Github user DaveBirdsall commented on a diff in the pull request:
https://github.com/apache/incubator-trafodion/pull/379#discussion_r56198891
--- Diff: docs/cqd_reference/src/asciidoc/_chapters/query_execution.adoc ---
@@ -0,0 +1,239 @@
+////
+/**
+* @@@ START COPYRIGHT @@@
+*
+* Licensed to the Apache Software Foundation (ASF) under one
+* or more contributor license agreements. See the NOTICE file
+* distributed with this work for additional information
+* regarding copyright ownership. The ASF licenses this file
+* to you under the Apache License, Version 2.0 (the
+* "License"); you may not use this file except in compliance
+* with the License. You may obtain a copy of the License at
+*
+* http://www.apache.org/licenses/LICENSE-2.0
+*
+* Unless required by applicable law or agreed to in writing,
+* software distributed under the License is distributed on an
+* "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+* KIND, either express or implied. See the License for the
+* specific language governing permissions and limitations
+* under the License.
+*
+* @@@ END COPYRIGHT @@@
+ */
+////
+
+[[query-execution]]
+= Query Execution
+
+This section describes CQDs that are used to influence query execution.
+
+[[hbase-async-operation]]
+== HBASE_ASYNC_OPERATION
+
+[cols="25%h,75%"]
+|===
+| *Description* | Allows index maintenance to be performed
concurrently with base table operation.
+| *Values* | *'ON'* Index maintenance is allowed. +
+*'OFF'* Index maintenance is not allowed. +
+ +
+The default value is *'ON'*.
+| *Usage* | HBase `put` operations are blocking. When
the table has one or more indexes,
+then the insert/update/delete (IUD) operation response time is improved by
executing the index maintenance
+operations concurrently with the base table operation: the put operations
to these HBase tables are executed
+in different threads.
+| *Production Usage* | Yes. It is 'ON' by default. This feature
can be disabled by setting this CQD to 'OFF'.
+| *Impact* | IUD operations on tables with one or more
indexes can become slower.
+| *Level* | Query.
+| *Conflicts/Synergies* | Not applicable.
+| *Real Problem Addressed* | Not applicable.
+| *Introduced In Release* | Trafodion 1.3.0.
+| *Deprecated In Release* | Not applicable.
+|===
+
+<<<
+[[hbase-cache-blocks]]
+== HBASE_CACHE_BLOCKS
+
+[cols="25%h,75%"]
+|===
+| *Description* | Influences HBase to retain the data blocks
in memory after they are read.
+| *Values* |
+*'ON'/'OFF'/'SYSTEM'* +
+ +
+The default value is *'SYSTEM'*.
+| *Usage* | HBase maintains the block cache structure
to retain the data blocks in memory after they are read.
+In LRU block cache configuration, the amount of block cache retained in
memory is proportional to the amount of reserved maximum
+Java heap size of the region server. LRU Block Cache is the default in
HBase. +
+ +
+The Trafodion Optimizer determines whether a sequential scan of the HBase
table in a query would cause the full eviction of the
+data blocks cached earlier thereby impacting the performance of the random
reads. The cache blocks option is turned off for the table
+in such a case. +
+ +
+Set the CQD
<<hbase-region-server-max-heap-size,HBASE_REGION_SERVER_MAX_HEAP_SIZE>> value
to reflect the amount of java heap size reserved for the region servers.
+This CQD is used by the Trafodion Optimizer to evaluate if the block cache
to be turned off.
+| *Production Usage* | Leave the setting to be 'SYSTEM' when
HBase is configured to use LRU block cache. If needed,
+you can override this settings with 'ON' or 'OFF'. +
+ +
+With other HBase configurations, you need to set HBASE_CACHE_BLOCKS to
'ON' or 'OFF' based on your application needs.
+| *Impact* | Automatically retains the random read
performance.
+| *Level* | Query.
+| *Conflicts/Synergies* | Not applicable.
+| *Real Problem Addressed* | Not applicable.
+| *Introduced In Release* | Trafodion 1.3.0.
+| *Deprecated In Release* | Not applicable.
+|===
+
+<<<
+[[hbase-filter-preds]]
+== HBASE_FILTER_PREDS
+
+[cols="25%h,75%"]
+|===
+| *Description* | Allows push down of predicates to HBase
Region Servers using HBase filters and optimize the columns retrieved
+from Region Servers. Only supported for NON ALIGN FORMAT tables.
+| *Values* |
+*'OFF'*: Predicates are never pushed down. +
+*'ON'*: A first implementation targeted for deprecation is enabled.
Support simple predicate formed by a combination of AND only.
