Sql-server – Hash join spills to disk even though there is plenty of memory granted for the query

The cost is the same (1%) for both the slow and fast cases. Does that mean the warning can be ignored? Is there a way to show "actual" times or costs. That would be so much better! Actual row counts are the same for the operation with the spill.

The cost shown is always the optimizer's estimated cost of the iterator, computed according to its internal model. This model does not reflect your server's particular performance characteristics; it is an abstraction that happens to produce reasonable plan shapes most of the time for most queries on most systems. There is no way to show 'actual' costs/execution times per iterator.

Besides performing a manual text diff of xml execution plans to find the differences in warnings, how can I tell what the 1500% increase in runtime is actually due to?

Typically, you can't. Spill warnings (sorts, hashes, exchanges) are new in execution plans for 2012, but they are just an indication of something you should investigate and look to eliminate if possible. The impact of a particular spill is something that needs to be measured - it is not possible to say that a spill of a particular type will always result in an x% performance drop for example.

For slow case, tempdb before/after (select * sys.fn_virtualfilestats(db_id('tempdb'),null)) (only showing a few 100ms of latency)

Spilling to tempdb and back is certainly undesirable, but the overall impact is hard to assess. For sort and hash spills, the impact is largely due to the I/O and access pattern, which may be small-block synchronous I/O e.g. for sort spills. With ~100ms of latency, you don't need too many synchronous I/Os to introduce a significant delay. The nature of the process and I/O patterns means tempdb spills can still be a problem on very low latency storage systems like fusion-io.

For exchange spills, there is an extra delay. The intra-query deadlock must be detected by the regular deadlock monitor, which by default only wakes up once every 5 seconds (more frequently if a deadlock has been found recently).

The resolver must then choose one or more victims, and spool exchange buffers to tempdb until the deadlock is resolved. The amount of spooling needed and the complexity of the deadlock will largely determine how long this takes.

Ultimately, preserved ordering is a Very Bad Thing for parallelism in general. Ideally, we want multiple concurrent threads operating on data streams with no inter-dependence. Preserving sort order introduces dependencies, so producer and consumer threads in different parallel branches can become deadlocked waiting for order-preserving iterators to receive rows to decide which input sorts next in sequence.

The precise nature of the deadlock depends on data distribution and per-thread sort order at runtime, so it is typically very hard to debug. Hence my recommendation to avoid order-preserving iterators in parallel plans, especially at high DOP. I do explain a very simplified example of an order-preserving parallel deadlock in some talks I do, but real examples are always more complex, though the underlying cause is the same.

In case the concepts are not familiar, it may help to follow the following example, reproduced from the (somewhat epic) 1993 paper Query Evaluation Techniques for Large Databases by Goetz Graefe:

If a different partitioning strategy than range-partitioning is used, sorting with subsequent partitioning is not guaranteed to be deadlock-free in all situations. Deadlock will occur if (i) multiple consumers feed multiple producers, and (ii) each producer produces a sorted stream and each consumer merges multiple sorted streams, and (iii) some key-based partitioning rule is used other than range partitioning, i.e., hash partitioning, and (iv) flow control is enabled, and (v) the data distribution is particularly unfortunate.

Figure 37 shows a scenario with two producer and two consumer processes, i.e., both the producer operators and the consumer operators are executed with a degree of parallelism of two. The circles in Figure 37 indicate processes, and the arrows indicate data paths. Presume that the left sort produces the stream 1, 3, 5, 7, ..., 999, 1002, 1004, 1006, 1008, ..., 2000 while the right sort produces 2, 4, 6, 8, ..., 1000, 1001, 1003, 1005, 1007, ..., 1999.

The merge operations in the consumer processes must receive the first item from each producer process before they can create their first output item and remove additional items from their input buffers. However, the producers will need to produce 500 items each (and insert them into one consumer’s input buffer, all 500 for one consumer) before they will send their first item to the other consumer. The data exchange buffer needs to hold 1000 items at one point of time, 500 on each side of Figure 37. If flow control is enabled and the exchange buffer (flow control slack) is less than 500 items, deadlock will occur.

The reason deadlock can occur in this situation is that the producer processes need to ship data in the order obtained from their input subplan (the sort in Figure 37) while the consumer processes need to receive data in sorted order as required by the merge. Thus, there are two sides which both require absolute control over the order in which data pass over the process boundary. If the two requirements are incompatible, an unbounded buffer is required to ensure freedom from deadlock.

Update: One of your comments raised a really good point that I sadly didn't even think of first. This is actually a pretty common place to look at - the higher your MAXDOP - the more threads you can use. The more you can use, the more memory that can be consumed because you have more going on at once inside of a query - more parallel thread. You said your MAXDOP is 10 server wide. This "feels" high. The latest guidance is 8 if more than 8 logical cpus, or something between "0 and N" if you have less than 8. I tend to look at the number of cores per NUMA node and may go lower than 8 for some systems that are running OLTP workloads based on the number of cores per NUMA node, and sometimes I go higher. But you might look at these guidelines and consider looking at your instance wide MAXDOP and consider changing it and testing to see the impact. Going lower than 10 will reduce your memory grants for these queries.

This doesn't eliminate the need to tun queries, however, and I think there is probably room for that. Basically you need to either continually increase memory until you can satisfy all your grants, consider lowering MAXDOP, tune queries so they consume smaller memory grants, consider reducing the workload - lower the amount of queries executed at once, or get into Resource Governor settings which can artificially lower the grant amount but most assuredly increase TempDB dependence which the memory grants (memory grant % especially).

Those are the main approaches I take. If you can, starting with query tuning and perhaps considering MAXDOP changes are a good place to start. I tend to leave Resource Governor as a final resort, have good luck with tuning queries and playing with server wide or query specific DOP.

Original Answer Query memory grants are a part of SQL Server helping to ensure the right amount of memory is available before running a query. It is a way to try and make sure that you aren't running an excessive amount of high memory consuming queries that will all but guarantee a lot of memory consuming operations spilling to TempDB.

In this case I imagine your Top 3 queries are consuming a lot of memory. And you were right to look at their plans. Hashes and Sorts are probably one of the issues here. Looking at tuning queries so they are appropriate in their plans and, in turn, memory consumption is a great solution here. I doubt this is a bug - it is the way SQL Server was designed.

This article is a bit old, but the knowledge contained within still applies. It is a good place to begin understanding query memory grants.

If tuning the queries doesn't save the day here, there are other tricks that can be employed. I've also asked some questions in the comments section of your question that may help expand an answer from me or someone else.