Dontcheff

Oracle In-Memory for SAP Databases

In DBA, Oracle database, SAP HANA on March 19, 2016 at 14:02

The Japanese proverb “HANA YORI DANGO” means literally “Dumplings rather than flowers” meaning “to prefer substance over style, as in to prefer to be given functional, useful items (such as dumplings) instead of merely decorative items (such as flowers)”.

HANA is nowadays a very stylish and trendy concept among the database professionals but I would follow the Japanese saying and rather go with substance.

c82-wsj-ad-sap-cloud-motion

The two different points of view can be easily found on the internet:

SAP: What Oracle won’t tell you about SAP HANA
Oracle: Oracle Database In-Memory vs SAP HANA benchmark results

The following “Facts vs Claims” will most likely perplex and bewilder everyone. Just have a look and decide for yourself. Let us not go into details but consider this statement:

FACT: Oracle makes big data a bigger problem with 4 copies of the data (3 in-memory and 1 on disk).

I have been trying for quite some time to figure out the 3 copies of in-memory data that Oracle creates with no success. Perhaps this is a riddle?

An year ago, on March 31st 2015, SAP was certified to run on Oracle Database 12.1.0.2. Rather odd one would say, as in the past SAP were awaiting the 2nd release of the Oracle database in order to certify it for SAP. At least a sign that 12.1.0.2 is mature enough to be used for SAP production systems.

As of June 30th 2015, the Oracle Database In-Memory Option is supported and certified for SAP environments for all SAP products based on SAP NetWeaver 7.x. on Unix/Linux, Windows and Oracle Engineered Systems platforms running Oracle Database 12c – in single instance and Oracle Oracle Real Application Clusters deployments.

In-Memory is such a great feature – but (for both Oracle and SAP) often the challenge you’ll face are these 2 tricky questions:

1. Which of your tables, partitions, and even columns should you mark for In-Memory column store availability?
2. What should be the value of the SGA, PGA and IMCS size? That is, how much memory do I need (global_allocation_limit in SAP and inmemory_size for Oracle)?

Here you have a very strong advantage of Oracle over SAP. The answers are now easier to find with the new In-Memory Advisor which is available via download from MOS Note:1965343.1. Use of In-Memory Advisor for databases where the IM option has not yet been deployed does NOT require an Oracle Tuning Pack license.

For SAP, Quick Sizer is used for a new implementation of Business Suite powered by SAP HANA. Here is a How to Properly Size an SAP In-Memory Database. There are two different approaches for performing the sizing: user-based sizing, which determines sizing requirements based on the number of users in the system, and throughput-based sizing, which bases the sizing on the items being processed. The sizing rules for SAP Business Suite on SAP HANA that are outlined in SAP Note 1793345.

Memory Management in the Column Store: SAP HANA aims to keep all relevant data in memory. Standard row tables are loaded into memory when the database is started and remain there as long as it is running. They are not unloaded. Column tables, on the other hand, are loaded on demand, column by column when they are first accessed. With Oracle, only tables (or columns/partitions of the table) that need to be in memory are set in memory thus avoiding waist of unnecessary memory and having the ability to do more with less.

The Oracle approach is much more sophisticated but there are 10 major issues DBAs should pay attention to when using the In-Memory option with SAP databases:

1. To use Oracle Database In-Memory with SAP NetWeaver the following technical and business prerequisites must be met:
– Oracle Database 12c Release 1 Patch Set 1 (12.1.0.2)
– UNIX/Linux: Oracle Database SAP Bundle Patch June 2015 (SAP1202P_1506) or newer. Strongly recommended Oracle Database SAP Bundle Patch August 2015 (SAP1202P_1508)
– Windows: Windows DB Bundle Patch 12.1.0.2.6 or newer. Strongly Recommended Windows DB Bundle Patch 12.1.0.2.8
– SAP NetWeaver 7.x Version with minimum SAP Kernel 7.21_EXT

2. Indexes. It is not allowed to make any changes to the standard index design of the SAP installations. However, customer specific index design can be changed. That is, all indexes which belong to the Y or Z namespaces can be changed.

3. Database Buffer Cache. It is not allowed to reduce the size of the database buffer cache and assign the memory to the In-Memory column store.

4. SAP Dictionary Support. Full SAP Dictionary (DDIC) Support of in-memory attributes at the table level starts with the support package SAP_BASIS 7.40 SP12.

