Exadata Storage Index Aging – Part 2B

This is the third blog in a series looking at Storage Cell Storage Index Predicate Offloads. Continued from previous post. And there’s more here: Analysis for Test 2B click here.

For Test 2B – This corrects the columns with very little offload so that they get between 60-100 percent offload from the storage index. See the previous 2 blogs for more background information.

Now we have offload for each test segment and 8 columns offloaded for each In Line (IL) test as expected. We can even see a pattern on when columns are released (aged out) after repeated queries on the 8 columns.

The pattern I see is that after 10 queries columns 1, and 7-21 seemed to keep their offload status without aging out. But what is going on with column 1 that keeps it from aging out for the entire test. And what about 2-6 that seem to age out fairly fast after their query run with a storage index offload.

I see some patterns developing, but may need more eyes info or tests to figure this out. Will continue when I can.

Let me hear your ideas.

Here’s the test sequence again:

Test 2 - Test Plan

Test 2 – Test Plan

Here’s the full output in excel format; this time run to completion:

Exadata Storage Index Aging – Part 2A

Continuing from Storage Index Aging – part 1 the test was re-run. This time the In-Line (IL) test was run for the most recent test column and then proceeding backwards through the most recently run columns, then running the remaining columns in the 21 column set for the table.

Here’s a quick graphic that may be easier to visualize.

Test 2 - Test Plan

Test 2 – Test Plan

This set of tests are closer to what I expected, but those “pesky” columns 6, 7, 13, 14, 20 and 21 are still not offloading the way I’d like, so I’ll fix that in Test 2B (next blog).

On the bright side, we see more of the columns persisting in the IL tests. Looks like 6 out of our expected 8 columns are persisting without aging out. Also, although the columns above are “pesky”, they may also be telling us something about how the aging process works. But for now it’s still just an observation.

Also note the persistence of columns 18 and 19, still the same as in Test 1.

Output Data (Abbreviated):

Image with partial output data:

Partial Image of Data Results - Test 2a

Partial Image of Data Results – Test 2a

Excel file with more output data: dcox_exadata_storage_index_aging_test2