WIXLIB

Architecting and Implementing aFederated Cloud

Cloud Computing

I am here to helpbuzzetti@us.ibm.com

Worldwide Centers4Poughkeepsie, NYMontpellier, FranceMakuhari, JapanBoeblingen, Germany

What we do

How We Are Like Our Clients

Provide Service to Lines of Business

Must Support Production like Services

Make Effective Use of Resources

Flaws in the Current System

Discourages Sharing

Not Efficient

Limits Experimentation

Stifles Innovation

It's this guys fault15

Vision

How is it different from Hybrid Cloud ?

Architecting

Goals

Worldwide

Leverage Assets

Avoid Duplication of Efforts

Standardized Services

ITIL Aligned

High Level Requirements

Service Requester Requirements

Self Service

Service Catalog

View all services offered

Provider Requirements

Site Managed Catalog

Site Managed Security

Cookie Cutter Deployment

Federation Requirements

Articles of Federation

Consistent Identity

Consistent Operations

Transparency Between Clouds

Implementing

Federated CloudFederatedCloudRequest oningServicesServices SelectedBOEVirtualizedInfrastructurefor tualizedInfrastructure

IBM Cloud Computing RACloud ServiceConsumerCloud Service ProviderCloud ServiceDeveloperCloud ServicesIT capability provided to Cloud Service Consumer(Virtualized) Infrastructure – Server, Storage, Network, FacilitiesInfrastructure for hosting Cloud Services and Common Cloud Management PlatformPartner CloudsCommon Cloud Management PlatformBSS – Business Support ServicesBusiness-level functionality for management of Cloud ServicesConsumerIn-house ITOSS – Operational Support ServicesOperational-level functionality for management of Cloud ServicesSecurity & ResiliencyServiceDevelopmentTools

ProductsTechnologyHighly ThreadedThroughput-orientedScale Out CapableLower Quality of ServiceTechnologyCompute intensiveHigh I/O BandwidthHigh Memory BandwidthFloating pointScale out CapableTivoliMonitoringTSAM v7.2 Automate process ofinstantiating andmanaging adistributed ITenvironment.TechnologyScaleHigh Transaction RatesHigh Quality of ServiceHandle Peak WorkloadsResiliency and SecurityServiceAutomationMgrTechnologyScaleHigh Quality of ServiceLarge Memory FootprintResponsive rvisor(KVM,(vSphere)VMware, rtsHMCBillingReportsWorkflowsVM ControlNIMSystem p / SUNHypervisor(PowerVM)StorageNetworkVM PartitionCisco UCSServiceAutomationTemplatesVM Partition0. 2 be Wec af r et nI r es UUCSManagerVM PArtitionVMx86VM Partition Virtualized resources Virtualized aggregation Physical infrastructureBusinessApplicationsTivoli Process Automation EngineOrchestration workflowsVMVirtualizedInfrastructure LayerServiceRequest MgrTransactionProcessingand DatabaseHMCSystem zHypervisor(zVM)StorageNetworkEnd to End Service ManagementTivoli ServiceAutomationLayerAnalytics andHigh PerformanceComputingVM Partition Service measurement Service reporting Usage accounting Auditing and controlsWeb, Collaborationand InfrastructureVM PartitionWorkloads

Middleware StackTSAM voli Process Automation EngineOrchestration workflows0. 2 be Wec af r et nI r es LibraryWorkflows

Phase One

Focus on Multiple Platforms

Service Provider InfrastructureTSAM1System p222System z3System zSystem z33POKSystem pSystem pMOPBOE

Phase Two

Focus on Cross GeographyPoughkeepsie, NYMontpellier, FranceMakuhari, JapanBoeblingen, Germany

Service Provider InfrastructureTSAM111System p2System z2System zSystem z33POKSystem pSystem p23TSAMTSAMMOPBOE

Two Service ebWeb2.02.0TSAMFederated ServiceCatalogFederated ServiceCatalogFederated ServiceCatalogLocal ServiceCatalogLocal ServiceCatalogLocal ServiceCatalogService AService BService C

Phase Three

Focus on Sharing Services

Service Provider InfrastructureTSAM111System p2System z2System zSystem z33POKSystem pSystem p23TSAMTSAMMOPBOE

Two Service ebWeb2.02.0TSAMFederated ServiceCatalogFederated ServiceCatalogFederated ServiceCatalogLocal ServiceCatalogLocal ServiceCatalogLocal ServiceCatalogService AService BService CService AService BService CService AService BService C

Future

Questions

Backup Charts

Service Provider Infrastructure89az/VM Linux49ITUAMTUAMz/LinuxTDI359bITMz/LinuxTDI96VM Host

Cloud Computing08/03/102Cloud Computing is a new way to deploy IT services.This session assumes the audience is familiar withcloud concepts. If they are not, spending a few minutesgoing over the NIST definition would be a good idea. http://csrc.nist.gov/groups/SNS/cloud-computing/

I am here to helpbuzzetti@us.ibm.com3This is me. I am here to help. I include this chart sothat people can have my email.I work with customers. I am an IT specialist and myrole is to help customers understand newtechnologies by doing proof of concepts, whiteboarding sessions and design workshops.

Worldwide Centers4Poughkeepsie, NYMontpellier, FranceMakuhari, JapanBoeblingen, Germany08/03/104The different places where I work. This slide is toshow that the Design Center has a world widepresence, and that we can help customers in almost allgeographies. The Federated cloud is getting support from the geosas well as other. 4

What we do08/03/105The client centers do a wide variety of work. We needsystems and software to be able to do our jobs, andlike many of our customers we are broken up into siloswhich can make us less efficient

How We Are Like Our Clients08/03/106In many ways we are like our clients. We have similargoals and needs.

