[
https://issues.apache.org/jira/browse/ASTERIXDB-1556?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15414598#comment-15414598
]
Taewoo Kim commented on ASTERIXDB-1556:
---------------------------------------
The summary of a Skype call with [~dtabass], [~tillw], [~buyingyi], and
[~javierjia]:
#1. The following proposal written after another meeting is OK.
1) After each insertion, Data Table reports the number of frames used by DT (D)
and Hash table reports the number of frames used by HT (H).
2) If this insertion is failed. We spill a partition of Data Table to disk to
find space on Data table.
3) If this insertion is successful, we calculate D + H. If D + H >= F, then we
spill a partition of Data table to disk.
#2. The following is not OK. If hash table size becomes huge, then we need to
resolve the problem within the hash table itself. For instance, we can keep
track of how much spaces are wasted due to a hash slot migration. And we can
come up with a garbage collection method.
-4) If H >= 0.8 * F, we spill entire partitions of DT to disk and reset the
Hash table. 80% can be changed. This process is required not to let Hash table
occupy entire space.-
#3. Hash Table Size (= the number of possible hash values in Hash Table) should
come from the system-admin configuration file. And the compiler should observe
this setting. If this value is not provided in the file, it's OK to use the
default value specified in the codebase. Also, the sanity check between
hash_table_size and groupmemory should be done. i.e. the space occupation of
hash entries in Hash Table cannot exceed the specified budget (groupmemory).
#4. Query Sanity Check should be done. After a physical plan is computed, we
can calculate the maximum memory usage per operators in each stage. So, the
impossible execution of a query based on the global memory setting can be
checked and safely ignored rather than executing an Out Of Memory exception
error.
> Hash Table used by External hash group-by doesn't conform to the budget.
> ------------------------------------------------------------------------
>
> Key: ASTERIXDB-1556
> URL: https://issues.apache.org/jira/browse/ASTERIXDB-1556
> Project: Apache AsterixDB
> Issue Type: Bug
> Reporter: Taewoo Kim
> Assignee: Taewoo Kim
> Attachments: 2wayjoin.pdf, 2wayjoin.rtf, 2wayjoinplan.rtf,
> 3wayjoin.pdf, 3wayjoin.rtf, 3wayjoinplan.rtf
>
>
> When we enable prefix-based fuzzy-join and apply the multi-way fuzzy-join ( >
> 2), the system generates an out-of-memory exception.
> Since a fuzzy-join is created using 30-40 lines of AQL codes and this AQL is
> translated into massive number of operators (more than 200 operators in the
> plan for a 3-way fuzzy join), it could generate out-of-memory exception.
> /// Update: as the discussion goes, we found that hash table in the external
> hash group by doesn't conform to the frame limit. So, an out of memory
> exception happens during the execution of an external hash group by operator.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
