Hi, OOW 16 is in the history books. Didn’t seem there were near the volume of people this year. The sessions I sat in on were full but not packed. I have a lot of material to share! The ACE Briefing (always the Thursday and Friday before) is at the Oracle campus. I am under NDA for anything not yet announced by Oracle Corp. So, I check with the people on what I can share and not share. It was all about the cloud. Oracle is really pushing their cloud services. Oracle has released Oracle12.2 on their Cloud Exedata offering, Oracle Exadata Express Cloud Service. I believe this is an Oracle-managed instance but does contain Oracle12.2 RDBMS. Otherwise, one presentation listed Oracle12.2 (on-premise versions) as being released later this calendar year. I thought that Oracle12.2 would be released at this event. Interesting word, on-premise. So, watch for blogs on: New Features for Developers New Features for DBA’s New Features for Rest Services (the APEX listener has become a product) New Features for In-Memory Updates to the cloud offerings New terminology as I run across it. Last April, I posted a nice series on In-memory, material I had gathered from Oracle PM Andy Revenes (heck of a nice guy). Here are the Oracle12.2. new features for In-Memory from VP Tirthankar Lahiri. His slide simply said “Oracle is presenting features for Oracle database 12c Release 2 on Oracle cloud. Features and enhancements related to the on-premises versions of Oracle Database 12c Release 2 are not being announced at this time”. Overall improved performance. He discussed: 3x better performance on mixed workloads (row processing and column store processing) 10x faster in-memory table joins New Feature: Join Groups boasts 2-3x over already fast in-memory joins Join Groups can be monitored via SQL Developer’s explain plans New Statistic: Columnar Encodings Leveraged (true when set to 1) Can store SQL expressions as additional in-memory columns, 3-5x faster on complex queries JSON columns can participate in column store giving a 20-60x performance increase in access times In-memory on Active Data Guard Use inmemory on standby database Exadata has in-memory columnar flash cache boasting a 10x performance increase In-memory Fast Start for 2-5x faster restore of column store In-memory Join Group Syntax: CREATE INMEMORY JOIN GROUP v_deptno (DEPT(deptno), EMP (deptno)); Create the join group on common columns being joined between two tables such as the illustrated EMP and DEPT. In-memory Expressions: CREATE TABLE CUSTOMER_SALES ( CUST_NO number, PRICE number, QTY number, TAX number, TOTAL_SALE as (PRICE * QTY + TAX)) INMEMORY; The column expressions must have a 1 to 1 relationship to the rows in the table or column store. You can also use DECODE, UPPER, LOWER, etc. Expressions can be manually defined (as shown in above syntax) or automatically created by Expressions Statistics Store (ESS) which monitors workloads. When it notices repeating SQL expressions, it will use the DBMS_INMEMORY.IME_CAPTURE to capture the repeating expressions and DBMS_INMEMORY.IME_POPULATE to create the inmemory virtual columns. Exadata flash cache is managed by the keyword CELMEMORY. ALTER TABLE emp CELLMEMORY; ALTER TABLE emp NO SELL MEMORY; CREATE TABLE CUSTOMER_SALES ( CUST_NO number, PRICE number, QTY number, TAX number, TOTAL_SALE as (PRICE * QTY + TAX)) CELLMEMORY MEMCOMPRESS FOR QUERY; You can use this feature on tables, partitions, sub partitions, and materialized views. The MEMCOMPRESS clause supports FOR QUERY LOW and FOR CAPACITY LOW. The NO PRIORITY clause will populate this into inmemory upon first request. This feature also supports automatic data optimizations with this policy syntax: INMEMORY or NO INMEMORY Alter MEMCOMPRESS level After of no access After of creation After of no modification On Can run these policies manually (as in a batch job perhaps) DBMS_ILM.EXECUTE_ILM Some examples include: ALTER TABLE emp ILM ADD POLICY NO INMEMORY AFTER 6 months OF CREATION; ALTER TABLE emp ILM ADD POLICY NO INMEMORY AFTER 10 days no access; ALTER TABLE emp ILM ADD POLICY MEMCOMPRESS FOR QUERY after 5 days of creation; In-memory Fast Start syntax: Begin DBMS_INMEMORY_ADMIN.FastStart_ENABLE(‘ ’); End; Tablespace listed should be 2x larger than parameter INMEMORY_SIZE. The data is then check pointed and stored in the DBIMFS_LOGSEG$. The metadata about this feature is stored in the sysaux tablespace. The column store is then loaded from this area on database restart rather than rebuilt from scratch. Dan Hotka Oracle ACE Director Author/Instructor/CEO
↧