Wikipedia

Oracle Exadata: Difference between revisions

Browse history interactively
← Previous edit
Content deleted Content added
VisualWikitext
TS-220657 (talk | contribs)
60 edits
m More X8M updates
m →Use cases: reword
 
(45 intermediate revisions by 38 users not shown)
Line 1:
{{Short description|Computing platform specialized to the Oracle Database}}
{{Infobox software
| title = Oracle Exadata
Line 6 ⟶ 7:
| released = October 2008
| operating system = [[Oracle Linux]]
| platform = Exadata Database Machine, Exadata Database Service, Exadata Cloud@Customer
| license = Commercial
| website = {{URL|http://www.oracle.com/exadata}}
}}
The '''Oracle Exadata''' '''Database Machine''' ('''Exadata'''<ref name=":30">{{Cite web|url=https://www.oracle.com/a/ocom/docs/engineered-systems/exadata/exadata-x8m-2-ds.pdf|title=Oracle Exadata Database Machine X8M-2|last=Various|first=|date=September 2019|website=oracle.com|url-status=live|archive-url=|archive-date=|access-date=September 19, 2019}}</ref>) is a computing platform that is specialized and optimized for running [[Oracle Database]]. The goal of Exadata<ref name=":10">{{Cite web|url=https://www.youtube.com/watch?v=ZHlFDgci9Fc|title=Oracle Exadata: World's Fastest Database Machine|last=Loaiza|display-authors=etal|date=August 2, 2011|website=YouTube|archive-url=|archive-date=|dead-url=|access-date=July 27, 2018}}</ref> is to achieve higher performance and availability at lower cost by moving database algorithms and intelligence into storage and networking, bypassing the traditional processing layers.<ref>{{Cite web|url=https://www.oracle.com/technetwork/database/exadata/trn4113-exadatadeepdive-5187039.pdf|title=Oracle Exadata: Architecture and Internals Technical Deep Dive|last=Umamageswaran and Goindi|first=Kothanda and Gurmeet|date=October 23, 2018|website=www.Oracle.com/technetwork|archive-url=|archive-date=|dead-url=|access-date=April 21, 2019}}</ref>
[[File:Exadata x8-2M.jpg|alt=Exadata X8-2M Full Rack|left|thumb|275x275px|''Exadata X8M-2 Full Rack'']]
Exadata is a combined hardware and software platform that includes [[Scale out|scale-out]] compute servers, scale-out intelligent storage servers, ultra-fast networking, [[3D XPoint|persistent memory]] (PMEM), [[NVM Express|NVMe]] flash, and specialized Exadata Software<ref name=":27">{{Cite book|url=https://www.oracle.com/pls/topic/lookup?ctx=en/engineered-systems/exadata-database-machine&id=SAGUG20314|title=Oracle Exadata System Software User's Guide 19.2.0|last=Stern|first=Janet|publisher=Oracle Corporation|year=2019|isbn=|___location=|pages=}}</ref> in a wide range of shapes and price points. Exadata Storage uses high-performance servers to store data and run Exadata Software to run data-intensive database processing directly in the shared storage tier.
 
[[File:Larry Ellison and Exadata.jpg|thumb|[[Larry Ellison]] and Exadata (2009)]]
Exadata debuted<ref>{{Cite web|url=https://siliconangle.com/2019/06/20/oracles-pioneering-database-machine-exadata-turns-10-exclusive-look-ahead/|title=Exadata Turns 10|last=Vellante|first=Dave|date=June 20, 2019|website=siliconangle.com|archive-url=|archive-date=|dead-url=|access-date=June 21, 2019}}</ref> in 2008 as the first in Oracle Corporation's family of ''Engineered Systems''<ref>{{Cite web|url=http://chucksblog.typepad.com/chucks_blog/2015/09/grown-up-it-for-grown-up-applications.html|title=Chuck's Blog: Grown-up IT for Grown-Up Applications|last=Hollis|first=Chuck|date=September 9, 2015|website=Typepad|archive-url=|archive-date=|dead-url=|access-date=July 27, 2018}}</ref> for use in corporate [[Data center|data centers]] deployed as "[[Private Cloud|private clouds]]". In October 2015, Exadata became available in the [[Oracle Cloud]] as a subscription service, known as the ''Exadata Cloud Service''.<ref>{{Cite web|url=https://www.oracle.com/technetwork/database/exadata/exadataservice-ds-2574134.pdf|title=Oracle Database Exadata Cloud Service Technical Data Sheet|last=Various authors|date=|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=July 27, 2018}}</ref>
'''Oracle Exadata''' ('''Exadata'''<ref name=":30">{{Cite web |last=Various |date=July 11, 2024 |title=Oracle Exadata |url=https://www.oracle.com/engineered-systems/exadata/ |access-date=July 11, 2024 |website=oracle.com}}</ref>) is a computing system optimized for running [[Oracle Database]]s.
 
Exadata is a combined [[database machine]] and software platform that includes [[Scale out|scale-out]] [[x86-64]] compute and storage servers, [[RDMA over Converged Ethernet|RoCE]] networking, RDMA-addressable memory acceleration, [[NVM Express|NVMe]] flash, and specialized software.<ref name=":0">{{Cite web|last=Pedregal-Martin|first=Cristobal|title=Exadata: Why and What|url=https://blogs.oracle.com/exadata/exadata-why-and-what}}</ref>
<ref>{{Cite book|title=Oracle Database Exadata Cloud Service: A Beginner's Guide|last=Spendolini|first=Brian|publisher=Oracle Press|year=2019|isbn=978-1260120875|___location=Amazon.com|pages=}}</ref> Oracle databases deployed in the Exadata Cloud Service are 100% compatible with databases deployed on Exadata on-premises, which enables customers to transition to the Oracle Cloud with no application changes. Oracle Corporation manages this service, including hardware, network, Linux software and Exadata software, while customers have complete ownership of their databases.
 
Exadata was introduced in 2008 for on-premises deployment, and since October 2015, via the [[Oracle Cloud]] as a subscription service, known as the ''Exadata Database Service on Dedicated Infrastructure,''<ref>{{Cite web |last=Various |date=July 11, 2024 |title=Oracle Exadata Database Service on Dedicated Infrastructure |url=https://www.oracle.com/engineered-systems/exadata/#dedicated-infrastructure |access-date=July 11, 2024 |website=oracle.com}}</ref> and ''Exadata Database Service on Exascale Infrastructure''.<ref>{{Cite web |last=Various |date=July 11, 2024 |title=Exadata Database Service on Exascale Infrastructure |url=https://www.oracle.com/engineered-systems/exadata/#exascale |access-date=July 11, 2024 |website=oracle.com}}</ref> ''Exadata Cloud@Customer''<ref>{{Cite web |last=Various |date=July 11, 2024 |title=Oracle Exadata Cloud@Customer |url=https://www.oracle.com/engineered-systems/exadata/#exadata-cloudatcustomer |access-date=July 11, 2024 |website=oracle.com}}</ref> is a hybrid cloud (on-premises) deployment of Exadata Database Service.
In early 2017, a third Exadata deployment choice became available. ''Exadata Cloud at Customer''<ref name=":0">{{Cite web|url=http://www.oracle.com/technetwork/database/exadata/exacc-x7-ds-4126773.pdf|title=Oracle Database Exadata Cloud at Customer Technical Data Sheet|last=Various authors|date=|website=Oracle Corporation|archive-url=|archive-date=July 27, 2018|dead-url=|access-date=}}</ref> is Exadata Cloud Service technology deployed on-premises (behind the corporate firewall) and managed by Oracle Cloud experts. Like the Exadata Cloud Service, Exadata Cloud at Customer is owned and managed by Oracle, and licensed through a pay-as-you-go subscription. The Oracle Cloud at Customer<ref>{{Cite web|url=https://www.oracle.com/cloud/paas/oracle-cloud-at-customer/resources.html|title=Oracle Cloud at Customer Resources: Data Sheet et al|last=Various authors|date=|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=July 27, 2018}}</ref> program is intended to bring all the benefits of the Oracle public cloud while still satisfying security and regulatory constraints.
 
Starting December, 2023, Exadata Database Service became available for Microsoft Azure, Google and AWS public clouds within the ''Oracle Database@Azure'', ''Oracle Database@Google Cloud'' and ''Oracle Database@AWS'' [https://www.oracle.com/cloud/multicloud/ multicloud partnerships].
 
==Use cases==
Exadata is designed to optimally run anyall Oracle Database workload or combination of workloads, such as an [[OLTP]], applicationData running simultaneously withWarehousing, Analytics, processing.and TheAI platformVector isprocessing, frequentlyoften usedwith tomultiple consolidate manyconsolidated databases that were previously running on dedicated database serverssimultaneously. Exadata's scale-out architecture is naturally suited to running in the Oracle Cloud, where computing requirements can dynamically grow and sometimes shrink.
 
