EQUNIX - PPT 11DB-Postgres™.pdf
- 2. 11DB/Postgres™ For Enterprise
11DB/Postgres™ adalah sistem
Manajemen Basis Data Relasional
untuk sistem transaksi berkinerja
tinggi: Perbankan, Telekomunikasi,
Hyper-Retail dan industri lainnya.
11DB/Postgres™ dikembangkan dari
Open Source PostgreSQL dengan
layanan dan fitur yang dibutuhkan
oleh Enterprise.
11DB/Postgres™ merupakan solusi
Enterprise untuk kebutuhan anda.
- 3. 11DB/Postgres™ is a Commercial RDBMS derived
directly from PostgreSQL and well tuned for High
Performance delivery and Enterprise grade
features. It uses Subscription License.
Beside this commercial license, user can always
choose the Open Source one, and use our
Premium Maintenance Support as your warranty
support 24 hours a day, 7 days a week.
More info: https:/
/equnix.asia/product/11db
- 6. Why is it so great?
Easy Deployment and Management
Contrary to normal deployment of the Open Source Software or any kind of Proprietary software which
has its own complexity. 11DB is delivered as a Container image and therefore it is quite easy to deploy.
Integrated HA (High Availability)
High Availability is provided natively, no need to install anything, HA enabled by default.
Comprehensive Monitoring Capability
In other databases, usually monitoring is not provided (for Postgres) and therefore 11DB includes all
essential, yet comprehensive monitoring tools to give information of the running system.
Supports transaction implementation with Multiversion Concurrency Control (MVCC)
In a high load transaction system, usage from application will make a huge traffic at the same time. When
tuned well, 11DB can cater the load as high as possible, while controlling the data versions for each
transaction and maintaining Atomicity, Consistency, and Isolation.
Distributed by:
- 7. Why is it so great?
Has the ability to store data in the form of JSON that can be indexed.
Not only support SQL-level data storage, but 11DB also able to store and query JSON data type, to ease
application processing, especially for APIs
Supports various types of connectors such as JDBC, ODBC, and so on.
11DB has various supports for database-drivers. Therefore application can use modern general driver to
connect.
User defined functions, triggers, and stored procedures.
DBA or developers can create their own function logic, even data type, when necessary, in flexible way.
Able to process high Online Transactions.
11DB has proven to be able to serve under high Online Transactions, up to ~100,000 TPS.
Has a database replication feature.
No 3rd party tools required. No hassle. Replication can be done in no time, without additional complex tools,
since 11DB supports native replication.
Distributed by:
- 8. Why is it so great?
Easy Configuration
The Database has its own tools which queries administrators and scans independently for hardware
configuration.
Automatic Snapshot and Incremental Backup
Data backup is a very important task for administrators, it is a must thing to do after installation to ensure all
backup in place and run well.
Comprehensive metrics analysis for capacity planning
Vital metrics are stored historically every time, in real-time, therefore it can be analyzed to measure usage and
calculate for sizing.
Has good data security and log/audit management features.
Storing data consistently is not enough. Security also should be considered when maintaining sensitive data.
11DB has also native audit and security management capability.
Support view and materialism feature.
View as a short way in storing long and complex queries, can be materialized.
Distributed by:
- 12. 11DB/Postgres as RDBMS for High Transaction
and Processing System, require Vertical
Scalability for its Top Performance delivery.
When Optimization is done for hardware, Data
Structure, Query and Server switches (caching
strategy and timing) Performance able to be
increased up to 20x (depends to how many
optimization can be done)
11DB/Postgres is High Performance RDBMS for
Corporations, runs more than 150K TPS
While Horizontal Scalability is also possible with
some limitations:
1. Offloading Report Query to the Replica with
Logical replication (Slide 6)
2. Offloading simple select query to the Replica
with binary streaming.
3. Application or Data Structure redesign, so it
allow data sharding across clusters.
Executive Summary
- 13. 11DB/Postgrestm
Scalability Strategies
★ RDBMS, is the core of business transactions. It
proceed information exchange in atomic,
consistent, isolated, and durable way. There’s no
other better way than RDBMS to record and guard
transactions in business world.
