1. The Role of Grid in the IT Landscape
Robert Fogel
Intel Corporation, Principal Education Architect
OGF Vice-President

2. Grid and virtualization work together in the IT datacenter and beyond
Overview
Enterprise IT Pain Points:
Total cost of operation
Operational & energy efficiency
Business alignment and delivery of business value
Virtualization has struck a loud chord
Grid extends the familiar concept of virtualization
Increased utilization of IT assets
Break-down of silos
Federation of data and infrastructure
Grid and virtualization work together in the IT datacenter and beyond 2

3. 3

4. Virtualization Spectrum
Rack/Cluster
& Desktops
Virtual
Enterprise
Processor
Virtual
Ecosystem
Virtual
Datacenter
Platform
(micro)
Virtualization Spectrum
(macro)
VM (today)
“The Grid”
Enterprise Grid (today)
Service-Oriented IT
(Grid + SOA + Web 2.0) 4

5. Virtualization in the Future
Virtualized Compute, Network & Storage
Application and Data Services
SOA
Outsourcers
Firewall
Internet
Comm Service Providers
Digital Government
Mobile Consumers
Digital Homes
Digital Offices (SMB)
Enterprise
Mobile Workforce
Digital Office
Data Center
MOTE
RFID
Factory / Warehouse
Sensors
Business Processes & Workflows
Suppliers
Customers 5

6. Next Generation IT Infrastructure
Grid
Virtualize, Provision, and Aggregate Shared Resource Pools
Communication
Grid Enabled
Storage
Compute
Messaging
Directory
Data
Orchestration
Resources
Security
Web Services
Grid Middleware Services
Utility Computing Services
Optimization
Load
Balancing
Billing
Accounting
Metering
WSRF
Grid Services
Utility Computing
Policy-based SLA/QoS, dynamic resource allocation, and pay-for-use 6

7. The Virtual Datacenter
CPU
Blade/Server
Core … VM
Grid of VMs (or service containers) 7

8. IT Infrastructure Evolution
Evolving Toward The Virtual Datacenter
IT Infrastructure Evolution
Single-core processor systems
Multi-Core processor-based system
Server Consolidation
Shared Landing
Intel Host Platform
Common Application Components
Operating System
App A
App B
App C
App A
App B
App C
Enterprise Grid
Grid Middleware
VMM
App E
App D
App C
App B
App A OS
OS Partitioning 8

9. The Business Value of Virtualization
Server Consolidation
R&D Production OS
App … OS
App
VMM HW HW
VMM HW HW
Multiple App/OS – 10:1 in many cases
Enables rapid deployment
Disaster Recovery
Dynamic Load Balancing OS
App 2 OS
App 3
RICHARD WIRT
In the server environment, virtualization increase utilization by making it possible to host mixed operating systems on a platform. By migrating workloads from dedicated smaller servers to high-reliability enterprise platforms, administrators can increase efficiency, reduce energy costs, maintain high reliability, and reduce administration overhead.
By providing the ability to run development, test and production version of a software stack on the same machine, virtualization can reduce the number of machines need in the development environment. It can also speed deployment because the development and production platforms are guaranteed to be identical, the developer can plug into production workloads, and the promotion from test to production can be done in seconds. The isolation provided by the VMM ensures that test software cannot bring down the production applications.
Disaster recover is aided by virtualization because production software can be encapsulated in a virtual machine and moved as a unit from a failing machine to an alternate platform before the application fails. Or if the hardware does fail, the application can be quickly restored as a single VM, eliminating the sometimes long and laborious process of restore backup files and reconfiguring the OS and applications before the system can be restarted.
Extending the notion of VM migration, dynamic load balancing software monitors the performance of an application running in a VM and when the response falls below as target level the VM is given access to more resources on its current platform, or moved to a more powerful or less utilized platform. As a result, the average utilization of all the machines is higher and more consistent.
App 1
App 4 OS
App … OS OS
VMM
VMM HW HW
VMM
VMM
CPU Usage
CPU Usage HW HW
90%
30%
Upholding high-levels of
business continuity
Balancing utilization with head room
Higher Utilization, Isolation, Lower TCO 9

