Download Trumping Oracle Exadata

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Microsoft SQL Server wikipedia , lookup

Relational model wikipedia , lookup

Extensible Storage Engine wikipedia , lookup

Database wikipedia , lookup

Microsoft Jet Database Engine wikipedia , lookup

Database model wikipedia , lookup

Serializability wikipedia , lookup

Clusterpoint wikipedia , lookup

ContactPoint wikipedia , lookup

Oracle Database wikipedia , lookup

Concurrency control wikipedia , lookup

Transcript
Trumping Oracle Exadata
A short story
Introduction
One of the world’s most prominent technology
companies recently evaluated Oracle’s Exadata
against two Fusion Powered servers.
Sometimes two can be bigger than Exa. 
Initial Situation
• Prominent technology company needs fast
and reliable infrastructure to service tens to
hundreds of thousands of requests per minute
• Architecture based on a pod-like design with
many pods aggregated for scale
• Transaction-driven application requires
disproportionate amount of storage resources
and disk-optimizing acrobatics
Stage 1 – Single Pod Design
• Two Sun T5440 servers
• 2x NetApp 6080s
• 1X Performance
Acceleration Modules
• 12 disk shelves
• Output
– 75,000 IOPS
• Result
– Insufficient performance
Stage 2 – Single Pod Design
• Two Sun T5440 servers
• 2x NetApp 6280s
• 2X Performance
Acceleration Modules
• 12 disk shelves
– 4 shelves with SSDs
• Output
– 208,000 IOPS
• Result
– Insufficient performance
Stage 3 – Single Pod Design
• Exadata (½ rack)
– Bundled servers
– Bundled SSDs
• Output
– 9,000 transactions per
second
• Result
– Insufficient performance
– VERY EXPENSIVE
Stage 4 – Final Single Pod Design
• Two HP DL 580 G7 servers
• 10 ioDrives per server
– 1.28TB each, 12.8TB total
• Output
– Over 39,000 transactions
per second from single
server
– 2nd server for failover
• Result
– Best-in-class performance
– Best-in-class price
High Availability – no problem
• Oracle Dataguard
– Already paid for with site license
– RAC not required
– Data Guard configuration consisting of a primary database and a
physical standby database. From the primary database, redo is
being transmitted and applied to the standby database. Log
apply services apply the redo out of the standby redo log files to
the standby database.
Source: Oracle.com
Final Comparison
NetApp
Disk Farms
Exadata
(½ rack)
Transactions per
(IOPS based) 9,000
second
Two
Fusionpowered
servers
Improvement
39,000+
4.3X
Rack space
Four 42U
(servers +
NTAP)
21U
8U
2.6X
Cost
Pricey
~$800,000
~$400,000
2X
Dollars per
transaction
$89
$10
9X
Transactions per
rack unit
429
4875
11.3X