Horizontal scaling

As explained in MongoDB 8.0 in Action, Third Edition

Horizontal scaling is the practice of increasing capacity by distributing workloads across additional machines or partitions.

Horizontal scaling increases database capacity by spreading data and workload across additional servers. It contrasts with vertical scaling, which adds CPU, RAM, or storage to one server. Scaling out can be more cost-effective as data grows, but a larger network of servers adds management complexity. section 9.3

MongoDB scales out through sharding: a sharded collection is partitioned across shards, each holding part of its data. MongoDB balances data and distributes load as machines are added. Its self-contained documents help because retrieving one document need not require joins across servers. section 1.2 section 9.3

Figure 1.1 MongoDB’s sharding architecture enables horizontal scaling by distributing data across multiple servers. (Image © MongoDB 2024 CC BY-NC-SA 3.0)

The cluster diagram shows the application reaching shards through mongos query routers, rather than choosing a server for each query itself. section 1.2

Keep these operational distinctions in mind:

• Replication copies data to secondary members for availability and read distribution; writes still pass through the primary. Sharding partitions data, allowing writes to different shards to reach their respective primaries. Replication within each shard helps keep its data accessible after a server failure. section 9.0
• Sharding is chosen per collection. Unsharded collections remain on a database’s primary shard; MongoDB 8.0+ provides moveCollection to move them to another shard. section 9.3
• Atlas offers sharding as an option, but its cluster autoscaling does not add shards or convert a replica set into a sharded cluster. Do not mistake automatic tier or storage changes for horizontal scaling. section 9.0 section 10.2.3

When one server becomes the limit

If growing data exceeds a server’s RAM or strains disk I/O, upgrading that server may help only until its available configurations run out. Sharding offers another path: add servers and distribute the collection’s data and workload, while planning for the added operational complexity. section 9.3

How MongoDB 8.0 in Action, Third Edition covers it

How deep it goes

Explains sharding at collection and cluster level, including trade-offs, but provides no shard-creation commands in the supplied passages.

Examples

Uses RAM and disk capacity limits, a sharded-cluster diagram, and Atlas autoscaling behavior as concrete reference points.

What it assumes

Defines scaling up and out from basic server resources, while assuming familiarity with databases and collections.

What you can do after

Evaluate/understand when sharding addresses capacity or write bottlenecks, and why replication or Atlas tier autoscaling is not a substitute.

How it connects

Connects horizontal scaling to vertical scaling, sharding, replication, and autoscaling while distinguishing their roles.

liveBookmanning.com

Read MongoDB 8.0 in Action, Third Edition on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Amazon Web Services in Action, Third Edition: An in-depth guide to AWSHorizontal scaling is increasing or decreasing system capacity by adding or removing equivalent computing instances.
• 

  liveBookmanning.com
  Elasticsearch in Action, Second EditionHorizontal scaling increases system capacity by adding more nodes or machines rather than increasing the resources of an existing node.
• 

  liveBookmanning.com
  Distributed Machine Learning PatternsA scaling technique that increases capacity by adding equivalent service instances or machines rather than enlarging one instance.
• 

  liveBookmanning.com
  Building Web APIs with ASP.NET CoreHorizontal scaling increases capacity by adding or removing service instances rather than enlarging one instance.
• 

  liveBookmanning.com
  ScyllaDB in ActionCluster scaling is the operation of increasing a cluster's capacity by adding nodes while retaining its existing nodes.
• 

  liveBookmanning.com
  Refactoring to RustHorizontal scaling is increasing service capacity by running additional instances of a system.
• 

  liveBookmanning.com
  Microsoft Azure in ActionA scaling technique that adds or removes database or virtual-machine instances to change aggregate computing capacity.
• 

  liveBookmanning.com
  Kubernetes in Action, Second EditionScaling out is increasing application capacity by running additional application instances.
• 

  liveBookmanning.com
  Bootstrapping Microservices, Second Edition: With Docker, Kubernetes, GitHub Actions, and TerraformHorizontal scaling is the technique of increasing capacity by running multiple instances of a component and distributing its workload among them.
• 

  liveBookmanning.com
  Fast Python: High performance techniques for large datasets
• 

  liveBookmanning.com
  Build AI-Enhanced Web AppsHorizontal scaling is the addition or removal of application instances to change processing capacity.
• 

  liveBookmanning.com
  Grokking ConcurrencyA scaling approach that increases performance by distributing load across existing and additional processing resources.

Compare how 13 books explain horizontal scaling →

Related concepts

• vertical scalingalternative to · Vertical scaling is increasing or decreasing a system’s capacity by changing the resources available to an existing machine, node, or instance.
• replicauses · A replica is a copy of data or state, or an equivalent running instance of a component, maintained alongside others in a replicated system.
• scalabilitykind of · Scalability is the ability of a system, process, or method to accommodate increasing workload, size, or demand while maintaining acceptable operation.
• load balancinguses · Load balancing is the distribution of requests or computational work across multiple resources to manage demand, reduce workload imbalance, and improve utilization or availability.
• autoscalingkind of · Autoscaling is the automatic adjustment of computing resource capacity or running instances in response to workload demand or measured system conditions.
• dataset splitimplemented by · A dataset split is the division of a dataset into distinct subsets assigned different purposes, such as training, validation, or testing.

This explainer was generated by Manning's knowledge graph from the text of MongoDB 8.0 in Action, Third Edition. The section links point to the parts of the book it draws on.