Orran Krieger: VMware vCloud Deck

Cloud Computing Orran Krieger Senior Staff Engineer, VMware, Inc. Lead Kvetcher VMware Cloud team Adjunct Professor CMU 1 Wednesday, April 15, 2009

Outline • Background on VMware • Cloud computing today • What should it be • The vCloud initiative • Research ideas 2 Wednesday, April 15, 2009

Evolution of Virtualization PHYSICAL 3 Wednesday, April 15, 2009

Evolution of Virtualization VIRTUALIZED PHYSICAL 3 Wednesday, April 15, 2009

Evolution of Virtualization POOLED VIRTUALIZED PHYSICAL HW HW HW HW HW HW HW HW HW HW HW HW HW HW HW HW HW HW Logical Resource Pooling (RP) Distributed Resource Scheduler (DRS) 3 Wednesday, April 15, 2009

Evolution of Virtualization POOLED R R P P 1 2 HW HW HW HW HW HW HW HW HW HW HW HW HW HW HW HW HW HW Logical Resource Pooling (RP) Distributed Resource Scheduler (DRS) 3 Wednesday, April 15, 2009

Power Savings with DPM • 5 hour VMmark run • 110 VMs, 4 server cluster 2000 1800 1600 1400 Watts 1200 1000 800 600 400 200 0 4 Wednesday, April 15, 2009

Power Savings with DPM • 5 hour VMmark run • 110 VMs, 4 server cluster 2000 No DPM 1800 1600 1400 Watts 1200 1000 800 600 400 200 0 4 Wednesday, April 15, 2009

Power Savings with DPM • 5 hour VMmark run • 110 VMs, 4 server cluster 2000 No DPM 1800 1600 1400 Watts 1200 1000 800 600 400 With DPM 200 0 50% Savings during VMmark 4 Wednesday, April 15, 2009

Distributed Virtual Switch core distribution access hosts 5 Wednesday, April 15, 2009

Distributed Virtual Switch • Network-centric view: core virtualization blurs the host/network boundary distribution access hosts VMs 5 Wednesday, April 15, 2009

Distributed Virtual Switch • Network-centric view: core virtualization blurs the host/network boundary distribution • Virtual network switches and topologies within a host access • VMs can move anywhere • The virtual network switch is a new access layer. hosts VMs 5 Wednesday, April 15, 2009

Distributed Virtual Switch • Network-centric view: core virtualization blurs the host/network boundary distribution • Virtual network switches and topologies within a host access • VMs can move anywhere • The virtual network switch is a new access layer. • Network administrators can hosts no longer rely on physical port-based access control VMs 5 Wednesday, April 15, 2009

Distributed Virtual Switch • Network-centric view: core virtualization blurs the host/network boundary distribution • Virtual network switches and topologies within a host access • VMs can move anywhere • The virtual network switch is a new access layer. • Network administrators can hosts no longer rely on physical port-based access control Distributed Virtual Switch VMs Distributed virtual switch to extend network access control and management to virtual environments 5 Wednesday, April 15, 2009

Availability Hot-plug resources 6 Wednesday, April 15, 2009

Availability Add/remove capacity on demand Hot-plug resources Improve application availability 6 Wednesday, April 15, 2009

Site Recovery Manager Production Recovery VMware Infrastructure VMware Infrastructure Site Recovery Manager transforms disaster recovery 7 Wednesday, April 15, 2009

Security: VMware VMsafe API that enables protection of VMs by inspection of virtual components in conjunction with hypervisor Isolation of protection engine from malware Application Broad ranging coverage of virtual machine CPU, memory, storage Operating System Protection Engine VMware Infrastructure 8 Wednesday, April 15, 2009

vApp: A new type of Software Policies • Properties • Policy-based operations 1. Product: eCommerce 2. Topology • Multi-tier 3. Resources Req: CPU, Mem, • Distributed as an OVF Disk,Bandwidth package 4. Only port 80 is used 5. DR RPO: 1 hour • Built by: 6. VRM: Encrypt w/ SHA-1 7. Decommission in 2 month • ISVs / Virtual Appliance Vendors Websphere Fire • By internal IT shops Exchange Wall Tomcat • By IT administrators • SI/VARs SAP 9 Wednesday, April 15, 2009

Desktops Centralized Virtual Desktop Infrastructure (VDI) Virtual Desktops VMware Infrastructure 3 10 Wednesday, April 15, 2009

Desktops Centralized Virtual Desktop Infrastructure (VDI) Virtual Desktops • Full desktops run as VMs in the data center • Connect securely to desktop from anywhere • Centralized management • More efficient resource usage VMware Infrastructure 3 • Higher availability • Online and offline mode VMware Virtual Desktop Manager Thick or Thin Clients 10 Wednesday, April 15, 2009