5. Individual Columns. It is not supported to load individual columns of an SAP table or partition into the IM column store. It is also not supported to exclude individual columns from an SAP table or partition from the IM column store. An SAP table is a database table used by an SAP application.

6. The database where you want to run In-Memory Advisor must have XDB component installed as the In-Memory Advisor relies on functions provided by XDB. In 12c XDB is installed by default.

7. The In-Memory Advisor is contained in the SAP Bundle Patch as patch 21231656.

8. For SAP applications it is strongly recommended to use a reasonable time window of collected AWR data. At least 2-3 days of AWR data should be used for the In-Memory Advisor. It absolutely makes no sense to use data from a 1-2 hour time window.

9. Use these In-Memory Advisor Parameter Name and Value:

– WRITE_DISADVANTAGE_FACTOR = 0.7
– LOB_BENEFIT_REDUCTION = 1.2
– MIN_INMEMORY_OBJECT_SIZE = 1024000
– READ_BENEFIT_FACTOR = 2

10. For SAP systems therefore the following init.ora parameters should be used:

inmemory_max_populate_servers = 4
inmemory_clause_default = “PRIORITY HIGH”
inmemory_size should be set to the value (+ ~20% for metadata and journals) used in the generate recommendation step of the In-Memory Advisor or set to value calculated by the SAP_IM_ADV Package for all tables/partitions to be loaded into the In-Memory column store.

Using RAT shows the benefit of the In-Memory option. Compare the DB time of the 2 replays after going to Exadata with the In-Memory option. DB time is the best and possibly only metric to compare captures with replays.

IMDB_Advisor

SAP Notes: 2178980 – Using Oracle Database In-Memory with SAP NetWeaver
MOS Notes: 1292089.1 – Master Note for Oracle XML Database (XDB) Install / Deinstall & 1965343.1 – Oracle Database In-Memory Advisor
Using SAP NetWeaver with Oracle Database In-Memory

Bottom line: if my SAP application runs on top of an Oracle database, I would rather put it on Exadata with the In-Memory option enabled that move it to SAP HANA.

Oracle Database Cloud Service vs Amazon Relational Database Service

In Cloud, Consolidation, DBA, Oracle database on February 28, 2016 at 15:00

How to compare Oracle’s Database Public Cloud with Amazon’s Relational Database Service (RDS) for enterprise usage? Let us have a look.

Oracle’s Database has 4 editions: Personal Edition, Express Edition (XE): free of charge and used by very small businesses and students, Standard Edition (SE): light version of Enterprise Edition and purpose designed to lack most features needed for running production grade workloads and Enterprise Edition (EE): provides the performance, availability, scalability, and security required for mission-critical applications.

In the comparison in this post, we will evaluate Oracle and Amazon in relation to the Enterprise Edition of Oracle’s database.

Oracle_Public_Database_Cloud

Oracle Public Database Cloud consists of 4 DB Cloud offerings: DBaaS, Virtual Image, Schema Service and Exadata Service. Here are few characterizations:

– Oracle supports Exadata, RAC & all DB options
– Simple pricing structure with published costs representing actual costs (unlimited I/Os, etc.)
– Hourly, Monthly & Annual pricing options
– Lowest cloud storage pricing across all major IaaS vendors

Amazon RDS for Oracle Database supports two different licensing models – “License Included” and “Bring-Your-Own-License (BYOL)”. In the “License Included” service model, you do not need separately purchased Oracle licenses. Here are few characterizations:

Enterprise Edition supports only db.r3.large and larger instance classes, up to db.r3.8xlarge
– Need to choose between Single-AZ (= Availability Zone) Deployment and Multi-AZ Deployment
– For Multi-AZ Deployment, Amazon RDS will automatically provision and manage a “standby” replica in a different Availability Zone (prior to failover you cannot directly access the standby, and it cannot be used to serve read traffic)
– Only 2 instance types support 10 Gigabit network: db.m4.10xlarge and db.r3.8xlarge
– Amazon RDS for Oracle is an exciting option for small to medium-sized clients and includes Oracle Database Standard Edition in it’s pricing
– Several application with limited requirements might find Amazon RDS to be a suitable platform for hosting a database
– As the enterprise requirements and resulting degree of complexity of the database solution increase, RDS is gradually ruled out as an option

So, here is high level comparison:

Oracle_Cloud_vs_Amazon_RDS
Notes:

– Oracle’s price includes the EE license with all options
– Amazon AWS is BYOL for EE
– Prices above are based on the EU (Frankfurt) region
– Amazon’s Oracle database hour prices vary from $0.290 to $4.555 for Single AZ Deplyoments and from $0.575 to $9.105 for Multi-AZ Deployments
– Oracle’s database hour prices vary from $0.672 to $8.569

Sources:

Oracle Archive Storage Pricing
Amazon Glacier Storage Pricing
Amazon Database Pricing
Oracle Database Pricing
Amazon Options for Oracle Database Engine
Oracle on Amazon RDS Support & Limitations

So, Amazon RDS is not an option if you need any of the following: Real Application Clusters (RAC), Real Application Testing, Data Guard / Active Data Guard, Oracle Enterprise Manager Grid Control, Automated Storage Management, Database Vault, Streams, Java Support, Locator, Oracle Label Security, Spatial, Oracle XML DB Protocol Server or Network access utilities such as utl_http, utl_tcp, utl_smtp, and utl_mail.

Interesting articles related to this topic:

1. Burning question for Oracle: What’s your response to Amazon? by Barb Darrow
2. Shootout: Oracle DB Cloud vs. Amazon RDS by Jan Navratil
3. The Oracle Database Cloud Service vs Oracle on Amazon RDS by Ranko Mosic
4. A Most Simple Cloud: Is Amazon RDS for Oracle Right for you? by by Jeremiah Wilton
5. Oracle RAC and AWS: A Hybrid Cloud Solution by Lindsay Van Thoen
6. How Much Does It Cost to Run Relational Database (RDS) Options on AWS by Yoav Mor
7. Oracle vs. Amazon: The Cloud Wars by Chris Lawless

The James Bond of Database Administration

In Data, DBA, Golden Gate, Oracle database, Oracle Engineered Systems on October 27, 2015 at 07:23

“Defending our systems needs to be as sexy as attacking others. There’s really only one solution: Bond.”

That is what ‘The Guardian’ wrote recently in an article entitled “The Man with the Golden Mouse: why data security needs a James Bond“.

Attending the annual Oracle ACE Director Briefing at Oracle HQ awoke up an interesting debate on the following question: What will happen in the near future with the DBA profession? Who is now the James Bond of Database Administration?

JB007

According to TechTarget, big data tools are changing data architectures in many companies. The effect on the skill sets required by database administrators may be moderate, but some new IT tricks are likely to be needed. GoldenGate is the new Streams, Exadata is the new RAC, Sharding the new Partitioning, Big Data is the new data (Texas is an exception), you name it…

Having the privilege to work throughout the years with some of the best database experts in the world has, for all it matters, proved to me that Double-O-Sevens are in fact more like Double-O-six-hundreds. Meaning that there are 100s of DBAs that qualify with no hesitation whatsoever as the James Bonds of Database Administration. I have learned so much from my ex Nokia colleagues, from my current Enkitec and Accenture colleagues. Not to mention friends from companies like eDBA, Pythian, Miracle, etc.

A DBA needs to have so many skills. Look for instance at Craig S. Mullins’ suggested 17 skills required of a DBA. Kyle Hunter’s article The evolution of the DBA and the Data Architect is clearly pointing to the emerging skillsets in the Data Revolution.

In the IT business, and in database administration in particular, it is not that important how well you know the old stuff, it is more important how fast you can learn the new things. Here are some of the tools that help every modern Oracle DBA:

Oracle Enterprise Manager 12c
ORAchk
Metalink/MOS
Developer Tools
Oracle Application Express (APEX)
SQL Developer
Oracle JDeveloper
SQL Developer Data Modeler
And last but not least SQL*Plus®

7init

These additional Metalink tools might be often of great help:

Diagnostic Tools Catalog – Note ID 559339.1
OS Watcher (Support Tool) – Note 301137.1
LTOM (Support Tool) – Note 352363.1
HANGFG (Support Tool) – Note 362094.1
SQLT (Support Tool) – Note 215187.1
PLSQL Profiler (Support Script) – Note 243755.1
MSRDT for the Oracle Lite Repository – Note 458350.1
Trace Analyzer TRCANLZR – Note 224270.1
ORA-600/ORA-7445 Error Look-up Tool – Note 153788.1
Statspack (causing more problems than help in 12c)

The Man with the Golden Mouse is the James Bond of Database Administration. The best DBA tools are still knowledge and experience.

GoldenMouse

Follow

Get every new post delivered to your Inbox.

Join 371 other followers