Oracle Exadata is a packaged solution offering from Oracle, configured with bundled hardware, storage and database, which is touted to be optimized for handling scalable data warehouse-type workloads in query and analysis.
2. What is Exadata?
⢠Complete Oracle database server
⢠Disk storage system
⢠CPU, memory, network hardware
⢠Operating system
⢠Database software
⢠Fully installed and configured
3. Exadata Unified Workload Transformation
Single Machine forâŚ
⢠Data Warehousing
⢠OLTP
⢠Database Cloud
OLTP with Analytics and Parallelism of
Warehousing
Warehousing with Interactivity,
Availability, and Security of OLTP
5. Exadata Architecture
Complete Database platform using standard servers for Compute and Storage
1. Scale-Out Database Servers
⢠128 to 160 CPU Cores per rack
⢠2TB to 4TB DRAM
1. Scale-Out Intelligent Storage Servers
⢠2-socket storage servers, Exadata
Storage Software
⢠Up to 500TB disk per rack
⢠22TB Flash memory cards per rack
1. InfiniBand Network
⢠Unified internal connectivity ( 40
Gb/sec )
8. What is standard about Exadata?
⢠Off the shelf hardware
⢠Intel x86 processors
⢠Standard Sun disk drives, memory
⢠Infiniband network adapters and switches
⢠Linux operating system
⢠Oracle 11gR2 database software
9. What is unique about Exadata?
⢠Storage Servers have Oracle database specific features
⢠Smart Scan â returns only needed data
⢠Storage Indexes â only accesses needed disk blocks
⢠Hybrid Columnar Compression â fits data into fewer disk
blocks
11. Exadata Intelligent Storage Grid
⢠Data Intensive processing runs in Exadata
Storage Grid
1. Filter rows and columns as data streams
from disks (168 Intel Cores)
2. Example: How much product X sold last
quarter
⢠Exadata Storage Reads 10TB from disk
⢠Exadata Storage Filters rows by
Product & Date
⢠Sends 100GB of matching data to DB
Servers
3. Scale-out storage parallelizes execution
and removes bottlenecks
12. Exadata Smart Scan
1. Database Servers
⢠Perform complex database
processing such as joins,
aggregation, etc.
2. Exadata Storage Servers
⢠Storage Server is smart storage, not
a DB node
⢠Search tables and indexes, filtering
out data that is not relevant to a
query
⢠Cells serve data to multiple
databases enabling OLTP and
consolidation
⢠Simplicity, and robustness of storage
appliance
13. Exadata Storage Index
Transparent I/O Elimination with No Overhead
⢠Exadata Storage Indexes maintain
summary information about table data in
memory
⢠Store MIN and MAX values of columns
⢠Typically one index entry for every MB of
disk
⢠Eliminates disk I/Os if MIN and MAX can
never match âwhereâ clause of a query
⢠Completely automatic and transparent
15. Exadata Hybrid Columnar Compression
Highest Capacity, Lowest Cost
1. Data is organized and compressed by
column
⢠Dramatically better compression
2. Speed Optimized Query Mode for Data
Warehousing
⢠10X compression typical
⢠Runs faster because of Exadata
offload!