Evolution of Virtualization Separate Consolidate Aggregate Automate Liberate CapEx OpEx Savings Business No Physical Savings Agility Boundaries Automation Automation Virtual Virtual Virtual Infrastructure Infrastructure Infrastructure Management Management Management Management Hypervisor Hypervisor Hypervisor Hypervisor Hypervisor Cloud Scale Test and Capacity Server Self-Managing Computing Development On Demand Consolidation Datacenter 11 Wednesday, April 15, 2009

The cloud started with SaaS/Web Traditional software model Test Develop Install Release Operate Configure SaaS model Test Develop Operate 12 Wednesday, April 15, 2009

The reality: success is very hard Test Operate Develop 13 Wednesday, April 15, 2009

The reality: success is very hard Test Operate Develop • Most SW companies don’t have the in-house skill to operate at scale: • e.g. loadbalancer, network config, security, disaster recovery, …& 70% of investment spent not enhancing application 13 Wednesday, April 15, 2009

The reality: success is very hard Operate Test Develop • Most SW companies don’t have the in-house skill to operate at scale: • e.g. loadbalancer, network config, security, disaster recovery, …& 70% of investment spent not enhancing application 13 Wednesday, April 15, 2009

The reality: success is very hard Operate Test Develop • Most SW companies don’t have the in-house skill to operate at scale: • e.g. loadbalancer, network config, security, disaster recovery, …& 70% of investment spent not enhancing application • Enormous investment required in application level to scale. 13 Wednesday, April 15, 2009

The reality: success is very hard Operate Develop Test • Most SW companies don’t have the in-house skill to operate at scale: • e.g. loadbalancer, network config, security, disaster recovery, …& 70% of investment spent not enhancing application • Enormous investment required in application level to scale. 13 Wednesday, April 15, 2009

The reality: success is very hard Operate Develop Test • Most SW companies don’t have the in-house skill to operate at scale: • e.g. loadbalancer, network config, security, disaster recovery, …& 70% of investment spent not enhancing application • Enormous investment required in application level to scale. • So, successful SaaS vendors started building re-usable platforms… 13 Wednesday, April 15, 2009

Cloud offerings • Software as a Service (SaaS): • Application hosted in the cloud. • e.g., gmail, google apps, salesforce.com, wikipedia… • Platform as a Service (PaaS): • Application development environment and runtime hosted in cloud. • e.g., engine yard, Google App Engine, Force.com • Infrastructure as a Service (IaaS): • Developers/administrators obtain general computing, storage, messaging services… • e.g., terramark, Amazon AWS, Mosso 14 Wednesday, April 15, 2009

Example Amazon IaaS 15 Wednesday, April 15, 2009

Example • Animoto April 2008: Peak EC2 instances: • Mon 50, Tues 400, Wed 900, Friday 3400 16 Wednesday, April 15, 2009

The problems with this evolution • Vertical offered by a single vendor • Just a few hundred/thousand developers enhancing offering • No on-premise offering for enterprise/university/SaaS vendor • Who wants to trust a single company? • Limited largely to web applications: • Limited support legacy, HPC, hosted client, grid, • Disintermediates OEMs, infrastructure vendors, traditional management stacks… (VARs for SaaS) • Ignores the key value in the virtualization needed for general purpose workloads: over provisioning, SRM, DRS, DPM, OVF, SVI … 17 Wednesday, April 15, 2009

Disclaimer… 18 Wednesday, April 15, 2009

Disclaimer… • At this point, there is the mandatory… okay there is a lot of hype here, but… 18 Wednesday, April 15, 2009

Disclaimer… • At this point, there is the mandatory… okay there is a lot of hype here, but… • The transformation is more profound that is yet understood, cloud done right will: • be used for all applications, • change how we deploy and develop applications, • enable new OSes, new programming models, new servers, new storage solutions • enable new markets for computer services • …, in other words, its gonna change everything 18 Wednesday, April 15, 2009

What do we really want 19 Wednesday, April 15, 2009

What do we really want Original vision of Utility/grid computing: 19 Wednesday, April 15, 2009

What do we really want Original vision of Utility/grid computing: ”If computers of the kind I have advocated become the computers of the future, then computing may someday be organized as a public utility just as the telephone system is a public utility... The computer utility could become the basis of a new and important industry.” 19 Wednesday, April 15, 2009

