Professional Documents
Culture Documents
Data Warehousing
OLTP
Database Consolidation
Exadata Evolution
Three generations of experience
2008
2009
2010
HP Oracle Database
Machine (V1)
Data Warehousing
Data Warehousing
OLTP
Database Consolidation
Data Warehousing
OLTP
Database Consolidation
Performance improved
17x with no changes to
our application
Jim Duffy, BNP Paribas
Quarter
Half
Full, Multi-Rack
Full-Rack
Multi-Rack
Storage Grid
14x 2-socket storage servers
(2U)
5 TB PCI Flash
Fully
Oracle Linux + Exadata SW
Redundant
Delivered ready-to-run
Tested
Highly supportable
No unique configuration issues
Identical to config used by Oracle Engineering
Deploy in Days,
Not Months
Software by Oracle
X2-8 or X2-2
Full Rack
X2-2
Half Rack
X2-2
Quarter Rack
100 TB
50 TB
21.6 TB
336 TB
168 TB
72 TB
5.3 TB
2.6 TB
1.1 TB
45 TB
22.5 TB
9.5 TB
150 TB
75 TB
31.5 TB
30 TB
15 TB
6.5 TB
100 TB
50 TB
21.5 TB
1- Capacity calculated using normal space terminology of 1 TB = 1024 * 1024 * 1024 * 1024 bytes.
2 - Actual space available for a database after mirroring (ASM normal redundancy) and allowing one disk (Quarter and
Half) or two disks (Full Rack) of free space to automatically remirror after disk failures.
3 - Actual space available for the database computed after triple mirroring (ASM high redundancy).
X2-2
Half Rack
X2-2
Quarter
Disk Data
Bandwidth1,3
25 GB/s
12.5 GB/s
5.4 GB/s
14 GB/s
7 GB/s
3 GB/s
Flash Cache
Data Bandwidth1,3
75 GB/s
37.5 GB/s
16 GB/s
64 GB/s
32 GB/s
13.5 GB/s
50,000
25,000
10,800
25,000
12,500
5,400
1,500,000
750,000
375,000
12 TB/hr
6 TB/hr
3 TB/hr
Disk IOPS
Flash IOPS2,3
Data Load Rate4
1 - Bandwidth is peak physical disk scan bandwidth achieved running SQL, assuming no compression. Effective
data bandwidth will be much higher when compression is factored in.
2 - IOPS Based on read IO requests of size 8K running SQL. Note that the IO size greatly effects flash IOPS.
Others quote IOPS based on 2K, 4K or smaller IOs that are not relevant for databases. Exadata Flash read IOPS
are so high they are typically limited by database server CPU, not IO.
3- Actual Performance varies by application.
4 Exadata load rates are typically limited by database server CPU, not IO. Rates vary based on load method,
indexes, data types, compression, and partitioning
Multi Rack
8+ Racks
96 TB Disk
1.5 TB Flash
4 Storage Servers
48 CPU cores
216 TB Disk
3.4 TB Flash
9 Storage Servers
108 CPU cores
432 TB Disk
InfiniBand
6.75 TB Flash
Connected
18 Storage Servers
216 CPU cores
Query
- Column Projection
- Decompression
- Row Filtering
Benefits Multiply
Prerequisites - SAP
SAP Technologies
- SAP NetWeaver 7.x (x>=0) and products, industry specific
solutions including enhancement packages based on this
NetWeaver stack
Examples: SAP ERP (ECC) 6.0, SAP BW 7.x,SAP CRM 2005 /
2007 / 7.x, SAP PLM 6.0 / 7.0x, SAP SRM 2005 / 2007 / 7.x,
SAP SCM 2005 / 2007 / 7.x ...
SAP Deployments
- Only 3-Tier installations
SAP Instances cannot run on database nodes
- SAP Central Services may run on database nodes
SAP Instances must run on separate server(s)
Exalogic with Oracle Linux or any combination of
hardware/operating systems for SAP Instances and Exadata
(Linux/Solaris) allowed
Prerequisites Oracle
Oracle Exadata Database Machine X2-2 or X2-8
Oracle GI, RDBMS and Exadata Storage Software
- Oracle Grid Infrastructure 11.2.0.2
- Oracle RDBMS 11.2.0.2
Oracle Exadata Database Bundle Patch 6 (BP6)
SAP Database Bundle Patch (June 10th) for Exadata
Goals
Expose customer's business-critical real SAP workload to an Exadata
environment (X2-2 Half Rack + Application servers on x86 running Solaris)
Measure performance increase (throughput & response times)
Show how Oracle technology can help to minimize the impact of hard/software failures for the business
Rules/Conditions
No lab tuning, no DB tweaking
Use standard Exadata database setup without optimization
6.5 TB DB migration from Oracle 10.2 (SPARC) via Oracle 11.2 (x86/Linux)
into an Exadata ASM/RAC environment
all improvements should be as well achievable by customer's staff for their
productive systems
Runtimes:
11h 30min on customer's current environment
3h 45min on Exadata.
Two active nodes of a Half-Rack Exadata had only been utilized
to 25%.
Throughput:
one node RAC: 7,500 users with 4s thinktime:
=> ~1s response time and 64% CPU utilization (one node)
four node RAC: 6,000 users with 4s thinktime:
=> ~0.9s response time and 17% CPU utilization (all nodes)
Applications
Enterprise Manager
WebLogic
Server
Coherence
Tuxedo
Solaris
Installed on premises
EL X2-2
Extreme Performance
Certified
Enterprise Java,
Fusion Middleware 11g
Linux or Solaris
Applications
Flash
Accelerated
Fully
Redundant
Compute Nodes
EL X2-2
Scale on Demand
Start small and grow
Field upgradable to
larger
configurations
Delivered and
installed by Oracle
8 Compute
Nodes
16 Compute
Nodes
30 Compute
Nodes
Spine Switch
10GbE
Standard
Oracle
Database
Compute Nodes
Exabus
Data Center
Service Network
(10GbE)
Exadata
Exalogic
FMW + Application
Execution
Storage
Management Switch
Data Center
Mgmt Network
(GbE)
Oracle Homes
WLS Domains
Logs
Binaries
Patches
User Files
GbE
WebLogic
WebLogic
WebLogic
WebLogic
WebLogic
GridLink
GridLink
GridLink
GridLink
GridLink
GridLink
XA
80%
20%
Continuous Connections
Even with RAC Changes
Exalogic Performance
Advantage
5X or more is
typical
performance
gain for many
application
types:
450%
400%
350%
25%
minimum
gain for
standard
Linux or
Solaris
applications
300%
250%
200%
150%
100%
50%
0%
+25% +50%
Exalogic ROI
2X
3X
5X
Java EE and
Fusion
Middleware
Oracle
Applications
OLTP with
Oracle Exadata
Extreme Performance