[SERVER-9120] Load based balancing Created: 25/Mar/13  Updated: 16/Jan/24

Status: Blocked
Project: Core Server
Component/s: Performance, Sharding
Affects Version/s: None
Fix Version/s: None

Type: Improvement Priority: Major - P3
Reporter: Scott Hernandez (Inactive) Assignee: Backlog - Cluster Scalability
Resolution: Unresolved Votes: 3
Labels: balancer, balancingStrategy, load, oldshardingemea
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
Depends
Related
related to SERVER-5047 Better chunk selection when balancing Blocked
is related to SERVER-9114 Add balancer strategy that balances b... Blocked
Assigned Teams:
Cluster Scalability
Backwards Compatibility: Fully Compatible
Participants:

 Description   

The balancer could use a load assessment for deciding what chunk should be balanced where.

Any combination of the following:

  • Most efficient working set
  • Least faults
  • Least writes
  • Least network traffic
  • Fastest query response time
  • Lowest CPU usage
  • Least memory used for sorting

Policy 1:

  1. Find most active chunk by query usage (counts + size)
  2. Find shard with most efficient working set (least faults)
  3. Do move

Policy 2:

  1. Find largest active chunk (in last X minutes)
  2. Find shard with least number of chunks
  3. Do move

Generated at Thu Feb 08 03:19:26 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.