When breaking monoliths into smaller services, the hardest part is actually breaking up the data that lives in the database of the monolith. To extract a data-rich service, it is useful to follow a series of steps which retain a single write-copy of the data at all times. The steps begin by making a logical separation in the existing monolith: splitting service behavior into a separate module, then separating data into a separate table. These elements can be separately moved into a new autonomous service.
There is a major shift in the industry away from monoliths towards smaller services. A key reason why organizations are investing in this shift is because smaller services built around business capabilities increase developer productivity. Teams that can own these smaller service/s can be “masters of their own destiny” which means they can evolve their service/s independently of other services in the system.
When breaking monoliths into smaller services, the hardest part is actually breaking up the data that lives in the database of the monolith. It is relatively easy to chop up the logic in the monolith into smaller pieces while still connecting to the same database. In this case, the database is essentially an IntegrationDatabase which gives the semblance of a distributed system that can evolve independently but in fact is a single tightly coupled system at the database level. For services to be truly independent and thus teams to be “master of their own destiny”, they also need to have an independent database – the schema and the corresponding data for the service.
In this article, I will be talking about a pattern, which is a series of steps, for extracting a data-rich service from a monolith while causing minimum disruption to the service consumers.
…take a look at the tag:
I would love to thank Martin Fowler for hosting this article and being generous with his time in reviewing this article. His review comments have truly taken this article to the next level. I would also like to thank Jorge Lee for providing critical commentary. I would like to thank my ThoughtWorks colleagues Joey Guerra, Matt Newman, Vanessa Towers, Ran Xiao and Kuldeep Singh for their comments on our internal mailing list.
30 August 2018: Published the rest of the article
29 August 2018: Published sixth and seventh step
28 August 2018: Published fifth step
27 August 2018: Published fourth step
25 August 2018: Published third step
24 August 2018: Published second step
23 August 2018: Published first step