Historically, specialized [[database computing platformsmachine]]s were designed for a particular workload, such as Data Warehousing, and poor or unusable for other workloads, such as OLTP. Exadata hasspecializes optimizations for all database workloads, implemented such thatin mixed workloads sharesharing system resources fairly.with Resourceresource management features also allow for prioritized allocation of system resourcesprioritization, such as always favoring workloads servicing interactive users over reporting and batch. Long running requests, evencharacterized ifby theyData Warehouses, reports, batch jobs and Analytics, are accessingreported theto samerun datamany times faster compared to a conventional, non-Exadata database server.<ref>{{Cite web |last=Various |date=July 11, 2024 |title=Exadata Customer Success Stories |url=https://www.oracle.com/customers/?search=exadata |access-date=July 11, 2024 |website=oracle.com}}</ref><ref>{{Cite web |last=Various |date=July 11, 2024 |title=Gartner Peer Insights: Oracle Exadata Database Machine |url=https://www.gartner.com/reviews/market/integrated-systems/vendor/oracle/product/oracle-exadata-database-machine |access-date=July 11, 2024 |website=Gartner.com}}</ref>
 
== Release History ==
Long running requests, characterized by Data Warehouses, reports, batch jobs and Analytics, are reputed to run many times faster compared to a conventional, non-Exadata database server.<ref>{{Cite web|url=https://www.oracle.com/search/customers?Ntt=exadata&Dy=1&Nty=1&Ntk=S1|title=Exadata Customer Success Stories|last=|first=|date=|website=|archive-url=|archive-date=|dead-url=|access-date=July 16, 2018}}</ref><ref>{{Cite web|url=https://www.gartner.com/reviews/market/data-warehouse-solutions/vendor/oracle/product/oracle-exadata-database-machine|title=Gartner Peer Insights: Oracle Exadata Database Machine|last=Various|date=|website=Gartner.com|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref> Customer references often cite performance gains of [http://www.oracle.com/us/products/database/exadata-reference-booklet-400018.pdf 10x or greater]. Analytics workloads can also use the ''Oracle Database In-Memory''<ref>{{Cite web|url=http://www.oracle.com/technetwork/database/options/database-in-memory-ds-2210927.pdf|title=Data Sheet: Oracle Database In-Memory|last=Various|first=|date=2019|website=Oracle Technology Network|archive-url=|archive-date=|dead-url=|access-date=April 21, 2019}}</ref> option on Exadata for additional acceleration, and In-Memory Databases on Exadata have been extended to take advantage of Flash memory capacity, many times larger than the capacity of [[Dynamic random-access memory|DRAM]]. Exadata’s ''Hybrid Columnar Compression''<ref name=":27" /> feature is intended to reduce the storage consumption of Data Warehouses and archival data as well as increase performance by reducing the amount of IO.
{| class="wikitable"
 
! Exadata Release
Transactional (OLTP) workloads on Exadata benefit from the incorporation of persistent memory and Flash memory into Exadata’s storage hierarchy, and the automatic "tiering" of data into DRAM, persistent memory, Flash or disk storage. Special algorithms optimize persistent memory and Flash for response time sensitive database operations such as log writes. For the most demanding OLTP, all-Flash storage eliminates the latency of disk media completely.
! Primary Software Enhancements
 
! Primary Hardware Enhancements
== Design concepts ==
The hardware components that make up a typical database computing platform are a compute server connected over a network to a storage array. The database software runs on the compute server and sends or receives database information to and from the storage array over the network. The hardware components use standard software protocols to "talk" to each other. This separation via standard interfaces is what allows a computing platform to run a wide variety of software and hardware from different vendors. All of the application logic and the processing of the data is performed on the compute server, to which all the data must be sent. With this approach, a computing platform can be used for a wide range of software applications, though it will not be optimized for any particular application.
 
The goal of Exadata was to create a complete [[Solution stack|stack]] of software and hardware focused on the Oracle Database, that allowed processing to be moved to its optimal ___location. If Exadata is only processing Oracle Database requests it can take advantage of that in all the software layers. The hardware design could include elements that are most advantageous to Oracle Database applications, such as very fast [[InfiniBand]] and Ethernet networking, [[Flash memory]] and [[3D XPoint|persistent memory]]. Given the importance of data storage to databases, Oracle was particularly focused on optimizing that aspect of the Exadata platform.
 
Oracle wanted a storage layer for Exadata that could easily [[scale out]] and parallelize Oracle Database requests. It also recognized the opportunity for storage to cooperate in the processing of database requests beyond just storing and shipping data. For example, rather than send an entire database table across the network to the compute server to find a small number of records, such data filtering could be done in storage and only the resulting records sent across the network. The addition of Flash memory and later persistent memory to Exadata Storage Servers also opened up a range of possibilities for optimizing performance in the storage layer. Over time, as the performance and capacity of Flash storage increased at a rapid rate, the network became a performance bottleneck for traditional database platforms and Exadata's offloading of database processing into Exadata Storage Servers avoided that problem. The subsequent addition of persistent memory in the storage layer exacerbated the traditional platform bottleneck even further.
 
The foundation of Exadata is the ''Exadata Storage Server''<ref>{{Cite web|url=https://www.oracle.com/technetwork/database/exadata/exadata-technical-whitepaper-134575.pdf|title=A Technical Overview of the Oracle Exadata Database Machine and Exadata Storage Server|last=Weiss|first=Ronald|date=2012|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 2, 2018}}</ref><ref name=":14">{{Cite web|url=https://www.oracle.com/technetwork/database/exadata/exadata-x8-2-ds-5444350.pdf|title=Data Sheet: Oracle Exadata Database Machine X8-2|last=|first=|date=2019|website=Oracle Corporation|archive-url=|archive-date=|access-date=April 21, 2019|dead-url=}}</ref>, invented by Oracle to replace the traditional storage array. Also important is Oracle's ownership of all the main software and hardware components of Exadata, enabling changes to be deeply integrated and released in coordinated fashion. A further benefit for customers is the ability to support the entire Exadata platform from one vendor.
 
== Software ==
Software enhancements specific to Exadata achieved better performance in some areas due to the integration of hardware and software. For example, when an OLTP application commits a transaction to the database software, that request is viewed by Exadata as a critical operation and prioritized accordingly within the network and in the storage servers. The commit request will move in front of less important messages on the network and in the IO queue<ref name=":13">{{Cite web|url=https://www.youtube.com/watch?v=5SWPA3FA-HQ&index=5&list=PLKCk3OyNwIzsMPOS-d73HAuEbDUp9Hglq|title=Guaranteeing Fast Response for OLTP Messages|last=Loaiza|first=Juan|date=July 19, 2016|website=YouTube|archive-url=|archive-date=|dead-url=|access-date=August 5, 2018}}</ref>.
 
Another example is the use of Flash memory for caching data that is being accessed by Analytics applications. Because a columnar data format is more effective for Analytics workloads, when Exadata moves the data row from disk to Flash, it automatically reformats the data into a columnar format<ref>{{Cite web|url=https://blogs.oracle.com/in-memory/columnar-formats-in-exadata-flash-cache|title=Columnar Formats in Exadata Flash Cache|last=Rivenes|first=Andy|date=October 30, 2017|website=Oracle Corporation blogs|archive-url=|archive-date=|dead-url=|access-date=August 6, 2018}}</ref>.
 
Both examples require the Oracle Database to understand the intent of the application and to cascade this understanding to the network and storage software, which behave accordingly.
 
Since Exadata's debut in 2008, there have been at least two significant Exadata software releases per year, delivering dozens of "smart" software enhancements. Most of these enhancements are based on a smaller number of technical foundations, as explained below.
 