Provide Service to Lines of Business08/03/107Each unit has to provide service to its line of business.For instance, the benchmarking centers need to beable to show customers that we can run their workload, and do it in such a way that it helps closebusiness.The same is true for the Briefing Centers and theDesign Centers.

Must Support Production like Services08/03/108Image 1325035/sizes/l/ We have to be production like. We might not loose amillion dollar an hour like our customers, but it is verycostly to us. Many times we are doing work with thecustomer before a sale is completed and any outagemight prevent the sale from going smoothly.

Make Effective Use of Resources08/03/109We do not have an infinite budget. We are not sales.We must do the best with what we have.

The current system has a number of faults thatFederated Cloud aims to fix.These flaws you will probably see in your own shop.

Discourages Sharing08/03/1011Each group busy and maintains its own hardware.Since they pay for it, getting resources from outsideyour groups is difficult.Each group busy – each group is busy

Not Efficient08/03/1012Benchmarks eat up a lot of resource, but they are notrun every single day of the year, in every geo. Thereare a lot of idle cycles that could be used better.

Limits Experimentation08/03/1013We can't do sand boxing to test new software andtechnologies as easily as we want to.

Stifles Innovation08/03/1014Lots of time is spent build and rebuilding systems. Wecould do a lot more innovative stuff if we shifted thoseresources.

15

Vision08/03/1016The end state. A cloud of clouds supporting a numberof groups, with resources from all over the world.

How is it different from Hybrid Cloud ?08/03/1017I get this question a lot. Basically, Hyrbid cloud is whenyou are using resources from two or more clouds of adifferent type.Federated cloud is when you are sharing the same setof services across different siloed resources.

Architecting08/03/1018Phase one of this project when designing how it wouldwork. And the first step is.

Defining the goals of the project. Theseare very high level, and they are userbased of the flaws in the currentsystem.

Worldwide08/03/1020Federated cloud needs to have support from all theGeos

Leverage Assets08/03/1021Photos fromhttp://www.flickr.com/photos/alishav/3259542We have to leverage assets amongstthe centers effectively.

Avoid Duplication of Efforts08/03/1022We have to avoid duplication of effort.If a group in France builds a really coolservice that will build WAS on zLinux, Ishould be able to use that same serveron their resources or my own.

Standardized Services08/03/1023We have to have standard servicesthat are supported in every geo. Thingslike basic IaaS.

ITIL Aligned08/03/1024We need to ensure that we have proper alignment withindustry best practices such as ITIL. After all we areproviding services, and we have to make sure that wehave proper service level agreements and life cycles.

High Level Requirements08/03/1025Image from google mapshttp://maps.google.com/maps?hl en&q ariel poughkeepsie&iHey, that is where I work!These goals mimic that of NIST but are called out herefor clarity.

Service Requester Requirements08/03/1026The requirements are broken down by the role thatwould need them.

Self Service08/03/1027One has to be able to obtain the usage of a servicewithout intervention from the service provider.

Service Catalog08/03/1028There has to be a catalog that is federated from allmembers that an end user can have access to.

View all services offered08/03/1029End users have to be able to view every service that isoffered. This helps them avoid asking for a duplicateservice.

Provider Requirements08/03/1030These are the requirements that a provider or memberof the federation is interested it.

Site Managed Catalog08/03/1031A site should be able to manage there own servicecatalog and pick and choose what services they wishto offer up in to the federation.

Site Managed Security08/03/1032Each site should be able to manage the security oftheir environment as if they were not in a federatedcloud.Each site may have different network restrictions forinstance, and we must be able to support that.

Cookie Cutter Deployment08/03/1033If you request a service be deployed in MOP, it shouldlook similar if not identical to that in POK

Federation Requirements08/03/10The rules that make this a federation.34

Articles of Federation08/03/1035Image fromhttp://www.flickr.com/photos/caveman 92223/2898686447/sizeThis is still a work in progress but this is the documentthat represents all the bylaws of the federation. Itstates what is supported and what is not as well ashow to join and leave the federation.

Consistent Identity08/03/10Same user profile as you go from site to site.36

Consistent Operations08/03/1037Things like monitoring and chargeback should beconsistent across all members of the federation.

Transparency Between Clouds08/03/1038It should not matter what cloud your server is going tobe instantiated in. To the end user they should all bethe same.

Implementing08/03/1039How did we get from the goals and requirements to aworking system ?

Federated CloudFederatedCloudRequest oningServicesServices SelectedBOE08/03/10VirtualizedInfrastructurefor edInfrastruc tureFirst we drew a pretty Infrastructure40

IBM Cloud Computing RACloud ServiceConsumerCloud Service ProviderCloud ServiceDeveloperCloud ServicesIT capability provided to Cloud Service Consumer(Virtualized) Infrastructure – Server, Storage, Network, FacilitiesInfrastructure for hosting Cloud Services and Common Cloud Management PlatformPartner CloudsCommon Cloud Management PlatformBSS – Business Support ServicesServiceDevelopmentToolsBusiness-level functionality for management of Cloud ServicesConsumerIn-house ITOSS – Operational Support ServicesOperational-level functionality for management of Cloud Services08/03/10Security & Resiliency41We looked at what was around at IBM and where therewas already a large amount of work done in the cloudspace. We are trying to reduce duplication here.In this diagram we chose to focus Federated Cloudaround the OSS level.