MongoDB Performance
ch5 Performance on Clusters

tags: MongoDB University M201

Performance Considerations in Distributed Systems

• Distributed System in MongoDB

  • Replica Cluster
  • Shard Cluster

• Working with Distributed Systems

  • Consider latency
  • Data si spread across different nodes
  • Read implications
  • Write implications

Before Sharding

• sharding is an horizontal scaling solution
• Have you reached the limits of our vertical scaling
• Need to understand how data grows and how data is access
• Sharding works by defining key base ranges
• It's important to have a shardkey

Increasing Write Performance with Sharding

• frequency
• Rate of change - Avoid monotonically increasing or decreasing values

db.collection.bulkWrite([op1, op2], { order: <bool> })

order: Execute operations one by one if "true" else execute operations in parallel

Reading from Secondaries

When reading from secondary is a good idea

• Offloading work
• Local reads

When reading from secondary is a bad idea

• In general
• Reading from shard directly

Aggregation Pipeline on a Sharded Cluster