TempDB: The Bottleneck You Can’t See in the Execution Plan

TempDB has always been a shared resource, but modern SQL Server workloads place more pressure on it than ever before — often without making the problem obvious.

Larger and adaptive memory grants, increased parallelism, version store usage, and feedback driven optimisation all mean TempDB activity is no longer limited to “bad queries” or obvious spills. In many cases, the execution plan looks reasonable, yet the instance still suffers from unexplained slowdowns, concurrency issues, or unpredictable performance.

In this session, we’ll look at how TempDB is used by the engine today, why traditional troubleshooting techniques often fall short, and how to identify systemic TempDB contention that affects every workload on the instance. You’ll learn how to correlate runtime behaviour, waits, memory grants, and TempDB usage to understand what’s really happening — even when the execution plan doesn’t tell the full story.

This isn't a TempDB fundamentals "you need more data files" session. It’s a practical, focused look at how TempDB becomes a hidden bottleneck in modern SQL Server environments, and most importantly what you can do about it.

After attending this session, you will be able to: - Explain why modern SQL Server features increase TempDB pressure even when queries appear well tuned - Identify instance wide TempDB contention using runtime signals, waits, and memory grant behaviour - Diagnose TempDB related performance problems that aren't visible in execution plans