3. Space Optimized Archival Mode for
infrequently accessed data
⢠15X to 50X compression typical
16. Exadata X3 Database In-Memory
1. X3 mass memory hierarchy delivers extreme
Machine
performance
⢠Automatically moves all active data from disk
to memory
2. DRAM memory expanded to 2 or 4 TB for hottest
data
⢠4 to 40 TB of compressed user data
3. Flash memory expanded 4X to 22 TB per rack
⢠40 to 200 TB of compressed user data â ALL
active data
⢠1.5 Million SQL random read I/Os per second
for OLTP
⢠100 GB/sec SQL data scan rate for reporting
and warehouses
18. Comparison of three DB server types
Compare two types of database servers to demonstrate
features unique to Exadata
â DB server attached to a Storage Area Network
(SAN)
â Exadata
⢠With Smart Scan
⢠Without Smart Scan
19. Rows, Columns, and Blocks
⢠A SQL table is a collection of rows â one row per sale
⢠A row is a list of columns â date, product, customer,
amount,âŚ
⢠A table is broken up into equal sized blocks each with
a number of rows = rows size/block size = 8000
bytes/80 bytes = 100 rows per block
⢠Results of queries are a subset of the columns and a
subset of the rows of a table
20. Database Server with Just Disks
DATABASE SERVER
INDEXES
DISKS
BLOCKS
COMPRESSION
MEMORY
CACHE
RESULTS
USER
21. Database Server with SAN
Disk Array (e.g. XP 24000)
Database Server
INDEXES
NETWORK
DISKS
BLOCKS
MEMORY
CACHE
BLOCKS
MEMORY
CACHE
COMPRESSION
RESULTS
USER
22. Exadata with Smart Scan
Cell Storage Server
Database Server
INDEXES
NETWORK
DISKS
BLOCKS
COMPRESSION
MEMORY
CACHE
RESULTS
USER
23. Exadata without Smart Scan
Cell Storage Server
Database Server
INDEXES
NETWORK
DISKS
BLOCKS
MEMORY
CACHE
BLOCKS
MEMORY
CACHE
COMPRESSION
RESULTS
USER
24. Observations
â˘
DB server with SAN differs from full Exadata system
â Full database blocks copied over SAN network While Results are
only copied over Exadata network
â Database server caches blocks from SAN
â˘
Exadata Smart Scan can be turned off
â Without Smart Scan Exadata works just like DB server with SAN
25. Why use Exadata?
â˘
â˘
â˘
â˘
New set of performance enhancing features
All other Oracle features still available
Easy to use new features
Easy to bypass new features
31. Scale to 18 Racks by Just Adding Cables
Full Bandwidth and Redundancy
Scale to more than 18 Racks by adding InfiniBand switches
32. Seamless Upgrades and Expansions
1. A single Database Machine can have
servers from different generations
2. Databases and Clusters can span across
multiple hardware generations
3. New software runs on older hardware
38. Things to watch out for
⢠Databases tuned for Exadata are not portable
â Datafiles with HCC wonât work on non-Exadata
system
â Applications that have been tuned to work well on
Exadata will be slow on non-Exadata
⢠Canât control, no documentation for storage indexes
â What columns are they on?
â What column types can they be on?
39. Things to watch out for â page 2
⢠HCC and tables > 255 columns buggy
â Wrong results, errors
â Wrong optimizer stats
⢠Newness issues
â Bugs, hangs, crashes, unexpected results
â Lack of documentation
â Lack of trained people
One Stop Shop â everything from Oracle â Like IBM mainframe of oldPreconfigured hardware optimized for Oracle DB.Oracle Exadata is a packaged solution offering from Oracle, configured with bundled hardware, storage and database, which is touted to be optimized for handling scalable data warehouse-type workloads in query and analysis.
Oracle has repeatedly claimed that the Exadata X2 system has significantly improved performance and scalability system-wide, resulting in scalability up to petabytes and beyond. There are actually two Oracle Exadata Database Machine products: Exadata X2-2 and Exadata X2-8. The X2-8 is a full-rack only system, primarily intended for large OLTP and consolidation environments. Both platforms consist of Oracle Database 11g Release 2, Oracle RAC (Real Application Clusters) Database server grid, an InfiniBand interconnect, the Oracle Enterprise Linux operating system, and the Exadata Storage Server Grid using either high-performance (600GB) or high capacity (3TB) disk storage.
Hundreds of engineer years spent optimizing and hardening the system end-to-end â Frees I/T talent to focus on business needsď§ Standard platform improves support experienceď§ Runs all existing Oracle Database workloadsď§ Building block of the Oracle CloudCan save from multiple vendor involvements
Each cell server takes up 3 rows on the rack. It has three rows of four disks each. Is a Linux server with bunch of disks. One small green light per disk, one large green light per server. 7 servers in this picture â top of rack.High-performance storage server built from industry standard components ⢠12 disks - 600 GB 15000 RPM High Performance SAS or 3TB 7200 RPM High Capacity SAS ⢠2 Six-Core Intel Xeon Processors (E5-2630L) ⢠Dual ported 40 Gb/sec InfiniBand ⢠4 x 400 GB Flash Cards⢠Intelligent Exadata Storage Server Software
Can buy hardware, OS, database separately â theoretically could reconfigure as standard Oracle database server(s) with RAC
storage servers = cell servers â disk I/O subsystem like SAN. These are the three main PERFORMANCE features
Exadata Smart Scan processes queries at the storage layer, returning only relevant rows and columns to the database server. As a result, much less data travels over fast 40GB InfiniBand interconnects--dramatically improving the performance and concurrency of simple and complex queriesExadata storage servers also run more complex operations in storage â Join filtering â Incremental backup filtering â I/O prioritization â Storage Indexing â Database level security â Offloaded scans on encrypted data â Data Mining Model Scoring ď§ 10x reduction in data sent to DB servers is common
A storage index is an in-memory structure that holds some information about the data inside specified regions of physical storage. This information tells the storage cell which areas of the disk do not contain the values the query is interested in, so those areas are not accessed during a scan.Storage indexes reside in the memory of the storage serversâalso called storage cellsâand significantly reduce unnecessary I/O by excluding irrelevant database blocks in the storage cells. Physical disk is divided into exadata cells normally 1MB in size. On every 1MB storage index is applied.