=== Technical concepts ===
 
*''Offloading to Storage'' - refers to the execution of data-intensive database operations within the Exadata Storage Servers, such as data scans, table joins and filtering of rows and columns. This reduces the amount of data that must ultimately be returned over the network to the compute servers, thus avoiding the network bottleneck. Offloading is possible because Exadata storage is built on standard servers, capable of running database functions in coordination with the Oracle Database, simultaneous with storage IO. Over time, more database functions and more data types have been offloaded. In addition, "reverse offloading" will push an operation back to the compute servers if Exadata storage is too busy.
*''Storage Indexes'' - enable the avoidance of IO by tracking column values within small regions of storage. If the Storage Index confirms that an IO to a region will not find a match, that IO isn't performed, and a costly and unnecessary operation is avoided. Storage indexes are automatically maintained and kept in memory on Exadata Storage Servers. Initially Storage Indexes tracked value ranges within a small number of columns. Over time, more columns and more sophisticated value tracking have been added, so that additional IO operations can be avoided.
*''Flash and PMEM Caching'' - delivers the low latency (fast response) of Flash and persistent memory, while preserving the lower cost of disk for storing large databases. The goal is the best IO performance at the lowest cost. In general, a small percentage of a database is active at any one time. If the active data is all held in Flash, for instance, the IO performance would be equal to all-Flash storage, at a much lower cost. Exadata Flash and PMEM caching use intelligence about the applications currently running to keep the most active data in Flash or persistent memory, in the optimal format. For example, Exadata knows when an IO is part of a database backup, and not an indication of an active data block, whereas traditional platforms view any IO as a "hot" block. Flash caching will also reformat rows into columnar format in Flash if the data is being accessed for Analytics. Initially, Flash caching was only used for reading data, then extended to log writes and all other write IO. Later, Flash was used as an extension of Oracle's Database In-Memory columnar data store, for significantly larger in-memory databases than DRAM capacity alone. PMEM caching debuted with Exadata X8M, adding an even faster cache in storage and substantially improving IO performance and latency.
*''Hybrid Columnar Compression (HCC)'' - reduces the amount of storage consumed by Data Warehouses and other infrequently updated data. Without compression, Data Warehouses can grow to enormous sizes. Exadata's Hybrid Columnar Compression benefits from the greater compressibility of columnar formats and avoids the performance pitfall of an all-columnar approach. Conventional data compression algorithms yield between 2x and 4x compression, whereas HCC averages between 10x and 15x compression. Such a large reduction in the amount of IO can also substantially improve performance. Initially, HCC tables did not support row-level locking, limiting their use with OLTP applications. In 2016, support for row-level locking was added to HCC on Exadata, improving the performance of mixed workloads with HCC data.
*''Resource Management'' - allocates Exadata system resources, such as CPU, IO and network bandwidth, to databases, applications or users based on priorities. When Exadata is used for consolidating many databases, Resource Management is usually employed to ensure appropriate quality of service. IO Resource Management debuted in V1 of Exadata. Network Resource Management was added in Exadata X4.
*''In-Memory Databases'' - offer exceptional performance for Analytics workloads, leveraging DRAM on compute servers, a complement to Exadata's emphasis on storage and networking. Oracle Database In-Memory became available in 2014 on Exadata, leveraging its fast InfiniBand network for in-memory Fault Tolerance. To support larger in-memory databases, Exadata Storage Servers implement in-memory routines and data formats in Exadata Flash, as an extension of the same in-memory processing that occurs on compute servers.
 
=== Software enhancements ===
A more detailed listing of software enhancements is below, grouped by their value to Analytics or OLTP workloads, or their impact on Availability. Similar enhancements cannot be duplicated on other platforms because they require software and [[Application programming interface|API]] modifications and integration across database software, operating systems, networking and storage.
 
Refer to the Exadata documentation<ref>{{Cite book|title=Oracle Exadata Database Machine Documentation Release (19.2)|last=Stern|first=Janet|publisher=Oracle Corporation|year=2019|isbn=|___location=https://docs.oracle.com/en/engineered-systems/exadata-database-machine/index.html|pages=}}</ref> and Data Sheet<ref name=":30" /> for descriptions of these features.
{|
|- style="vertical-align:top"
|
{|
|-
! For ANALYTICS
|-
|Database@AWS
|Exadata Database Service available with AWS
|
* Automatically parallelize and offload data scans to storage
|-
| rowspan="6" |X11M - Jan 2025
|
|AI Vector search acceleration - up to 55% faster
* Filter rows in storage based on 'where' clause
|25% faster compute core performance
|-
|Analytics scan throughput increase - 2.2x faster
|
|33% greater server memory bandwidth
* Filter rows in storage based on columns selected
|-
|Transaction processing acceleration - 25% faster
|
|11% faster storage core performance
* JSON and XML offload
|-
|OLTP read latency acceleration - up to 21% faster (14 microseconds)
|
|PCIe 5 performance-optimized flash
* Filter rows in storage based on join with other table
|-
|Intelligent power management - reduce CPU cores, cap power consumption, optimize power utilization
|
|X11M-Z database and storage servers
*''Hybrid Columnar Compression''
|-
|Available on-premises, Oracle Cloud, Cloud@Customer and multicloud (Azure, Google Cloud, AWS)
|
*''[[Storage Index]]'' data skipping
|-
|Database@Google Cloud
|Exadata Database Service available with Google Cloud
|
*''IO Resource Management'' by user, query, service, DB
|-
| rowspan="3" |Exadata Exascale
|
July, 2024
* Automatic transformation to columnar format in Flash Cache
|Fully elastic pay-per-use architecture. Users specify the cores and storage capacity needed, reducing entry-level infrastructure costs for Exadata Database Service and aligning costs with usage
| rowspan="3" |None
|-
|Large pools of shared compute and storage allow databases to quickly scale over time without concern for server-based size limitations or disruptive migrations
|
*''Smart Flash Cache'' for table scans
|-
|Rapid and efficient database snapshots and thin cloning
|
* Offload index fast full scans
|-
|Database@Azure
|Exadata Database Service available with Microsoft Azure
|
* Offload scans on encrypted data, with FIPS compliance
|-
| rowspan="5" |X10M - June 2023
|
|Exadata RDMA Memory (XRMEM) DRAM cache
* Storage offload for LOBs and CLOBs
|3x increase in compute cores (96-core AMD EPYC)
|-
|Oracle Linux 8 and UEK 6 kernel updates
|
|1.5x higher memory capacity
* Storage offload for min/max operations
|-
|New In-Memory Columnar compression algorithm
|
|2.5x faster DDR5 memory
* Data Mining offload
|-
|Optimized Smart Scan for more complex queries
|
|2.4x higher flash storage capacity (in all-flash storage)
* All ports active InfiniBand messaging
|-
|Faster decryption and decompression
|
|22% more disk storage capacity
* Reverse offload to DB servers if storage CPUs are busy
|-
| rowspan="6" |X9M - Sept, 2021
|
|Secure RDMA fabric isolation
* Database In-Memory automatic memory population/depopulation
|PCIe 4.0 dual-port active-active 100&nbsp;Gb RoCE network
|-
|Smart Flash Log write-back
|
|33% increase in compute cores
* In-Memory support for external tables
|-
|Storage Index and Columnar Cache persistence
|
|33% increase in memory capacity
* In-Memory optimized arithmetic
|-
|Faster decryption and decompression Algorithms
|
|28% increase in disk capacity
* Automatics statistic and indexing
|-
|Smart Scan performance optimizations
|
|1.8x greater internal fabric bandwidth (PCIe 4.0)
* Just in time smart columnar decryption
|-
|
|1.8x greater flash bandwidth (PCIe 4.0)
* Smart aggregation with columnar cache
|}
|
|
{|
|-
| rowspan="4" |X8M - Sept, 2019
! For AVAILABILITY
|RoCE: RDMA over Converged Ethernet
|Persistent Memory (PMEM) in storage
|-
|Persistent Memory Data Accelerator
|
|100 Gbit/s internal fabric (2.5x increase)
* Instant detection of node or cell failure
|-
|Persistent Memory Commit Accelerator
|
| rowspan="2" |
*''In-Memory Fault Tolerance''
|-
|KVM virtual machine support
|
* Sub-second failover of IO on stuck disk or flash
|-
| rowspan="3" |X8 - April, 2019
|
|AIDE: Advanced Intrusion Detection Environment
* Offload backups to storage servers
|Storage Server Extended (XT)
|-
|ML-based monitoring and auto-indexing
|
|40% increase in disk capacity
* Exadata data validation (H.A.R.D.)
|-
|Real-time updates of optimizer statistics
|
|60% increase in storage processor cores
* Instant data file creation
|-
| rowspan="3" |X7 - Oct, 2017
|
|In-memory database in flash storage
* Prioritize rebalance of critical files
|2x increase in flash capacity
|-
|DRAM cache in storage
|
|25% increase in disk capacity
* Automatic hard disk scrub and repair
|-
|Large-scale storage software updates
|
|25 Gbit/s data center Ethernet support
* Power cycle failed drives to eliminate false drive failures
|-
|Exadata Cloud@Customer
|Exadata Cloud Service on-premises
|
* Avoid reading predictive failed disks
|-
| rowspan="3" |X6 - April, 2016
|
|Exafusion direct-to-wire OLTP protocol
* Cell software transparent restart
|2x increase in flash capacity
|-
|Smart Fusion Block Transfer
|
|10% increase in compute cores
* Flash and disk life cycle management alert
|-
|Smart Flash Log
|
|2x increase in memory capacity
* Confinement of temporarily poor performing drives
|-
|Exadata Database Service
|Exadata on Oracle Cloud Infrastructure (OCI)
|
* Prevent shutdown if mirror server is down
|-
| rowspan="5" |X5 - Dec, 2014
|
|In-memory database fault tolerance
*''Automatic Software Updates'' on an entire "fleet" of Exadata systems with one operation
|2x increase in flash & disk capacity
|-
|Database snapshots
|
|Elastic configurations
* Hot pluggable Flash cards
|-
|Xen virtual machine support
|
|All-flash storage server option
* Keep standby database consistent when NO FORCE logging is used
|-
|NVMe flash protocol support
|
|50% increase in compute cores
* Fast, secure eraser of disk and Flash
|-
|IPv6 support
|
|50% increase in memory capacity
* Advanced Intrusion Detection Environment (AIDE) detects and alerts when unknown changes to system software are made
|-
| rowspan="4" |X4 - Nov, 2013
|
|Network Resource Management
* Automatic monitoring of CPU, network and memory using Machine Learning
|2x increase in flash capacity
|}
|
|
{|
|-
|I/O latency capping
! For OLTP
|2x increase in memory capacity
|-
|Capacity-on-Demand licensing
|
|50% increase in compute cores
* Database Aware PCI Flash
|-
|Active/Active InfiniBand (2x increase)
|
|33% increase in disk capacity
* Exadata ''Smart Flash Logging''
|-
| rowspan="5" |X3 - Sept, 2013
|
*''Write-back|Smart Flash Cache'' write-back
|Eighth-Rack configuration
|-
|Improved management of slow disks/flash
|
|4x increase in flash capacity
* IO Prioritization by DB, user, or workload to ensure QoS
|-
|Sub-second brownout after storage failure
|
|33% increase in compute cores
*''Direct-to-Wire Protocol''
|-
|Simplified disk replacement
|
|75% increase in memory capacity
*''Network Resource Management''
|-
|Bypass predictive disk failure
|
|2x increase in data center bandwidth
* EXAchk full-stack validation
|-
| rowspan="7" |X2 - Sept, 2010
|
|Smart Flash Log
* Full-stack security scanning
|8-socket (X2-8) configuration
|-
|Auto Service Request
|
|Storage Expansion Rack
* NVMe flash interface for lowest latency IO
|-
|Secure Erase of storage
|
|Hardware-based decryption
* Active AWR includes storage stats for end to end monitoring
|-
|Platinum Services
|
|50% increase in compute cores
* Database scoped security
|-
| rowspan="3" |
|
|2x increase in memory capacity
* Cell-to-cell rebalance preserving flash cache
|-
|50% increase in disk capacity
|
* In-Memory commit cache
|-
|8x increase in data center bandwidth
|
* Memory optimized OLTP and IoT lookups
|-
| rowspan="5" |v2 - Sept, 2009
|
|Storage Indexes
* Automatics statistic and indexing
|Flash storage
|-
|Database-aware Smart Flash Cache
|
|Quarter-Rack configuration
* RDMA over Converged Ethernet
|-
|Hybrid Columnar Compression
|
|2x increase in memory & disk capacity
* Persistent memory data accelerator (PMEM cache)
|-
| rowspan="2" |
|
|3x increase in data center bandwidth
* Persistent memory commit accelerator (PMEM log)
|}
|}
 
