data pipelines to maintain. Delivering both without compromises is the harder part. Earlier this year, Databricks claimed to have cracked the problem, promising "one data, zero compromises, zero copies." It works with Reyden – a new compute engine – and Lakebase, its serverless PostgreSQL on open object storage.

竞争对手指出,该系统在某一层面上曾短暂存在两份数据副本。SingleStore也通过内存行存储和带分层存储的磁盘列存储来应对这一问题,“这意味着事务先命中内存,随后再落到磁盘存储上”,从而在单一系统上同时支持分析与事务。Oracle和SAP多年来也一直在为事务型平台添加分析能力,以缩短数据记录与分析之间的延迟。

Rivals pointed out there were briefly two copies of the data in the system at some level. SingleStore also tackles the problem with an in-memory row store and an on-disk column store with tiered storage, "meaning transactions hit memory first, and then they roll off to disk storage," allowing analytics and transactions on a single system. Oracle and SAP have also spent years adding analytical capabilities to transactional platforms to reduce the delay between recording data and analyzing it.

MongoDB承诺无需客户自行管理管道即可实现实时分析。这是否也意味着避免数据的单独副本,则不那么明确。自动同步并不能说明信息在底层是如何存储或移动的。已要求MongoDB澄清其架构。MongoDB此前已有过类似尝试。2022年,它推出了列式存储索引,旨在帮助开发者将分析查询构建到自己的应用中。其思路是在不改变文档结构、也不将数据迁移到另一系统的情况下加速这些查询。

MongoDB promises real-time analytics without a customer-managed pipeline. Whether that also means avoiding separate copies of the data is less clear. Automatic synchronization doesn't explain how the information is stored or moved underneath. The has asked MongoDB to clarify the architecture. MongoDB has been here before. In 2022, it unveiled columnstore indexing designed to help developers build analytical queries into their applications. The idea was to accelerate those queries without changing the document structure or moving data to another system.

Firebolt为MongoDB提供了又一件用于熟悉工作的工具:让分析功能在其运营数据库旁也能自然运作。这套底层机制能否兑现宣传,仍有待观察。®

Firebolt gives MongoDB another tool for a familiar job: making analytics feel at home beside its operational database. Whether the plumbing lives up to the pitch remains to be seen. ®

阅读媒体原文 ↗