There are several categories of scaling. A system may use one or more scaling techniques.
Decomposition
This is effectively microservices. This breaks down functional requirements into discrete services; each a domain of functional requirements.
Scaling System
Domain slices may or may not have the same needs as other domains in the system. Micro-services enables domains to not only scale independently from one another, but the taken approaches to scale may be different as each domain may tolerate or have different trade-offs as part of its requirements.
Scaling Organization
Also enables scaling the number of developers that can work on a system in an organization. Teams can own slices of the domain (services). This allows them to work within that domain without as many dependencies on other teams of developers.
Challenges
- System complexity increases. Although functionally the complexity remains the same, there is an increase in system complexity. This is due to the physical boundaries between domains; both in terms of infrastructure and domain communication.
- The impact of cross domain dependencies are exacerbated in comparison to a monolithic system. In a monolithic system, dependencies occur mostly at the code level. Microservices elevate these to a systems level; both in terms of implementation and organizationally. For example, a team depending on another team’s roadmap and work completion is an example of the increased impact these cross domain dependencies can cause.
Vertical Scaling
Scaling the computational power/resources of a single server. Has a rather finite limit to how much it can scale a system.
Horizontal Scaling
Services are scaled across multiple servers. This requires a coordinator (load balancer) to distribute requests across all servers. There are different approaches to doing this, such as weighted round robin, sticky sessions, etc.
Partitioning
This approach splits requests across a physical boundary (typically network or geographic location). This is a common approach to scale both databases (sharding) and systems across data centers.
