Scribd
Upload
180 views2 pages

In-Memory Columnar Formats in Flash Cache

Uploaded by

Nand Lal Riar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
180 views2 pages

In-Memory Columnar Formats in Flash Cache

Uploaded by

Nand Lal Riar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

When data is initially populated into the IM column store it is read directly from disk in its row format,

converted to a columnar
format and then compressed. The faster you can read the data, the faster you can complete the population process. Exadata storage
offers outstanding IO performance ensuring the data population process is not I/O bound.
The population process is conducted by a set of background worker processes. These worker processes can operate in parallel to
populate the IM column store as fast as data can be read off of disk and CPUs can process that data. This is where the high I/O
performance and CPU resources of Exadata come into play to make the population of the IM column store as fast as possible. The
number of background worker processes can also be controlled to take further advantage of Exadata's scalability.

Database In-Memory will also repopulate IMCUs when the number of stale entries in an IMCU reaches a staleness threshold. Again,
with Exadata's high I/O performance this can occur in the background with no noticeable effect on application performance.

IN-MEMORY COLUMNAR FORMATS IN FLASH CACHE


Exadata uses a unique set of software algorithms to implement database intelligence in storage, NVMe flash storage, and RDMA over
Converged Ethernet. A full rack Exadata X8-2 Database Machine with HC Storage Capacity offers 360TB of flash, which is nearly 30X
the capacity of DRAM, and can achieve up to 350 GB/second of analytic scan bandwidth from SQL1.
It is now possible to store data in the In-Memory columnar format in the flash cache in an Exadata environment. This enables all of
the In-Memory optimizations (accessing only the compressed columns required, SIMD vector processing, storage indexes, etc.) to be
used on a much larger amount of data.
When the INMEMORY_SIZE parameter is set to a non-zero value objects accessed using a Smart Scan will be brought into Exadata
flash cache and will be automatically converted into the In-Memory columnar format. The data will be rewritten in the background
into Database In-Memory columnar format, and this will result in all subsequent accesses to the data benefitting from all of the In-
Memory optimizations when that data is retrieved from the flash cache.

EXCEED DRAM LIMITS AND TRANSPARENTLY SCALE ACROSS MEMORY, FLASH


AND DISK
With Exadata, your application can make use of all storage tiers (memory, flash, & disk) without having to be aware of where the data
resides or suffer suboptimal performance when not all of the data resides in-memory in the IM column store. On Exadata data can
reside in the IM column store, in the database buffer cache, in flash storage in columnar or row format, or on disk storage and your
application never needs to be aware of data location because Oracle Database can seamlessly access that data.
ature or In-Memory Column Cache can dramatically
accelerate Oracle Database processing by speeding I/O operations. Exadata Smart Flash Cache provides intelligent caching of database
objects to avoid physical disk I/O. Exadata storage also provides an advanced compression technology, Hybrid Columnar Compression
(HCC), that typically provides 10x level of data compression and boosts the effective data transfer by an order of magnitude.

1 Exadata Database Machine X8-2 Data Sheet

4 WHITE PAPER | Why Exadata is the Best Platform for Database In-Memory | Version 3.2
Copyright © 2020, Oracle and/or its affiliates | Confidential - Public
This means that all data access, and not just data that has been populated into the IM column store in DRAM, will be as efficient as
possible.

Exadata also includes Smart Scan, a unique technology that offloads data-intensive SQL operations into the Oracle Exadata storage
servers. This is similar to and complements Database In-Memory processing by pushing SQL processing into the Exadata storage
servers when data is not in the IM column store. Data filtering and processing occurs immediately and in parallel across all storage
servers as data is read from disk or flash. Exadata Smart Scan reduces database server CPU consumption and greatly reduces the
amount of data moved between storage and database servers. This enables scaling and efficient SQL processing across all storage tiers
whether data resides in the IM column store, on flash storage or on disk storage.

EXADATA IS ORACLE'S DATABASE IN-MEMORY DEVELOPMENT PLATFORM


Exadata is the development platform for Database In-Memory. Thus, Database In-Memory issues are discovered and fixed on Exadata
first. Exadata is also the primary platform for Oracle Database testing, HA best practices validation, integration and support. The same
reasons it is the best platform for Oracle Database apply to Database In-Memory.

EXADATA IS A DATABASE CONSOLIDATION PLATFORM AND DATABASE IN-


MEMORY FURTHER ENABLES CONSOLIDATION OPPORTUNITIES
Database consolidation is one of the major strategies that organizations use to achieve greater efficiencies in their operations.
Increasing the utilization of hardware resources while reducing administrative costs are primary goals of consolidation projects.
Exadata is optimized for Data Warehouse and OLTP database workloads, and its balanced database server and storage grid
infrastructure make it an ideal platform for database consolidation. Exadata is a modern architecture featuring scale-out industry-
standard database servers, scale-out intelligent storage servers and a high-bandwidth low-latency InfiniBand network that connects
all servers and storage. In many ways Database In- in-memory performance techniques that
are similar to those that are used by Exadata on flahs and disk. Exadata allows customers to simultaneously optimize performance and
cost for analytic workloads by using Database In-Memory columnar formats in-memory and flash in conjunction with existing
Exadata features to increase consolidation capacity for all data. The result is a solution that gives the speed of DRAM, the IOPs of
flash, and the cost effectiveness of disk.

IN-MEMORY FAULT TOLERANCE


Given the shared nothing architecture of the IM column store in a RAC environment, some applications may require a fault-tolerant
option. On Exadata it is possible to mirror the data populated into the IM column store by specifying the DUPLICATE subclause. This
means that each In-Memory Compression Unit (IMCU) populated into the IM column store will have a mirrored copy placed on one
of the other nodes in the RAC cluster. Mirroring the IMCUs provides in-memory fault tolerance as it ensures data is still accessible
via the IM column store even if a node goes down. It also improves performance, as queries can access both the primary and the
backup copy of the IMCU at any time.

IMCUs duplicated on
another node in RAC
cluster

Should a RAC node go down and remain down for some time, the only impact will be the re-mirroring of the primary IMCUs located
on that node. Only if a second node were to go down and remain down for some time would the data have to be redistributed.

5 WHITE PAPER | Why Exadata is the Best Platform for Database In-Memory | Version 3.2
Copyright © 2020, Oracle and/or its affiliates | Confidential - Public

You might also like