+Could be counter-productive when applied on nullable columns. +
+*'1'*: Same as *'ON'*. +
+*'2'*: Full feature is enabled. +
+ +
+An explain plan can show whether predicates are successfully pushed down
to the Region Servers and what columns are really retrieved. +
+ +
+The default value is *'OFF'*.
+| *Usage* | Used to improve performance by reducing
the number of columns retrieved to a strict minimum
+and filter out rows as early as possible.
+| *Production Usage* | Please consult the Trafodion community.
+| *Impact* | Using this CQD increases the amount of
work done in the HBase Region Servers.
+| *Level* | System or Session.
+| *Conflicts/Synergies* | Not applicable.
+| *Real Problem Addressed* | Not applicable.
+| *Introduced In Release* | Trafodion 2.0.0.
+| *Deprecated In Release* | Not applicable.
+|===
+
+<<<
+[[hbase-hash2-partitioning]]
+== HBASE_HASH2_PARTITIONING
+
+[cols="25%h,75%"]
+|===
+| *Description* | Treat salted Trafodion tables as
hash-partitioned on the salt columns.
+| *Values* |
+*'OFF'*: Salted Trafodion tables are not hash-partitioned on the salt
columns. +
+*'ON'*: Salted Trafodion tables are hash-partitioned on the salt columns. +
+ +
+The default value is *'ON'*.
+| *Usage* | If, for any reason, there are issues with
parallel plans on salted tables (especially with data skew) then try setting
this CQD to OFF.
+| *Production Usage* | Yes.
+| *Impact* | Not applicable.
+| *Level* | System or Session.
+| *Conflicts/Synergies* | Not applicable.
+| *Real Problem Addressed* | Not applicable.
+| *Introduced In Release* | Trafodion 2.0.0.
+| *Deprecated In Release* | Not applicable.
+|===
+
+<<<
+[[hbase_num_cache_rows_max]]
+== HBASE_NUM_CACHE_ROWS_MAX
+
+[cols="25%h,75%"]
+|===
+| *Description* | Determines the number of rows obtained
from HBase in one RPC call to the HBase Region Server in a sequential scan
operation,
+| *Values* |
+Numeric value. +
+ +
+The default value is *'10000'*.
+| *Usage* | This CQD can be used to tune the query to
perform optimally by reducing the number of interactions to the HBase Region
Servers during
+a sequential scan of a table. +
+ +
+You need to consider how soon the maximum number of rows are materialized
on the Region Servers. When filtering is pushed down to Region Servers,
+then it can take a longer time depending upon the query and the predicates
involved. This can result in HBase scanner timeouts.
+| *Production Usage* | Use the default setting and reduce the
value to avoid HBase scanner timeouts. Consult with the Trafodion community
+if you think that you need to use this CQD.
+| *Impact* | Not applicable.
+| *Level* | Query.
+| *Conflicts/Synergies* | Not applicable.
+| *Real Problem Addressed* | Not applicable.
+| *Introduced In Release* | Trafodion 1.3.0.
+| *Deprecated In Release* | Not applicable.
+|===
+
+<<<
+[[hbase-rowset-vssb-opt]]
+== HBASE_ROWSET_VSSB_OPT
+
+[cols="25%h,75%"]
+|===
+| *Description* | Allows INSERT, UPDATE, and DELETE (IUD)
operations to be performed as an HBase batch `put` operation.
+| *Values* |
+*'ON'*: Perform IUD operations as an HBase batch `put` operation. +
+*'OFF'*: Do not perform IUD operations as an HBase batch `put` operation. +
+ +
+The default value is *'ON'*.
+| *Usage* | When IUD operation involves multiple
tuples, then the Trafodion Optimizer evaluates whether these operations
+can be done in a batch manner at the HBase level thereby reducing the
network interactions between the client applications and the HBase Region
Servers. +
+ +
+If possible, then the query plan involves VSBB operators. The Virtual
Sequential Block Buffer(VSBB) name is retained in Trafodion though it is
unrelated to HBase.
+| *Production Usage* | Yes.
+| *Impact* | IUD operations can become slower if this
CQD is set to 'OFF'.
+| *Level* | Query.
+| *Conflicts/Synergies* | Not applicable.
+| *Real Problem Addressed* | Not applicable.
+| *Introduced In Release* | Trafodion 1.3.0.
+| *Deprecated In Release* | Not applicable.
+|===
+
+<<<
+[[hbase-rowset-vssb-size]]
+== HBASE_ROWSET_VSSB_SIZE
--- End diff --
Typo. Should be HBASE_ROWSET_VSBB_SIZE
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
