将大型 (6TB) Oracle 数据库的更新流复制到另一个非 DBMS 系统的最佳方法是什么？ 我不需要“批量加载”oracle 数据库，而只想将所有更新近乎实时地（10 秒或更短的延迟）传输到另一个本地系统中。 更新速度为 150 行/秒，相当于每秒 10 兆字节。
为了清楚起见，让我强调一下，我不是从一个数据库复制到另一个数据库。 这是一个应用程序集成问题：我需要从数据库复制到内部非数据库应用程序。 我曾考虑过使用企业服务总线，但这似乎不合适。
What is the best way to replicate the update stream to a large (6TB) oracle db into another non-DBMS system? I don't need to "bulk load" the oracle db, but merely want to flow all updates into another home-grown system in near realtime (10s latency or less). Updates happen at the rate of 150 rows/second representing 10s of megabytes per second.
For clarity, let me emphasize that I am not replicating from one db to another. This is an application integration problem: I need to replicate from a db into an in-house, non-db application. I've thought of using an enterprise service bus, but that seems inappropriate.
发布评论
评论（1）
那么您想要实现一个基于日志的复制方案，将事务日志泵送到不同的文件系统吗？ 有许多商业产品可以让您做到这一点，最著名的是 SharePlex< /a> 和 GoldenGate，此外甚至还有一个名为 Zizzy，但我不能说我已经尝试过。
编辑 1：如果您不想将更新提供到另一个数据库，GoldenGate 的事件驱动架构 提供了一种将复制连接到 JMS 消息总线的方法。 对于应用程序集成，这是我会选择的选项。
编辑 2：如果使用消息总线不是您的方式想去，我能看到的唯一其他明智的选择是亲自动手并使用 Oracle 的 LogMiner API。 IIRC，这就是 Oracle 用于 Streams 和 DataGuard 等功能的方式，并且它可能是大多数变更数据捕获解决方案的基础。
So you're wanting to implement a log based replication scheme that just pumps the transaction log off to a different file system? There are a number of commercial products that will let you do this, the better known ones being SharePlex and GoldenGate, plus there's even an open source one called Zizzy, but I can't say that I've tried it.
edit 1: if you're not wanting to feed the updates into another database, GoldenGate's Event-Driven Architecture provides a means to hook up the replication to a JMS message bus. For application integration, this is the option that I'd go for.
edit 2: if using a message bus isn't the way you want to go, the only other sensible alternative that I can see is to get your hands dirty and use Oracle's LogMiner API. IIRC, this is what Oracle uses for features like Streams and DataGuard and it's likely what most Change Data Capture solutions are based on, under the hood.