==Database software==
Exadata compute servers run the [[Oracle Linux]] 7 operating system and Oracle Database 11g Release 2 Enterprise Edition through Oracle Database 19c. Exadata system resources can be optionally virtualized using the Xen-based Oracle VM (prior to X8M) or KVM-based Oracle VM (X8M and later). All Oracle Database options, such as Real Application Clusters, Multitenant, Database In-Memory, Advanced Compression, Advanced Security, Partitioning, Active Data Guard and others are optionally available with Exadata. Applications that are certified to a supported version of the Oracle Database are automatically compatible with Exadata. No additional modifications or certifications are required<ref name=":11">{{Cite web|url=https://www.oracle.com/technetwork/database/exadata/exadata-statementofdirection-2417679.pdf|title=Oracle Exadata Statement of Direction|last=Various authors|first=|date=2019|website=|archive-url=|archive-date=|dead-url=|access-date=April 21, 2019}}</ref>.
 
The same database software that runs on Exadata on-premises will run in the Exadata Cloud Service and Exadata Cloud at Customer. In addition, on-premises software licenses are eligible for the BYOL<ref>{{Cite web|url=http://www.oracle.com/us/cloud-essentials-byolauc-4419224.pdf|title=Bring Your Own License and Universal Credits|date=2018|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 7, 2018}}</ref> (Bring Your Own License) transfer into the Oracle Cloud or Cloud at Customer.
 
==Networking==
Exadata provides high-speed networks for internal and external connectivity. A 100 gigabits per second (100 Gbit/s) Ethernet fabric is used for internal connectivity between compute and storage servers and 25, 10 and 1 Gbit/s Ethernet ports are included for data center connectivity. The Ethernet fabric is also used as the cluster interconnect between compute servers.
 
Exadata uses a custom-designed, database-oriented protocol over the Ethernet fabric to achieve higher performance. It makes extensive use of [[remote direct memory access]] (RDMA over Converged Ethernet - RoCE) to improve efficiency by avoiding operating system overhead and data copies when moving data between servers. Exadata also has a ''direct-to-wire protocol''<ref>{{Cite web|url=https://www.youtube.com/watch?v=OMZogy7rkoI&index=9&list=PLKCk3OyNwIzsMPOS-d73HAuEbDUp9Hglq|title=Direct Access from Database to Network for Much Faster OLTP|last=Loaiza|first=Juan|date=|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref> that allows the database to "talk" directly to the Ethernet hardware.
 
Exadata also takes advantage of RoCE Class of Service in its ''Network Resource Management''<ref name=":13" /> feature to prioritize important traffic across the network. In this feature the Oracle Database software tags network messages that require low latency, such as transaction commits, lock messages and IO operations issued by interactive users, enabling them to bypass messages issued by less critical high-throughput workloads such as reporting and batch. The result is analogous to how an emergency vehicle with its siren on is able to move more quickly through heavy traffic - high-priority network messages are moved to the front of the server, network switch, and storage queues, bypassing lower-priority messages and resulting in shorter and more predictable response times.
 
==Management software==
Oracle Enterprise Manager Cloud Control (EM) manages Oracle software and hardware, including the Exadata Database Machine. EM integrates with the built-in Exadata management tooling, as well as with customers' existing systems management and helpdesk tools. The Exadata plug-in for EM provides an integrated view of compute servers, storage servers, switches, and topology. It also provides discovery, monitoring and alerting capability for Exadata systems management.
 
The Exadata Cloud Service and Exadata Cloud at Customer platforms are managed by Oracle, while customers control and manage the software and databases running on the compute servers. Exadata Cloud platforms include a web-based provisioning wizard through which customers can quickly provision their chosen Exadata system and subsequently their database instances. Customers also perform familiar database administration and OS administration tasks aided by cloud automation for backup, patching, and upgrades
 
== Hardware ==
Exadata is available in two models: one based on two-[[CPU socket|socket]] compute servers and the other based on eight-[[Cpu socket|socket]] compute servers. The two models differ only in the hardware used for the compute servers. The networking, storage servers and software are the same in both models.
 
The most recent Oracle Exadata Database Machine is the X8M<ref>{{Cite web|url=https://video.oracle.com/detail/video/6086091482001|title=Introducing Oracle Exadata X8M|last=Loaiza|first=Juan|date=September 2019|website=oracle.com|url-status=live|archive-url=|archive-date=|access-date=September 19, 2019}}</ref> generation, introduced in September 2019.
 
The X8M-2 compute servers feature a small form factor, 1 RU (Rack Unit) in height. They employ 2-socket Intel Xeon processors; each socket with 24 compute cores for 48 total cores per compute server. Memory starts at 384 gigabytes (GB) and can be expanded to 1.5 terabytes (TB).
 
The Exadata Database Machine base configuration has 2 compute servers and 3 storage servers, referred to as a ''Quarter Rack''. The same hardware is also available in an ''Eighth Rack'' configuration with half of the processing and half of the storage capacity. As the database workload and/or data size increases, additional compute and storage servers may be added to increase the volume of work performed in parallel, using Exadata's ''Elastic Configuration.''
 
The X8M-8 compute server uses eight-socket compute servers that consume 5 RU in height and have greater memory capacity than the X8M-2, with 3 to 6 TB per server. Whereas each X8M-2 compute server contains 48 compute cores, each X8M-8 server contains 192 compute cores. This allows large database workloads to easily [[Scalability|scale-up]] within a compute server while still supporting Exadata’s [[Scalability|scale-out]] expandability across multiple servers. The larger memory capacity of the X8M-8 also favors In-Memory Databases and very large OLTP, consolidation, and Data Warehouse workloads. Like the X8M-2, the Exadata X8M-8 base configuration has 2 compute servers and 3 storage servers, but consumes a ''Half Rack'' of space. Additional compute and storage servers may be added until the rack is full.
 
