- If partition scheme allocates more data to one partition than the others, it leads to a “hot spot”
- The process by which hot spots emerge is called “partition skew”
- This greatly undermines the scalability benefits of partitioning
- Hot spots can emerge gradually due to changes in the distribution of the data
- Can be rectified through partition rebalancing
- This is usually quite situational, so the data systems don’t do it themselves
- In some systems, such as social networks,