10. Consolidated Environment
# 2
CPU
App C
OS B
# 3
App D
CPU
OS B
# 1
CPU
App A
OS A
App B
# 5
CPU
App F
OS B
C#4A
CPU
OS B
C#4B
C#4C
App E
Dissimilar configurations
Low System Utilization (<20% on avg.) 10

11. Starting on Virtualization
App E
Grid SW
# 2A
CPU
App C
App D
SW VMM
OS B
# 2
CPU
App C
OS B
# 3
App D
CPU
OS B
# 1
CPU
App A
OS A
App B
# 1
CPU
App A
App B
SW VMM
OS A
OS B
App F
# 5
CPU
App F
OS B
#4A
#4B
#4C
App E
OS B
OS B
OS B
CPU
CPU
CPU
SW OS Partitioning
Grid
Dissimilar configurations
Cluster or silo heterogeneous environment enabled for maximal performance & server utilization 11

12. Better Virtualization
App E
Grid SW
# 2A
VT CPU
App C
App D
VT Enab. VMM
OS B
# 1A
VT CPU
App A
App B
VT Enabled VMM
OS A
OS B
App F
#4B
#4A
CPU
OS B
#4C
CPU
OS B
CPU
SW VMM
CPU
SW VMM
SW OS Partitioning
HW Assisted OS Partitioning
Grid
HW assisted OS Partitioning
Better reliability over SW based OS Partitioning 12

13. Grid + OS Partitioning Enhanced virtualization
App E
App E
Grid SW
Grid SW
# 2A
VT CPU
App C
App D
VT Enab. VMM
OS B
# 4VA
VT CPU
VT Enabled VMM
OS A
OS B
#4B
#4A
CPU
OS B
# 4VB
VT CPU
VT Enabled VMM
OS A
OS B
#4C
CPU
OS B
App A
App B
App F
OS Partitioning
Grid + OS Partitioning
Grid
Enhanced virtualization
Resources available on demand 13

14. Completing Virtualization
App A
App B
App F
App E
Grid SW
Grid SW
# 2A
VT CPU
VT Enab. VMM
OS B
# 4VA
VT CPU
VT Enabled VMM
OS A
OS B
# 4VB
VT CPU
VT Enabled VMM
OS A
OS B
#4C
CPU
OS B
App C
App D
Grid + OS Partitioning
Extending cross enterprise
Resource allocation shifted based on SLA and QoS 14

15. Future Scaling Much simpler scale-out App C App D App A App B App F
App E
Grid SW
Grid SW
# 2A
# 4VA
VT CPU
VT Enabled VMM
OS A
OS B
# 4VB
VT CPU
VT Enabled VMM
OS A
OS B
#4C
CPU
OS B
# N1
VT CPU
VT Enab. VMM
OS B
OS A
OS B
OS B
VT Enab. VMM
VT CPU
Much simpler scale-out 15

16. The Best Combination 80% 90% 65% 30% 15% 1000% Enterprise Grid
Application Modification May Be Required
(many vendors)
No Application Modification Required
App A
App B
App C
Enterprise Grid
Business Grid SW
VMM
App E
App D
App C
App B
App A OS
OS Partitioning
Server Utilization
90%
Operation Costs Reduction
30%
Application Performance
1000%
Server Utilization
80%
Operation Costs Reduction
65%
Application Performance
15%
Source: Intel Customers including Hewitt, Wachovia, CSFB and others 16

17. Summary and Conclusion
Grid and VM technologies are mainstream
Grid and VM work together to provide better optimization and new capabilities of the IT datacenter
Grid setups for extending optimization and new capabilities beyond the datacenter 17