===Storage servers===
There are three choices for Exadata storage servers: ''Extreme Flash (EF),'' ''High Capacity (HC)'' ''and Extended (XT)''. The X8M-2 Extreme Flash Storage Server is all-flash storage containing 8 [[PCI Express|PCIe]] flash drives for a total of 51.2 TB of raw storage capacity and 1.5 TB of persistent memory. The X8M-2 High Capacity storage server contains 12 disks, 14 TB each, for a total of 168 TB of raw storage capacity, plus 25.6 TB of PCIe flash and 1.5 TB of persistent memory to cache active data blocks. Exadata’s ''Smart Cache'', ''Smart Log'', ''Columnar Flash Cache'' and ''Write Back Flash Cache'' features determine how and when to use flash.
 
In addition to adding storage servers into an Exadata Database Machine base configuration, storage servers may also be acquired with or added to ''Exadata Storage Expansion''<ref name=":28">{{Cite web|url=http://www.oracle.com//technetwork/database/exadata/exadata-storage-expansion-x7-2-ds-4428337.pdf|title=Data Sheet: Oracle Exadata Storage Expansion Rack X7-2|date=2017|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref> racks.
 
Storage performance specifications for a full rack Exadata configuration are as follows:
{| class="wikitable sortable"
|-
|40 Gbit/s internal fabric (2x increase)
! Exadata Storage Server !! Scan Rate !! Read IOPS !! Write IOPS
|-
| rowspan="6" |v1 - Sept, 2008
| X8M-2 Extreme Flash || 560 GB/sec || 16,000,000* || 6,580,000
|Oracle Enterprise Linux
|Scale-out 4-socket compute servers
|-
|Smart Scan (storage offload)
| X8M-2 High Capacity || 350 GB/sec || 12,000,000 || 6,580,000
|Scale-out 4-socket storage servers
|}
<small>* Based on an elastic configuration of 11 compute and 11 storage servers. All other rates are based on a standard full rack configuration of 8 compute and 14 storage servers<ref>{{Cite web|url=http://www.oracle.com/technetwork/database/exadata/oeca-download-2817713.html|title=Oracle Exadata Configuration Assistant (OECA)|website=Oracle}}</ref></small>
 
<small>Note: IOPS = 8K IO Operations per Second from SQL</small>
 
With the introduction of Exadata X8, Oracle added ''Exadata Storage Server Extended (XT)'' for low-cost storage of infrequently accessed data. The XT storage server contains the same disk capacity as the X8M-2 High Capacity Storage Server, but does not contain Flash storage or require the Exadata Storage Server Software license.
 
=== Memory-level performance with shared storage ===
Architects of traditional computing platforms have always had to cope with technology change affecting the design of their systems. The goal is to eliminate bottlenecks so that the output of storage moves through the network and is processed by compute servers without any slowdown. Solving an imbalance generally involves adding faster or more network connections or compute servers. This was before the advent of ultra-fast PCIe [[Flash memory]] and the [[NVM Express|NVMe]] Flash interface, and before persistent memory.
 
Flash memory started to became mainstream in corporate computing around 2010, as a cache in front of hard disks or as a replacement for disks entirely. Every year thereafter Flash capacity and performance increased significantly. In 2017, leading-edge Flash performance crossed a threshold, where the most advanced networks were unable to match the performance of Flash, and became a substantial bottleneck. As an example, a popular all-Flash storage system with 480 Flash cards is rated at only 75 GB/s of data throughput, whereas without a network bottleneck, that many Flash cards should produce over 2,600 GB/s of data throughput. Plugging Flash directly into a compute server removes the network bottleneck, but also removes the ability to share storage with multiple compute servers. Offloading to storage in Exadata is a way to bypass this network bottleneck by filtering out unneeded data in storage before sending the remaining data across the network. The addition of persistent memory, faster than Flash, increased the value of Exadata offloading even further.
 
=== Hardware specifications ===
{| class="wikitable"
|-
|IORM (I/O Resource Manager)
! Exadata Generation
|20 Gbit/s internal fabric (InfiniBand)
(2-socket)*
! V1 !! V2!! X2-2!! X3-2!! X4-2!! X5-2!! X6-2!! X7-2
!X8-2
!X8M-2
|-
|Join filtering (Bloom filters)
| Date introduced || Sep-2008 || Sep-2009 || Sep-2010 || Sep-2012 || Nov-2013 || Jan-2015 || Apr-2016 || Oct-2017
|1 Terabyte disks
|Apr-2019
|Sep-2019
|-
|Incremental backup filtering
|Last ship date
|1 Gbit/s data center network (Ethernet)
|Oct-2009
|Oct-2010
|Sep-2012
|Feb-2014
|Mar-2015
|Jul-2016
|Nov-2017
|still shipping
|still shipping
|still shipping
|-
|Smart file creation
|Specifications|| {{n/a}}
|
|<small>Datasheet</small><ref>{{Cite web|url=http://www.oracle.com/us/industries/healthcare/058454.pdf?ssSo%20urceSiteId=ocomjp|title=Data Sheet: Sun Oracle Database Machine|date=2009|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 13, 2018}}</ref>
|<small>Datasheet</small><ref name=":18">{{Cite web|url=http://www.oracle.com/technetwork/database/exadata/dbmachine-x2-2-datasheet-175280.pdf|title=Data Sheet: Oracle Exadata Database Machine X2-2|date=2012|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref>
|<small>Datasheet</small><ref name=":19">{{Cite web|url=http://www.oracle.com/technetwork/database/exadata/exadata-dbmachine-x3-2-ds-1855384.pdf|title=Data Sheet: Oracle Exadata Database Machine X3-2|date=2013|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref>
|<small>Datasheet</small><ref name=":20">{{Cite web|url=https://www.aamratechnologies.com/wp-content/uploads/2014/09/Oracle-Exadata-Database-Machine-X4-2-Data-Sheet.pdf|title=Data Sheet: Oracle Exadata Database Machine X4-2|date=2013|website=AAMRA Technologies|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref>
|<small>Datasheet</small><ref name=":21">{{Cite web|url=https://www.dynamicsystemsinc.com/Downloads/exadata-x5-2-ds-2406241.pdf|title=Data Sheet: Oracle Exadata Database Machine X5-2|date=2015|website=Dynamic Systems Inc.|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref>
|<small>Datasheet</small><ref name=":22">{{Cite web|url=http://www.oracle.com/technetwork/database/exadata/exadata-x6-2-ds-2968790.pdf|title=Data Sheet: Exadata Database Machine X6-2|date=2017|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref>
|<small>Datasheet</small><ref>{{Cite web|url=https://www.oracle.com/technetwork/database/exadata/exadata-x7-2-ds-3908482.pdf|title=Oracle Exadata Database Machine X7-2 data sheet|last=various|first=|date=2019|website=oracle.com/exadata|archive-url=|archive-date=|dead-url=|access-date=April 21, 2019}}</ref>
|<small>Datasheet</small><ref name=":14" />
|<small>Datasheet</small><ref name=":30" />
|-
| Operating system || Linux || Linux || Linux || Linux || Linux || Linux || Linux || Linux
|Linux
|Linux
|-
| Disk storage (raw TB) || 168 || 336 || 504 || 504 || 672 || 1344 || 1344 || 1680
|2352
|2352
|-
| Flash cache (raw TB) || {{n/a}} || 5.3 || 5.3 || 22.4 || 44.8 || 89.6 || 179.2 || 358.4
|358.4
|358.4
|-
|Persistent memory (TB)|| {{n/a}} || {{n/a}} || {{n/a}} || {{n/a}} || {{n/a}} || {{n/a}} || {{n/a}} || {{n/a}} || {{n/a}}
|21
|-
| Extreme flash (raw TB) || {{n/a}} || {{n/a}} || {{n/a}} || {{n/a}} || {{n/a}} || 179.2 || 358.4 || 716.8
|716.8
|716.8
|-
| Compute cores || 64 || 64 || 96 || 128 || 192 || 288 || 352 || 384
|384
|384
|-
| Max memory (Gio) || 256 || 576 || 1,152 || 2,048 || 4,096 || 6,144 || 12,288 || 12,288
|12,288
|12,288
|-
| Ethernet (Gb/s) || 8 || 24 || 184 || 400 || 400 || 400 || 400 ||800
|800
|800
|-
! Exadata Generation
(8-socket)**
!! {{n/a}} !! {{n/a}} !! X2-8!! X3-8!! X4-8!! X5-8!! X6-8!! X7-8
!X8-8
!X8M-8
|-
| Date introduced || {{n/a}} || {{n/a}} || Sep-2010 || Sep-2012 || Jul-2014 || Nov-2015 || Apr-2016 || Oct-2017
|Apr-2019
|Sep-2019
|-
|Last ship date|| {{n/a}} || {{n/a}}
|Nov-2012
|Dec-2014
|Oct-2015
|Mar-2016
|Nov-2017
|still shipping
|still shipping
|still shipping
|-
|Specifications|| {{n/a}} || {{n/a}}
|<small>Datasheet</small><ref>{{Cite web|url=http://www.oracle.com/technetwork/database/exadata/dbmachine-x2-8-datasheet-173705.pdf|title=Data Sheet: Oracle Exadata Database Machine X2-8|date=2012|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref>
|<small>Datasheet</small><ref name=":23">{{Cite web|url=http://www.oracle.com/technetwork/database/exadata/exadata-dbmachine-x3-8-ds-1855388.pdf|title=Data Sheet: Oracle Exadata Database Machine X3-8|date=2012|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref>
|<small>Datasheet</small><ref name=":24">{{Cite web|url=http://www.oracle.com/technetwork/database/exadata/exadatax4-8datasheet-2243041.pdf|title=Data Sheet: Oracle Exadata Database Machine X4-8|date=2014|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref>
|<small>Datasheet</small><ref name=":25">{{Cite web|url=http://www.oracle.com/technetwork/database/exadata/exadata-x5-8-ds-2745934.pdf|title=Data Sheet: Oracle Exadata Database Machine X5-8|date=2015|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref>
|<small>Datasheet</small><ref name=":26">{{Cite web|url=http://www.oracle.com/technetwork/database/exadata/exadata-x6-8-ds-2968796.pdf|title=Data Sheet: Oracle Exadata Database Machine X6-8|date=2017|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref>
|<small>Datasheet</small><ref>{{Cite web|url=http://www.oracle.com/technetwork/database/exadata/exadata-x7-8-ds-3938980.pdf|title=Data Sheet: Oracle Exadata Database Machine X7-8|date=2017|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref>
|<small>Datasheet</small><ref>{{Cite web|url=https://www.oracle.com/a/ocom/docs/engineered-systems/exadata/exadata-x8-8-ds.pdf|title=Oracle Exadata Database Machine X8-8 data sheet|last=Various|first=|date=2019|website=oracle.com/exadata|archive-url=|archive-date=|dead-url=|access-date=April 21, 2019}}</ref>
|<small>Datasheet</small><ref>{{Cite web|url=https://www.oracle.com/a/ocom/docs/engineered-systems/exadata/exadata-x8m-8-ds.pdf|title=Oracle Exadata Database Machine X8M-8|last=Various|first=|date=September 2019|website=oracle.com|url-status=live|archive-url=|archive-date=|access-date=September 19, 2019}}</ref>
|-
| Operating system || {{n/a}} || {{n/a}} || Linux || Linux || Linux || Linux || Linux || Linux
|Linux
|Linux
|-
| Disk storage (raw TB) || {{n/a}} || {{n/a}} || 504 || 504 || 672 || 1344 || 1344 || 1680
|2352
|2352
|-
| Flash cache (raw TB) || {{n/a}} || {{n/a}} || 5.3|| 22.4 || 89.6 || 89.6 || 179.2 || 358.4
|358.4
|358.4
|-
|Persistent memory (TB)|| {{n/a}} || {{n/a}} || {{n/a}} || {{n/a}} || {{n/a}} || {{n/a}} || {{n/a}} || {{n/a}} || {{n/a}}
|21
|-
| Extreme flash (raw TB) || {{n/a}} || {{n/a}} || {{n/a}} || {{n/a}} || 179.2 || 179.2 || 358.4 || 716.8
|716.8
|716.8
|-
| Compute cores || {{n/a}} || {{n/a}} || 96 || 160 || 240 || 288 || 288 || 384
|384
|384
|-
| Max memory (TB) || {{n/a}} || {{n/a}} || 4 || 4 || 12 || 12 || 12 || 12
|12
|12
|-
| Ethernet (Gb/s) || {{n/a}} || {{n/a}}|| 176 || 176 || 180 || 180 || 180 ||540
|540
|540
|}
<nowiki>*</nowiki> Per 1 full rack with 8 compute servers and 14 storage servers
 
