[IMPALA-9382] Prototype denser runtime profile implementation

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: Impala 4.0.0
• Component/s: Backend
• Epic Color: ghx-label-2
• Assignee: Tim Armstrong
• Reporter: Tim Armstrong
• Created: 13/Feb/20 19:29
• Updated: 12/Feb/21 01:46
• Resolved: 12/Feb/21 01:46

Description

RuntimeProfile trees can potentially stress the memory allocator and use up a lot more memory and cache than is really necessary:

• std::map is used throughout, and allocates a node per map entry. We do depend on the counters being displayed in-order, but we would probably be better of storing the counters in a vector and lazily sorting when needed (since the set of counters is generally static after Prepare()).
• We store the same counter names redundantly all over the place. We'd probably be best off using a pool of constant counter names (we could just require registering them upfront).

There may be a small gain from switching thrift to using unordered_map, e.g. for the info strings that appear with some frequency in profiles.

However, I think we need to restructure the thrift representation and in-memory representation to get significant gains.

Issue Links

• breaks: IMPALA-10276 Release build sees SIGSEGV when updating the total time counter
• duplicates: IMPALA-1638 Investigate using c++ templates option when generating thrift and increasing the transport buffer
• is related to: IMPALA-6870 SummaryStatsCounter should be included in averaged profile