What do we really want Original vision of Utility/grid computing: ”If computers of the kind I have advocated become the computers of the future, then computing may someday be organized as a public utility just as the telephone system is a public utility... The computer utility could become the basis of a new and important industry.” John McCarthy, MIT Centennial in 1961 19 Wednesday, April 15, 2009

What do we really want Original vision of Utility/grid computing: ”If computers of the kind I have advocated become the computers of the future, then computing may someday be organized as a public utility just as the telephone system is a public utility... The computer utility could become the basis of a new and important industry.” John McCarthy, MIT Centennial in 1961 Virtualization converts computation into a fungible commodity 19 Wednesday, April 15, 2009

Why would this be transformative • Nicholas Carr: • “As with the factory-owned generators that dominated electricity production a century ago, today's private IT plants will be supplanted by large-scale, centralized utilities.” • Gets rid of key impediments to innovation: • Virtual appliance model for distributing installing applications. • Avoids need for broad HCL, OS support, … • Availability of massive capacity on demand. • Enables long-tail in SW 20 Wednesday, April 15, 2009

VMware’s vCloud initiative • Provide SW so that anyone can easily create cloud: e.g., SP, University, Enterprise • Supports all applications • Controlled programmatically: • From VMs, from capacity owner, from portal • OVF based API • Can scale up to massive data centers. • Can scale down to small department. • Can be accessed from VI Client • New end user interface provide simple self service experience. • Enable broad partner and research collaboration: • Researchers can replace any part of the service. • Researchers can replace the entire implementation and clone the API 21 Wednesday, April 15, 2009

VMware’s vCloud initiative • Provide SW so that anyone can easily create cloud: e.g., SP, University, Enterprise • Supports all applications • Controlled programmatically: • From VMs, from capacity owner, from portal • OVF based API • Can scale up to massive data centers. vCloud Software • Can scale down to small department. • Can be accessed from VI Client • New end user interface provide simple self service experience. • Enable broad partner and research collaboration: • Researchers can replace any part of the service. • Researchers can replace the entire implementation and clone the API 21 Wednesday, April 15, 2009

VMware’s vCloud initiative • Provide SW so that anyone can easily create cloud: e.g., SP, University, Enterprise • Supports all applications • Controlled programmatically: • From VMs, from capacity owner, from portal vCloud API • OVF based API • Can scale up to massive data centers. vCloud Software • Can scale down to small department. • Can be accessed from VI Client • New end user interface provide simple self service experience. • Enable broad partner and research collaboration: • Researchers can replace any part of the service. • Researchers can replace the entire implementation and clone the API 21 Wednesday, April 15, 2009

VMware’s vCloud initiative • Provide SW so that anyone can easily create cloud: e.g., SP, University, Enterprise VI • Supports all applications Client • Controlled programmatically: • From VMs, from capacity owner, from portal vCloud API • OVF based API • Can scale up to massive data centers. vCloud Software • Can scale down to small department. • Can be accessed from VI Client • New end user interface provide simple self service experience. • Enable broad partner and research collaboration: • Researchers can replace any part of the service. • Researchers can replace the entire implementation and clone the API 21 Wednesday, April 15, 2009

VMware’s vCloud initiative • Provide SW so that anyone can easily create cloud: e.g., SP, University, Enterprise VI • Supports all applications Self-service UI Client • Controlled programmatically: • From VMs, from capacity owner, from portal vCloud API • OVF based API • Can scale up to massive data centers. vCloud Software • Can scale down to small department. • Can be accessed from VI Client • New end user interface provide simple self service experience. • Enable broad partner and research collaboration: • Researchers can replace any part of the service. • Researchers can replace the entire implementation and clone the API 21 Wednesday, April 15, 2009

VMware’s vCloud initiative • Provide SW so that anyone can easily create cloud: e.g., SP, University, Enterprise VI • Supports all applications Self-service UI Client • Controlled programmatically: • From VMs, from capacity owner, from portal vCloud API • OVF based API • Can scale up to massive data centers. Alternative Implementation vCloud Software • Can scale down to small department. • Can be accessed from VI Client • New end user interface provide simple self service experience. • Enable broad partner and research collaboration: • Researchers can replace any part of the service. • Researchers can replace the entire implementation and clone the API 21 Wednesday, April 15, 2009

Key abstractions vCloud • Organization with users/ roles Organization 1 • Cloud to admin on his own VDC 1 VDC 2 terms: vApp vApp • data centers • L2 networks “VDCnet” “foo” • Persistent VMs (fenced) ND “bar” • SLA on vApp “VDCnet” (isolated) ND • overprovisioning “Private” • ... “Public” • OVF based REST API: • multi-tiered applications Mapped as network • serialized using OVF adapters into installation • configuration via OVF 22 Wednesday, April 15, 2009

