1. Smart Flash Cache
Exadata Storage Server layer includes some flash storage as a hardware
component, which has been implemented as a set of PCI flash cards. The main
benefit of that is of course faster access than standard, disk-based access. In
some cases you can order Exadata to KEEP whole table in Smart Flash Cache which
speeds up database layer access with Full Table Scans. On the other hand in
Write-Back mode you can empower DBWR or LGWR performance by putting written
data on flash cache first, and afterwards sync it with regular hard drives,
which of course is transparent to database engine.
2. Storage Indexes
It is
absolutely unique to Exadata. Storage Indexes are an intelligent storage
implementation. Generally speaking classic database index by it definition is
created to efficiently provide a location of a certain data key. To be honest
Storage Indexes in Exadata are focused on eliminating areas on storage as
possible place where data might exist. Online data maps which are completely
transparent for database layer are stored on the Flash Cache of Exadata Storage
Servers. To make the story short, when Exadata Storage Server scans through the
Storage Index and identifies the regions where predicate value falls within the
MIN/MAX for the region, only for the identified regions physical I/O occurs.
And as a consequence, even Full Table Scan has been planned by CBO on Database
Server layer, limited I/O operations use to be proceed.
3. Smart Scans and Cell Offloading
use to be called as secret sauce of Oracle Exadata. The main concept of
Offloading is to move processing from DB Nodes (Database Servers) to the
intelligent storage layer. What is even more important Offloading means the
reduction in the volume of data that returns to database server, which is one of
the major bottlenecks in terabytes or even more bigger databases. To eliminate
the time spent on transferring completely unnecessary data between storage and
the database tier is the main issue that Oracle Exadata has been built to
solve. Keep in mind that Offloading and Smart Scan terms could be used somewhat
4. Hybrid Columnar Compression
known as HCC, Hybrid Columnar Compression is one of the key features of Oracle
Exadata, and it is only available on this engineered system. HCC format of
compression will be used only when data arrives with direct path loads. There
are four levels of compression QUERY LOW (LZO, 4x), QUERY HIGH (ZLIB, 6x),
ARCHIVE LOW (ZLIB, 7x) and ARCHIVE HIGH (Bzip2, 12x). Keep in mind that HCC is
not a good option for OLTP systems. In case of HCC mechanics, HCC store data in
nontraditional format. Even data resides in Oracle blocks, with block header
for every block, in HCC data storage has been organised in logical structures
called compression units (CUs). Each CU consists of multiple Oracle blocks.
Resource Manager (IORM) is a Oracle Exadata features which enriches Oracle
Resource Manager from Database layer. IORM only actively manages I/O requests
when needed and when Storage Server is not fully utilised, it provide data
immediately. But when a disk is under heavy utilisation, Storage Server
software redirects the I/O requests to the appropriate IORM queue and schedules
I/O from there according to the policies defined in your IORM plans. Generally IORM
policies open the way to prioritise databases on intelligent storage layer,
which enable workload optimisation.

  • January 10, 2018 | 18 views
  • Comments