# Distributed Graphs Things to keep in mind: * Need to think about if the partition is balanced (randomness tends to help with this) * For certain scenarios this actually is a bad idea * Consider a social network. If you randomly partition then their is a high chance that a user will not be partitioned with their friends. Users rarely query the whole graph, but they frequently query their local friends * Likewise, if you consider a network of flights, likely there will be more flights that are close by (on the surface of the earth) that also fly to each other. Of course there will be long distance flights, but the number of long distance flights are likely lower than short small flights to nearby areas. So it could make sense to partition based on where the airports are in physical space (a cluster of sorts) * --- Date: 20211019 Links to: Tags: References: * [Using Graph Partitioning in Distributed Systems Design](https://youtu.be/QHkhyY9atkE)