<nowiki>**</nowiki> Per 1 full rack with 2 compute servers and 14 storage servers
 
== Elastic configurations ==
Prior to the X5-2 generation, Exadata systems were only available in fixed-size configurations of Eighth, Quarter, Half and Full Rack sizes. With the X5-2 Exadata release in January, 2015, ''elastic configurations'' were introduced. An elastic configuration has a customer-specified combination of database servers and storage servers. Elastic configurations allow individual storage or compute servers to be added to a base configuration until the physical rack is full. For example, an Exadata system optimized for in-memory database processing could be created by adding many compute servers, each with maximum memory. Conversely, an Exadata system optimized for a large data warehouse could be configured by adding many High-Capacity storage servers. The ratio of compute to storage servers can vary, depending on the characteristics of the intended workload. Elastic configurations may also be used to scale out earlier generation Exadata systems using the latest compatible servers. In addition, Exadata Database Machines have always been able to span multiple racks using the built-in network fabric connections. Thus, Exadata’s scale-out extends beyond a single physical rack.
 
== Evolution ==
Oracle Corporation releases a new generation of Exadata every twelve to eighteen months<ref name=":1">{{Cite web|url=http://www.oracle.com/us/corporate/press/017557_EN|title=Oracle Press Release: Oracle Introduces the HP Oracle Database Machine|last=Ledbetter|first=Letty|date=September 24, 2008|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=July 27, 2018}}</ref><ref name=":2">{{Cite web|url=http://www.oracle.com/us/corporate/press/033684|title=Oracle Press Release: Oracle Unveils Exadata Version 2: The First Database Machine for OLTP|last=Ledbetter|first=Letty|date=September 15, 2009|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=July 27, 2018}}</ref><ref name=":3">{{Cite web|url=http://www.oracle.com/us/corporate/press/173663|title=Oracle Press Release: Oracle Introduces Exadata Database Machine X2-8|last=Ledbetter|first=Letty|date=September 20, 2010|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=July 27, 2018}}</ref><ref name=":4">{{Cite web|url=http://www.oracle.com/us/corporate/press/1855412|title=Oracle Press Release: Oracle Announces Oracle Exadata X3 Database In-Memory Machine|last=Ledbetter|first=Letty|date=October 1, 2012|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=July 27, 2018}}</ref><ref name=":5">{{Cite web|url=http://www.oracle.com/us/corporate/press/2079925|title=Oracle Press Release: Oracle Launches Oracle Exadata Database Machine X4|last=Ledbetter|first=Letty|date=December 11, 2013|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=July 27, 2018}}</ref><ref name=":6">{{Cite web|url=https://www.oracle.com/corporate/pressrelease/data-center-012115.html|title=Oracle Press Release: Oracle Tackles Data Center Cost and Complexity with Next Generation Engineered Systems|last=Whitaker|first=Teri|date=January 21, 2015|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=July 27, 2018}}</ref><ref name=":7">{{Cite web|url=https://www.oracle.com/corporate/pressrelease/exadata-x6-database-machine-040516.html|title=Oracle Press Release: New Oracle Exadata X6 Database Machine Delivers Next-Generation Flash Performance, Massive Storage Capacity, and Real-Time OLTP|last=Maloney|first=Nicole|date=April 5, 2016|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=July 27, 2018}}</ref><ref name=":8">{{Cite web|url=https://www.oracle.com/corporate/pressrelease/oow17-oracle-exadata-x7-100217.html|title=Oracle Press Release: New Oracle Exadata X7 Delivers In-Memory Performance from Shared Storage|last=Maloney|first=Nicole|date=October 2, 2017|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=July 27, 2018}}</ref><ref name=":29">{{Cite web|url=https://www.oracle.com/corporate/pressrelease/oracle-exadata-builds-in-machine-learning-061219.html|title=New Oracle Exadata Builds in Machine Learning Advances, Supercharges Performance, Improves Cost Effectiveness|last=Maloney|first=Nicole|date=June 12, 2019|website=Oracle|archive-url=|archive-date=|dead-url=|access-date=June 14, 2019}}</ref>. At each release, Oracle refreshes most hardware components to the latest Intel Xeon processors, memory, disk, flash and networking. The hardware refreshes in themselves result in performance increases with every release. Exadata software is also refreshed with each generation and periodically in between, enhancing some combination of performance, availability, security, management and workload consolidation. In October 2015, features to support the Oracle Cloud were introduced<ref name=":9">{{Cite web|url=https://www.oracle.com/corporate/pressrelease/dbcs-oow-102615.html|title=Oracle Press Release: Oracle Expands Database Cloud Portfolio with Unprecedented Enterprise-Class Capabilities|last=Whitaker|first=Teri|date=October 26, 2015|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=July 27, 2018}}</ref>.
 
The emphasis of each Exadata generation is described below.
 