The biggest hump for the EHCC feature is its own comfort zone i.e. database with less transactions and low concurrency. On a database which does frequent transactions and reads the data, the feature stands defeated.
Caches Write I/Os in PCI flash in addition to Read I/Osď§ Transparently accelerates write intensive workloads â 20X more write IOPS than disk on X3 â 10X more write IOPs than disk on V2 and X2 ď§ Persistent write cache speeds database recovery ď§ Exadata Flash Cache is much more effective than flash tiering architectures used by others â Caches current hot data, not yesterdayâs â Caches data in granules 8x to 16x smaller than tieringď§ Greatly improves the effectiveness of flash
Demonstrate what is different about Exadata and what is not by comparing it to two other types of Oracle database servers. Compare/contrast.
blocks read from disk to memory and copied across SAN have unneeded rows and columns. Results can be small subset of entire table or full copy of table.
Similarity between this and cell server. In this case the indexes and compression operate at the database level. Red is bad â entire blocks copied. Green is good â small result set. Indexes reduce disk blocks read from disk. Compression fits more rows per block.
here the SAN looks like a cell server but there are no indexes, compression and the full blocks are copied over the SAN. SAN cache is like DB cache â caches blocks
Similar to the san in that the disks and db server are separate but the database like processing is in the cell. But then what about the DB server? Hash and sort areas would be good to show. Also, can mix both methods. Talk here about increasing hash and sort area size for OBIA but leaving buffer cache small. Explain PGA and SGA? Results go to PGA which is where hashing and sorting caching is.
Just like with SAN.
key point is blocks copied over fiber network for SAN versus results over infiniband for Exa. Last part leads to how to turn on off Smart scan. key point of smart scan slides.DB server with disks looks like Exadata Cell serverIndexesCompressionmemory cachereturns small subset of full data
New tools for our tool bagASM (auto storage management)Storage Indexes
Command line utility to manage Exadata cells (CLI)Oracle Integrated Lights Out Manager (ILOM)Oracle Integrated Lights Out Manager (ILOM) is the service processor embedded on all Oracle's SPARC Enterprise T-series and Sun Fire x86 servers, including all rack mounts and blades. Oracle ILOM enables full out-of-band management, providing a âJust like being thereâ remote management capability
Oracle Enterprise Manager Cloud Control 12c uses a holistic approach to manage the Exadata Database Machine and provides comprehensive lifecycle management from monitoring to management and ongoing maintenance for the entire engineered system.Enterprise Manager is a web-based graphical interface that enables you to perform management operations on the database like startup and shutdown, object creation, and schema management.
Start Small and Grow 1. Eighth Rack2. Quarter Rack3. Half Rack4. Full RackUnique Architecture Makes it Fastest at the Lowest Cost
ExadataShared everything = resource contentionParallelism is conditional and unpredictableScalability with diminishing returnsTeradataShared nothing = no resource contentionParallelism is unconditional and predictableLinear scalability with a slope of one
Unlike Teradata Database, which performsall query processing within the nodes,Exadata cells only perform initial columnprojections from the select list, and rowrestrictions using the WHERE clausepredicates. Exadata can also perform somefact table row restrictions based on joineddimensions (for star schema joins). TheExadata cells transmit the resulting rowsets to the Oracle database server, where allother query operations are executed, justas they have always been.
Flash memory (storage server cache) slowNormal RAM â 1000x faster than diskCell server cache (Flash) â 10x faster than disk