Query Performance Prediction for Concurrent Queries using Graph Embedding

Existing performance predictors cannot predict the performance of concurrent queries

• hard to capture the correlation between various queries. (lock conflict and buffer sharing)

This paper

• Proses a performance prediction system based on a graph embedding-based model.

• Propose a graph update and compaction algorithm to adapt to workload changes, since workloads may dynamically change.

For a set of queries, they get the physical plan for each query, and then add an edge between them if there are few conditions:

• parent-child relationship in the same query plan;
• (2) data-sharing: they access the same data, i.e., visiting the same table or index;
• (3) data conflict: they have read-write/write-write access conflicts;
• (4) resource-competition: they compete for the resource at the same time, e.g., competing memory, CPU, I/O bandwidth;

---

• ← Previous Post
• Next Post →