★ Processing multiple transactions at the same time,
in atomic and isolated way require delicate lock and
coordination span across domains, having multiple
host for single domain drags the locking speed and
therefore potentially occurring frequent deadlocks.
★ The best scalability for highly online transaction and
processing for RDBMS, is vertical scalability.
★ Plus offloading established state of rows into
Historical Database is a great approach to scale out
report inquiries.
★ When report request goes to OLAP Hist Database,
then transaction performance would increase
multiple times.
★ Scale out or vertical scalability is achieved by
hardware optimization, data structure optimization,
query tuning and caching strategy.
- 14. Hardware Optimization
❖ PG_WAL should use the fastest storage speed, in this
case: uses NVMe (PCIe SSD)
❖ Implement Size Provisioning for any installed SSDs in
the System
❖ Use CPU Clock as highest as possible: 2.5Ghz or more.
❖ Use the most dense Cores as possible: AMD EYPC Zen 3
has 64 real cores
❖ Use RAM as many as possible, dedicate ⅓ at least for
Shared Buffer, leave ⅔ for heap and kernel caches.
❖ Prepare SSD or NVMe for Index tablespaces.
❖
❖ …and many mores…
Data structure optimization,
❖ Ensure no blob/bytea or unique index in transaction
table
❖ Use ordinal type of primary key or unique indexes
❖ Referential Integrity constraint should be delivered
with housekeeping strategy/procedures
❖ Avoid using unnecessary stored procedures, every
stored procedure which doesn’t really close to data is
unnecessary.
❖ Implement table partitioning when needed, and able
to be pruned
❖
❖ …and many mores…
- 15. Query tuning
❖ Use prepared statement as much as possible;
prepared statement decrease parsing time and
impact faster in query processing.
❖ Arrange JOIN in appropriate way,
❖ Avoid OR, instead use UNION or IN
❖ Ensure there is an index for every searched filter
column in every queries.
❖ Avoid any kind of unnecessary: column, table,
SELECT cascades, JOIN’s
❖ Run housekeeping periodically, supervised and
anticipated.
❖ ..and others…
Caching strategy
❖ Shared buffer pool is the closest Caching mechanism
to the 11DB/Postgres
❖ Kernel Caches is 2nd layer of caches for file system
data access.
❖ Work mem is prepared per connection basis, and
therefore stated by scripts. Workmem allocation would
be based on index size.
❖ continues…
- 16. In first Quadrant:
★ Main Data Center, RDBMS is installed in High
Availability constellation (In red Square) on
mission critical production system
★ Replicas provided for some purposes: OLAP
Historical Data, Reports, Late Replication for
special purposes
★ Monitoring is included in every implementation
In second Quadrant:
★ Disaster Recovery Center, replicates Master from
its VIP and therefore failover wouldn’t impact DRC
installations.
★ OLAP Server is also replicates if necessary.
General implementation of RDBMS on Enterprises for their mission critical applications in production system.
- 17. In Third Quadrant:
★ Transaction Log Archiving [9] and weekly
snapshot backup. [A]
★ Run snapshot every weeks, incremental
backup for snapshots, incremental backup for
PITR (Point In Time Recovery) when needed.
★ Snapshot creates from last week snapshot
being updated incrementally into current
snapshot every week.
In Fourth Quadrant:
★ Data Warehouse, ETL for data transportation, transformation, Cleansing and Preprocessing (when needed)
★ After ETL, the data is stored accumulated in clustering Data Warehouse such as: Deepgreen, or Greenplum.
★ Business Intelligence can consume this data provided by datawarehouse, both directly or through Data mart.
- 18. 11DB/Postgres has
proven (again) as the
best RDBMS in the world
which able to crunch
transactions using great
hardware efficiently.
Not many great software
able to tune and push to
the limit of hardware
capabilities,
11DB/Postgres is one of
them.
More than 150 Thousands Transaction per second (read and write) shows its real capability
- 20. Equnix Appliance:
Jawaban terbaik untuk kebutuhan
komputasi yang mumpuni
Last but not least, if you wanted to have a complete solution for High
Performance delivery, you may consider: Equnix Appliance
More info: https:/
/equnix.asia/product/appliance