''Exadata V1''<ref name=":1" />, released in 2008, focused on accelerating Data Warehousing by delivering the full throughput of storage to the database. Per Oracle, Exadata achieved this by moving database filtering operations into storage, instead of sending all data to the compute servers and filtering it there. Oracle refers to this capability as ''Exadata Smart Scan''<ref>{{Cite web|url=https://www.youtube.com/watch?v=EhAlkOQwuAM&index=2&list=PLKCk3OyNwIzsMPOS-d73HAuEbDUp9Hglq|title=Exadata Smart Scan: Moving Queries to Data is Required to Realize Flash Performance|last=Loaiza|first=Juan|date=July 19, 2018|website=YouTube|archive-url=|archive-date=|dead-url=|access-date=August 5, 2018}}</ref><ref>{{Cite web|url=http://kerryosborne.oracle-guy.com/papers/Understanding%20Exadata%20Offloading.pdf|title=Understanding Exadata Offloading|last=Osborne|first=Kerry|date=August 3, 2011|website=Kerry Osborne's Oracle Blog|archive-url=|archive-date=|dead-url=|access-date=July 27, 2018}}</ref>''.'' Exadata V1 also supported a consolidation feature for allocating IO bandwidth between databases or workloads, called ''IORM'' (IO Resource Manager)<ref>{{Cite book|title=Oracle Exadata System Software User's Guide, 18c (18.1)|last=Stern|first=Janet|publisher=Oracle Corporation|year=2018|isbn=|___location=|url=https://docs.oracle.com/cd/E80920_01/SAGUG/exadata-storage-server-iorm.htm#SAGUG20421|pages=12}}</ref>.
 
Exadata V1 was available in ''Full Rack'' or ''Half Rack'' sizes, and the choice of ''High Performance'' or ''High Capacity'' storage servers.
 
''Exadata V2''<ref name=":2" /><ref name=":16">{{Cite web|url=https://www.oracle.com/technetwork/database/exadata/exadata-datasheet-1-129084.pdf|title=Data Sheet: Sun Oracle Exadata Storage Server|date=2010|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref><ref name=":17">{{Cite web|url=https://www.oracle.com/technetwork/oem/extensions/index-160766.html|title=Sun Oracle Database Machine and Exadata Storage Server|date=2009|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref>, released in 2009, added a ''Quarter Rack'' configuration and support for OLTP workloads via [[Flash memory|Flash storage]] and database-aware ''Flash Caching''.<ref>{{Cite web|url=http://www.oracle.com/technetwork/database/exadata/exadata-smart-flash-cache-366203.pdf|title=White Paper: Exadata Smart Flash Cache Features and the Oracle Exadata Database Machine|last=Subramaniam|first=Mahesh|date=December 1, 2013|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=July 27, 2018}}</ref>
 
Exadata V2 also introduced ''Hybrid Columnar Compression''<ref name=":15">{{Cite web|url=http://www.oracle.com/technetwork/database/database-technologies/performance/hybridcolumnarcompression-oow2010-200703.pdf|title=Hybrid Columnar Compression (HCC) on Oracle Database 18c|date=June 1, 2018|website=Oracle Technology Network|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref> to reduce the amount of storage consumed by large Data Warehousing tables.
 
''Storage Indexes''<ref>{{Cite web|url=https://www.youtube.com/watch?v=J64gw7Aal_c&index=1&list=PLKCk3OyNwIzsMPOS-d73HAuEbDUp9Hglq|title=Storage Indexes: By Understanding Data, Exadata Can Radically Reduce I/O|last=Loaiza|first=Juan|date=July 19, 2016|website=YouTube|archive-url=|archive-date=|dead-url=|access-date=August 5, 2018}}</ref> in Exadata V2 increased performance by eliminating the need to read entire regions of storage, based on knowledge of the data contained in the region.
 
''Exadata X2-2''<ref name=":18" />, the third generation, was released in 2010 and a second model of Exadata, ''Exadata X2-8''<ref name=":3" />, was introduced. The X2-8 and subsequent “8 socket” Exadata models feature Intel processors targeted at large memory, [[Scalability|scale-up]] workloads. The use of Flash storage beyond caching began in this release with a ''Smart Flash Logging''<ref>{{Cite web|url=https://www.youtube.com/watch?v=_2AyotxL0N8&index=3&list=PLKCk3OyNwIzsMPOS-d73HAuEbDUp9Hglq|title=Smart Logging Enables Fast OLTP on Flash|last=Loaiza|first=Juan|date=July 19, 2016|website=YouTube|archive-url=|archive-date=|dead-url=|access-date=August 5, 2018}}</ref><ref>{{Cite web|url=https://www.youtube.com/watch?v=KFTFKMnr5pc|title=Video: Oracle Exadata Smart Flash Log|last=Fusek|first=Peter|date=November 20, 2011|website=YouTube|archive-url=|archive-date=|dead-url=|access-date=August 3, 2018}}</ref> feature. Support for ''10 Gigabit per second (Gb/sec) Ethernet'' connectivity was also added.
 
Data security through encryption was encouraged with the incorporation of ''hardware decryption''<ref>{{Cite web|url=https://software.intel.com/en-us/blogs/2012/12/26/securing-databases-with-intel-advanced-encryption-standard-new-instructions-aes-ni|title=Securing Databases with Intel Advanced Encryption Standard New Instructions (AES-NI)|last=Nguyen|first=Khang T|date=December 26, 2012|website=Intel|archive-url=|archive-date=|dead-url=|access-date=August 3, 2018}}</ref> in Exadata X2-2, largely eliminating the performance overhead compared to software decryption.
 
A ''Storage Expansion Rack''<ref>{{Cite web|url=http://www.oracle.com//technetwork/database/exadata/exadata-storage-expansion-x7-2-ds-4428337.pdf|title=Data Sheet: Oracle Exadata Storage Expansion Rack X7-2|last=Various authors|date=2017|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 3, 2018}}</ref> based on Exadata X2-2 was added in 2011 to accommodate large, fast-growing Data Warehouses and archival databases. All subsequent 2-socket Exadata generations have included a new Storage Expansion Rack.
 
''Exadata X3-2''<ref name=":4" /><ref name=":12">{{Cite web|url=http://www.nocoug.org/Journal/NoCOUG_Journal_201305.pdf|title=A Brief History of Exadata Time|last=Gorbachev|first=Alex|date=May 1, 2013|website=NoCOUG Journal|archive-url=|archive-date=|dead-url=|access-date=August 5, 2018}}</ref><ref name=":19" /> and ''X3-8''<ref name=":23" /> were released in 2012, including a new ''Eighth Rack X3-2'' entry-level configuration. ''Flash storage capacity quadrupled'' and OLTP write throughput reportedly increased by 20x via the ''Write-Back Flash Cache''<ref>{{Cite web|url=http://www.oracle.com/technetwork/articles/database/exadata-write-back-flash-2179184.html|title=Oracle Exadata Database Machine - Write-Back Flash Cache|last=Gomez|first=Deiby|date=|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 3, 2018}}</ref> feature.
 
A number of availability enhancements were added, ''bypassing slow or failed storage media''<ref>{{Cite web|url=https://www.youtube.com/watch?v=4ScPLSbapS0&index=6&list=PLKCk3OyNwIzsMPOS-d73HAuEbDUp9Hglq|title=Instant Server Failure Detection|last=Loaiza|first=Juan|date=July 19, 2016|website=YouTube|archive-url=|archive-date=|dead-url=|access-date=August 5, 2018}}</ref>, ''reducing the duration of storage server brownouts'' and ''simplifying replacement of failed disks.''
 
''Exadata X4-2''<ref name=":5" /><ref name=":20" /> was released in 2013. ''Flash capacity doubled'' and ''Flash compression'' was added, effectively doubling capacity again. ''Network Resource Management''<ref name=":13" /> was introduced, automatically prioritizing critical messages. InfiniBand bandwidth doubled with support for ''active/active connections''.
 
''Exadata X4-8''<ref name=":24" /> released in 2014, plus ''Capacity on Demand''<ref>{{Cite book|title=Exadata Database Machine Licensing Information User's Guide, 18c|last=Stern|first=Janet|publisher=Oracle Corporation|year=2018|isbn=|___location=|url=https://docs.oracle.com/cd/E80920_01/DBMLI/exadata-capacity-on-demand.htm#DBMLI147|pages=5-6}}</ref> licensing, ''IO latency capping'' and ''timeout thresholds''.
 
