Grid vs. Cloud Computing · Cloud Computing: IT on Demand Capacity Utilization Most servers have utilization rates of just 5-20%; the other 80-95% is wasted capacity.1 Cloud provisioning
Post on 08-Jun-2020
0 Views
Preview:
Transcript
© 2008 Parabon Inc. All rights reserved. | 1© 2009 Parabon Computation, Inc. All rights reserved. |
Grid vs. CloudComputing
The similarities and differencesbetween Cloud Computing andExtreme-Scale Computation on Demand®
© 2009 Parabon Computation, Inc. All rights reserved.
Virtualized Data Center
Virtualization bringsnew capabilities to thedata center.
Capacity Utilization
© 2009 Parabon Computation, Inc. All rights reserved.
Virtualized Data Center
Virtualization bringsnew capabilities to thedata center.
Grid computing andcloud computing aretwo types of servicesthat benefit fromvirtualization.
Capacity Utilization
© 2009 Parabon Computation, Inc. All rights reserved.
Cloud Computing: IT on Demand
Capacity Utilization
Cloud computing enablesself-service provisioning ofvirtual machines.
© 2009 Parabon Computation, Inc. All rights reserved.
Cloud Computing: IT on Demand
Capacity Utilization
Cloud computing enablesself-service provisioning ofvirtual machines.
© 2009 Parabon Computation, Inc. All rights reserved.
Cloud Computing: IT on Demand
Capacity Utilization
Cloud computing enablesself-service provisioning ofvirtual machines.
Goal: “Simplify deployment ofOperating Systems and appservers.”
© 2009 Parabon Computation, Inc. All rights reserved.
Cloud Computing: IT on Demand
Capacity Utilization
Cloud computing enablesself-service provisioning ofvirtual machines.
Goal: “Simplify deployment ofOperating Systems and appservers.”
Characteristics: Smallnumbers of VM allocationsheld for long periods of time(days/months).
© 2009 Parabon Computation, Inc. All rights reserved.
Cloud Computing: IT on Demand
Capacity Utilization
Cloud computing enablesself-service provisioning ofvirtual machines.
Goal: “Simplify deployment ofOperating Systems and appservers.”
Characteristics: Smallnumbers of VM allocationsheld for long periods of time(days/months).
© 2009 Parabon Computation, Inc. All rights reserved.
Cloud Computing: IT on Demand
Capacity Utilization
Most servers have utilizationrates of just 5-20%; the other80-95% is wasted capacity.1
Cloud provisioning tools andvirtualization greatly improveutilization rates. Even so, mostdata centers have massiveamounts of capacity that goesunused.
[1] Taurus - A Taxonomy of the Actual Utilizationof Real UNIX and Windows Servers, David GHeap, Principal IT Consultant, IBM EnterpriseServer Group, January 2003. http://www-03.ibm.com/servers/library/pdf/taurus.pdf
UNUSED
© 2009 Parabon Computation, Inc. All rights reserved.
Grid: Extreme-Scale Computation on Demand®
Capacity Utilization
Grid computing solutionsenable parallel processingof computational tasks, oftenusing idle vs. dedicatedcapacity.
© 2009 Parabon Computation, Inc. All rights reserved.
Grid: Extreme-Scale Computation on Demand®
Capacity Utilization
Grid computing solutionsenable parallel processingof computational tasks, oftenusing idle vs. dedicatedcapacity.
Goals: “Accelerate throughputfrom decades to days orfrom months to minutes.”
© 2009 Parabon Computation, Inc. All rights reserved.
Grid: Extreme-Scale Computation on Demand®
Capacity Utilization
Grid computing solutionsenable parallel processingof computational tasks, oftenusing idle vs. dedicatedcapacity.
Goals: “Accelerate throughputfrom decades to days orfrom months to minutes.”
“Enable deep computationsthat are otherwise intractable.”
© 2009 Parabon Computation, Inc. All rights reserved.
Grid: Extreme-Scale Computation on Demand®
Capacity Utilization
Grid computing solutionsenable parallel processingof computational tasks, oftenusing idle vs. dedicatedcapacity.
Goals: “Accelerate throughputfrom decades to days orfrom months to minutes.”
