10.07.2015 Views

Expert Oracle Exadata - Parent Directory

Expert Oracle Exadata - Parent Directory

Expert Oracle Exadata - Parent Directory

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

CHAPTER 11 UNDERSTANDING EXADATA PERFORMANCE METRICSSQL> SELECT SUM(object_id) FROM t4_hcc_query_high WHERE owner LIKE 'S%';SUM(OBJECT_ID)--------------3.9741E+12SQL> SELECT * FROM v$sql WHERE sql_id = '0rv137mgwpnkx';(output pivoted and edited for better readability)IO_CELL_OFFLOAD_ELIGIBLE_BYTES: 6188924928IO_INTERCONNECT_BYTES : 720233312PHYSICAL_READ_REQUESTS : 6547PHYSICAL_READ_BYTES : 6188982272PHYSICAL_WRITE_REQUESTS : 0PHYSICAL_WRITE_BYTES : 0OPTIMIZED_PHY_READ_REQUESTS : 7LOCKED_TOTAL : 1PINNED_TOTAL : 2IO_CELL_UNCOMPRESSED_BYTES : 112741585171IO_CELL_OFFLOAD_RETURNED_BYTES: 720175968In this case the IO_CELL_OFFLOAD_RETURNED bytes is much smaller than the IO_CELL_OFFLOAD_ELIGIBLEbytes; thus, the Smart Scan definitely did help to reduce the data flow between the cells and thedatabase. Furthermore, the IO_CELL_UNCOMPRESSED_BYTES is way bigger than the PHYSICAL_READ_BYTES,which indicates that the EHCC helped to reduce the number of bytes that had to be read from disk bythe cells, thanks to compression. Note that the IO_INTERCONNECT_BYTES is not much greater than theIO_CELL_OFFLOAD_RETURNED_BYTES, which indicates that for this SQL, almost all the traffic was due to thedata returned by Smart Scans. There was no extra traffic due to other reasons like Temp table spacereads/writes thanks to non-optimal sorts, and there were no hash joins or other work-area operations ordatabase block I/Os, thanks to chained rows or in-database consistent read processing. Note: Smart Scanning makes data retrieval from segments faster, but it doesn’t do anything to speed upjoining, sorting, and aggregate operations, which are applied after the data has been retrieved from the segments.A notable exception is the Bloom filter pushdown to cells, which allows the cells to filter the data from the probetable using a hash bitmap built based on the driving row source’s data in the hash join.While this example used the V$SQL view, which shows SQL child cursor level statistics, you could alsouse V$SQL_PLAN_MONITOR (the columns PLAN_LINE_ID, PLAN_OPERATION, and so on) to measure thesemetrics for each execution plan line. This is useful because a single execution plan usually accesses andjoins multiple tables, and different tables may benefit from the Smart Scan offloading differently. We willlook at some scripts and tools that use this data in the next chapter.376

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!