''Exadata X5-2''<ref name=":6" /><ref name=":21" /> and ''X5-8''<ref name=":25" /> were released in 2015 with a major set of enhancements. ''Flash and disk capacity doubled''. ''Elastic configurations''<ref>{{Cite book|title=Exadata Database Machine System Overview|last=Stern|first=Janet|publisher=Oracle Corporation|year=2018|isbn=|___location=|url=https://docs.oracle.com/cd/E80920_01/DBMSO/exadata-elastic-configurations.htm#DBMSO22093|pages=14-17}}</ref> were introduced to enable expansion one server at a time. ''Virtualization'' was added as an option to Exadata along with ''Trusted Partitions''<ref>{{Cite web|url=https://blogs.oracle.com/exadata/controlling-software-costs-on-exadata|title=Controlling Software Costs on Exadata|last=Goindi|first=Gurmeet|date=February 19, 2015|website=Oracle Blogs|archive-url=|archive-date=|dead-url=|access-date=August 3, 2018}}</ref> for flexible licensing within a virtual machine. Database ''snapshots''<ref>{{Cite web|url=http://www.oracle.com/technetwork/database/exadata/learnmore/exadata-database-copy-twp-2543083.pdf|title=White Paper: Storage Efficient Database Copy Options with Exadata|last=Shah|first=Manish|date=March 1, 2015|website=Oracle Technology Network|archive-url=|archive-date=|dead-url=|access-date=August 3, 2018}}</ref> on Exadata storage enabled efficient development and testing. ''Oracle Database In-Memory'' on Exadata included ''Fault Tolerant''<ref>{{Cite web|url=https://www.youtube.com/watch?v=YFrF1N8lWpI&index=7&list=PLKCk3OyNwIzsMPOS-d73HAuEbDUp9Hglq|title=Duplicating In-Memory Data Assures Availability|last=Loaiza|first=Juan|date=July 19, 2016|website=YouTube|archive-url=|archive-date=|dead-url=|access-date=August 5, 2018}}</ref><ref>{{Cite web|url=https://www.youtube.com/watch?v=z-qSDImrBmo|title=Oracle Database In-Memory Fault Tolerance Explained|last=Colgan|first=Maria|date=March 10, 2017|website=YouTube|archive-url=|archive-date=|dead-url=|access-date=August 3, 2018}}</ref>''redundancy''. The High Performance Exadata storage servers were replaced with ''all-Flash (Extreme Performance) storage servers'' and Exadata became the first major vendor to adopt the ''[[NVM Express|NVMe]] Flash interface''. ''Columnar Flash cache'' was introduced to automatically reformat analytics data into row format in Flash. ''[[IPv6]] support'' was completed. ''Exadata Cloud Service''<ref name=":9" /><ref>{{Cite web|url=http://www.oracle.com/technetwork/database/exadata/exadataservice-ds-2574134.pdf|title=Data Sheet: Oracle Databade Exadata Cloud Service|date=2017|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 8, 2018}}</ref> was launched on the Oracle Cloud.
 
''Exadata X6-2''<ref name=":7" /><ref name=":22" /> and ''X6-8''<ref name=":26" /> were released in 2016. ''Flash capacity doubled''. E''xafusion Direct-to-Wire protocol''<ref>{{Cite web|url=https://www.youtube.com/watch?v=OMZogy7rkoI&index=9&list=PLKCk3OyNwIzsMPOS-d73HAuEbDUp9Hglq|title=Direct Access From Database to Network for Much Faster OLTP|last=Loaiza|first=Juan|date=July 19, 2016|website=YouTube|archive-url=|archive-date=|dead-url=|access-date=August 5, 2018}}</ref> reduced messaging overhead in a cluster and ''Smart Fusion Block Transfer''<ref>{{Cite web|url=https://www.youtube.com/watch?v=iBrfdAFVhNk&list=PLKCk3OyNwIzsMPOS-d73HAuEbDUp9Hglq&index=4|title=Non-Blocking Transfers of OLTP Data on Clusters|last=Loaiza|first=Juan|date=July 19, 2016|website=YouTube|archive-url=|archive-date=|dead-url=|access-date=August 5, 2018}}</ref> eliminated log write delays for OLTP applications in a cluster. ''Exadata Cloud at Customer''<ref>{{Cite web|url=https://www.oracle.com/corporate/pressrelease/expands-oracle-cloud-022717.html|title=Oracle Press Release: Oracle Expands Oracle Cloud at Customer Portfolio to Database Workloads with Oracle Exadata Cloud Machine|last=Maloney|first=Nicole|date=February 27, 2017|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=July 27, 2018}}</ref><ref name=":0" /> debuted, enabling Oracle Cloud benefits within corporate data centers.
 
''[https://www.oracle.com/technetwork/database/exadata/exadata-x7-2-ds-3908482.pdf Exadata X7-2]''<ref name=":8" /> and ''[https://www.oracle.com/technetwork/database/exadata/exadata-x7-8-ds-3938980.pdf X7-8]'' were released in 2017<ref>{{Cite web|url=https://video.oracle.com/detail/videos/featured-videos/video/5591322770001|title=Video: Oracle's Next Generation Exadata Database Machine X7|last=Loaiza|first=Juan|date=2017|website=Oracle Corporation|archive-url=|archive-date=|dead-url=|access-date=August 2, 2018}}</ref>. ''Flash capacity doubled''. Flash cards became ''Hot pluggable'' for online replacement. ''10 Terabyte (TB) disk drives'' debuted along with ''25 Gb/sec Ethernet'' connectivity. Oracle Database In-Memory processing was extended into Flash storage, and storage server DRAM was utilized for faster OLTP.
 
==Support Policy==
''[https://www.oracle.com/technetwork/database/exadata/exadata-x7-2-ds-3908482.pdf Exadata X8-2]''<ref name=":29" />and ''X8-8'' were released in April 2019. ''Exadata Storage Server Extended (XT)'' was introduced for low-cost storage of infrequently accessed data.''14 Terabyte (TB) disk drives'' debuted along with ''60% more compute cores'' in Exadata storage servers. Machine Learning algorithms were added to automatically monitor CPU, network and memory to detect anomalies such as stuck processes, memory leaks and flaky networks, and to automatically create (Auto index), rebuild or drop indexes. Optimizer statistics are also gathered in real-time as DML executes. For enhanced security, Advanced Intrusion Detection Environment (AIDE) was added to detect and alert when unknown changes to system software are made.
As the platform has been around since 2008, Oracle has published information related to the end-of-support for older Exadata generations. In Oracle's published document titled ''Oracle Hardware and Systems Support Policies'',<ref>{{Cite web|url=https://www.oracle.com/us/support/library/hardware-systems-support-policies-069182.pdf|title=Oracle Hardware and Systems Support Policies|access-date=March 5, 2021}}</ref> they mention "After five years from last ship date, replacement parts may not be available and/or the response times for sending replacement parts may be delayed." To look up the "last ship date" of a particular Oracle Exadata generation, Oracle published a document titled ''Oracle Exadata - A guide for decision makers''.<ref>{{Cite web |last=Various |title=Oracle Exadata - A guide for decision makers |url=https://www.oracle.com/a/otn/docs/exadata-decision-maker-guide.pdf |access-date=July 11, 2024 |website=oracle.com}}</ref>
 
Each generation of the [[Oracle Zero Data Loss Recovery Appliance]] shares components with similar generations of Exadata.
[https://www.oracle.com/a/ocom/docs/engineered-systems/exadata/exadata-x8m-2-ds.pdf Exadata X8M-2]<ref>{{Cite web|url=https://www.oracle.com/corporate/pressrelease/oow19-oracle-unleashes-worlds-fastest-database-machine-091619.html|title=Oracle Unleashes World's Fastest Database Machine|last=Brown|first=Victoria|date=September 16, 2019|website=oracle.com|url-status=live|archive-url=|archive-date=|access-date=September 19, 2019}}</ref> and [https://www.oracle.com/a/ocom/docs/engineered-systems/exadata/exadata-x8m-8-ds.pdf X8M-8] were released in September, 2019. Substantial performance increases resulted from the addition of [[3D XPoint|Intel Optane DC Persistent Memory]] in Exadata Storage Servers, and a new 100 Gbit/s internal network fabric based on RoCE (RDMA over Converged Ethernet), replacing the previous InfiniBand fabric. These changes increased Read IO throughput by 2.5x and lowered IO latency 10x. In addition. a new KVM based virtual machine replaced the previous Xen based one, doubling the amount of memory available to a guest VM.
 
== References ==
Line 496 ⟶ 267:
==External links==
{{Commons category|Oracle Exadata}}
* The Next Platform: [http://www.nextplatform.com/2016/04/14/oracle-steps-moores-law-rev-exadata-database-machines/ Oracle Steps With Moore’s Law To Rev Exadata Database Machines]
* [http://www.oracle.com/us/products/database/exadata-database-machine/ Oracle Website: Oracle Exadata Database Machine]
{{Oracle}}
*{{Citation | url = https://flashdba.com/history-of-exadata/ |last=Buckel|first=Chris| title = flashdba: History of Exadata | work = Flash DBA|volume=|pages=}}.
*[http://blog.tanelpoder.com/categories/exadata/ Tanel Poder’s Exadata blog, from an expert in performance tuning on Exadata]
* [http://arup.blogspot.com/ Arup Nanda’s blog, where you can search for Exadata-related posts]
*[https://www.accenture.com/us-en/service-infrastructure-services-accenture-enkitec-group The Accenture/Enkitec blog, from an organization specializing in Exadata and sponsoring yearly Exadata-centric conferences]
* How does Oracle Exadata work: [https://www.de.logicalis.com/was-ist-oracle-exadata/]
[[Category:Database management systems]]
[[Category:Data warehousing products]]
Retrieved from "https://en.wikipedia.org/wiki/Oracle_Exadata"