“Enable deep computationsthat are otherwise intractable.”
Characteristics: Largenumbers of work requestsrun for short periods of time(minutes / hours).
© 2009 Parabon Computation, Inc. All rights reserved.
Frontier® Grid Services
Capacity Utilization
The Frontier® Grid Platformcan share resources with cloudsolutions to provide a high-performance compute serviceusing otherwise wastedcapacity.
© 2009 Parabon Computation, Inc. All rights reserved.
Frontier® Grid Services
Capacity Utilization
The Frontier® Grid Platformcan share resources with cloudsolutions to provide a high-performance compute serviceusing otherwise wastedcapacity.
The Frontier Grid Servermanages execution of large-scale computational jobsacross hundreds or thousandsof compute nodes.
© 2009 Parabon Computation, Inc. All rights reserved.
Frontier® Grid Services
Capacity Utilization
The Frontier Compute Enginecan be deployed as a small,unobtrusive virtual applianceacross the entire data center.
© 2009 Parabon Computation, Inc. All rights reserved.
Frontier® Grid Services
Capacity Utilization
The Frontier Compute Enginecan be deployed as a small,unobtrusive virtual applianceacross the entire data center.
Frontier uses resources onlywhen its host has excesscapacity. And, it is designed to“back off” when higher priorityprocesses are present.
© 2009 Parabon Computation, Inc. All rights reserved.
Frontier® Grid Services
Capacity Utilization
Computational Work Units
Next, we’ll watch how Frontierexecutes a large-scale job inthe presence of other VMallocations.
© 2009 Parabon Computation, Inc. All rights reserved.
Frontier® Grid Services
Capacity Utilization
Computational Work Units
A large job comprised of manytasks is launched against theFrontier Grid Service.
© 2009 Parabon Computation, Inc. All rights reserved.
Frontier® Grid Services
Capacity Utilization
Computational Work Units
The Frontier Grid Serverassigns tasks to computeengines on the grid.
© 2009 Parabon Computation, Inc. All rights reserved.
Frontier® Grid Services
Capacity Utilization
Computational Work Units
The Frontier Grid Serverassigns tasks to computeengines on the grid.
© 2009 Parabon Computation, Inc. All rights reserved.
Frontier® Grid Services
Capacity Utilization
Computational Work Units
Frontier Compute Engineson hosts with excess capacitybegin executing tasks in turn,and results are returned backto the server.
© 2009 Parabon Computation, Inc. All rights reserved.
Frontier® Grid Services
Capacity Utilization
Computational Work Units
Capacity utilization in the datacenter increases and the jobfinishes quickly thanks to massparallelization.
© 2009 Parabon Computation, Inc. All rights reserved.
Frontier® Grid Services
Capacity Utilization
Computational Work Units
Capacity utilization in the datacenter increases and the jobfinishes quickly thanks to massparallelization.
© 2009 Parabon Computation, Inc. All rights reserved.
Frontier® Grid Services
Capacity Utilization
Computational Work Units
Because Frontier runs at lowpriority, other operations, suchas cloud provisioning of newVMs, can proceed as normal.
© 2009 Parabon Computation, Inc. All rights reserved.
Frontier® Grid Services
Capacity Utilization
Computational Work Units
Because Frontier runs at lowpriority, other operations, suchas cloud provisioning of newVMs, can proceed as normal.
© 2009 Parabon Computation, Inc. All rights reserved.
Frontier® Grid Services
Capacity Utilization
Computational Work Units
As the job completes, capacityis freed for use by other jobs.
© 2009 Parabon Computation, Inc. All rights reserved.
Frontier® Grid Services
Capacity Utilization
Computational Work Units
Our job, which would takeweeks or months on a singlecomputer, finishes in minutesor hours.
© 2009 Parabon Computation, Inc. All rights reserved.
Conclusion: Grids ≈ Clouds
Grid and Cloud computing share some common goals:
Deliver computation as a utility Virtualize use of computing resources Eliminate need for dedicated computing resources Dramatically improve computational price-performance
But, the services they provide are fundamentally different:
Cloud – Information Technology (IT) on Demand Grid – Extreme-Scale Computation on Demand®
top related