DMS latency

0

Hi, a customer is asking if DMS could be a viable option for real-time replica of an on-prem Oracle to RDS. The questions are:

  • what is the minimum achievable latency (besides network latency between on-prem and AWS)?

  • How does the DMS polling process work? Is there a defined sleep interval?

Thanks

AWS
Antonio
已提问 5 年前1553 查看次数
1 回答
0
已接受的回答

To answer your question simply, DMS is not a real-time replication engine. If you look at the settings for change management in DMS, without BatchApply turned on, DMS will collect transactions from the source database every second. It will attempt to collect the minimum number of transactions for a 1 second period before it applies those transactions to the target database.

https://docs.aws.amazon.com/dms/latest/userguide/CHAP_Tasks.CustomizingTasks.TaskSettings.ChangeProcessingTuning.html

It is also possible for DMS to gather changes from the source DB faster than it can write them to the target DB. It is also possible for these "buffered" transactions to fill the DMS instance memory therefore being written to disk. In these cases, latency will be impacted.

There are many options for tuning DMS based on the customer requirements, so if we look at the specific requirements for the customer's use case, then it might be viable option. However, simply asking for "real-time" replication is not enough to conclude that DMS is or isn't the right solution for them.

Minimum achievable latency depends on the use case and the DMS settings the customer is using for the use case.

Task Settings for DMS

I hope this helps!

已回答 5 年前

您未登录。 登录 发布回答。

一个好的回答可以清楚地解答问题和提供建设性反馈,并能促进提问者的职业发展。

回答问题的准则