Key abstractions vCloud UI & API Organization 1 Organization 2 VDC 1 VDC 1 VDC 2 VDC 2 vApp vApp vApp vApp “VDCnet” “VDCnet” “foo” “foo” (fenced) (fenced) ND ND “bar” “bar” “VDCnet” “VDCnet” (isolated) (isolated) ND ND “Private” “Private” “Public” “Public” Same color means same network 23 Wednesday, April 15, 2009

It should be easy, VMware already has • Rich service provider, OEM, ISV partner ecosystem • API for controlling virtualization that has become a defacto standard • Ability to deal with large numbers of hosts as a pool of resources • Support for backup, DR, resource management, power management, … • Rich community of users • Rich user interface loved by administrators •… 24 Wednesday, April 15, 2009

Not so fast… • Problems with our existing technology: • Core abstractions exposed physical as well as virtual • Scaling up to 10000 hosts & 100K VMs fundamentally different • Security more of a issue in multi-tenancy environment • Challenges with our approach: • Need to scale down • Need to enable partner ecosystem: • No one circumscribed approach. • Need to release SW to others to install, configure… • Need to enable others to innovate. 25 Wednesday, April 15, 2009

Difference between small & large scale 26 Wednesday, April 15, 2009

Scale up versus down • Large scale service: Service Oriented Architecture • Each service totally independent with own DB, own set of machines… all communication through interface: • Advantages • Fails independently & can identify failures • Own reliability based on requirements of data • Can evolve services independently • Can scale them independently • Small scale: shrink wrapped software • Single general purpose piece of software • Advantages: • Minimize overhead • Simplified installation 27 Wednesday, April 15, 2009

Modular design Large Scale Modularity also enables extensibility 28 Wednesday, April 15, 2009

Modular design Large Scale Small Scale Modularity also enables extensibility 28 Wednesday, April 15, 2009

Modular design Small Scale Modularity also enables extensibility 28 Wednesday, April 15, 2009

Key Technologies REST API : HTTP based resource oriented interface; All the characteristics of the WWW • Extensible without breaking client. • Client only has to know about what it cares about. • Can route, proxy, cache Spring: Standard component framework • Injects dependencies and wires together Spring beans • Forces programmer into maintainable design pattern; isolates dependencies OSGI: Standard dynamic module framework • Global registry of interfaces to instances • Dynamically load, unload, start, stop bundle Other technology • JMS publish/subscribe messaging bus isolates end points • Hibernate simplifies DB code & DB independence 29 Wednesday, April 15, 2009

Architecture/Implementation vCloud Self vCloud API client VMRC Transfer Client Service UI Loadbalancer Blaze DS Rest dispatcher vCloud Task vRP VC IP Pool VMRC Network Image Catalg Cloner OSGI Network Mgr Mgr Control Mgr Proxy srvc xfer Netwrk srvc Mgmt srvc vCloud Application Cell Network Storage Sched Authnt Security VC Invent Directory Handler Mgr Tasks Mgr Mgr User/ Message Bus Model DB Inventory DB Account Billing Mgmt Substrate Inventory VC Validator Diff Mgmt Listener Substrate ESX hosts Networking Running VM storage Offline storage 30 Wednesday, April 15, 2009

Architecture/Implementation vCloud Self vCloud API client VMRC Transfer Client Service UI Loadbalancer Blaze DS Rest dispatcher vCloud Task vRP VC IP Pool VMRC Network Image Catalg Cloner OSGI Network Mgr Mgr Control Mgr Proxy srvc xfer Netwrk srvc Mgmt srvc vCloud Application Cell Network Storage Sched Authnt Security VC Invent Directory Handler Mgr Tasks Mgr Mgr User/ Message Bus Model DB Inventory DB Account Billing Mgmt Substrate Inventory VC Inventory Validator Diff VC Listener Validator Diff Mgmt Listener ESX hosts Networking Running VM storage Offline Substrate ESX hosts Networking Running VM storage Offline storage storage 30 Wednesday, April 15, 2009

Architecture/Implementation vCloud Self vCloud API client VMRC Transfer Client Service UI Loadbalancer Blaze DS DS Rest dispatcher Blaze Rest dispatcher vCloud Task vRPvRP VC VC IP Pool VMRC VMRC Network Image Task IP Pool Catalg Cloner Network Image Network MgrMgr MgrMgr Control Catalg Cloner MgrMgr OSGI Proxy Network srvcsrvc xfer xfer Netwrk Control Proxy srvcNetwrk srvc srvcsrvc vCloud Mgmt vCloud Application Application Cell Network Storage Sched Authnt Security Network Storage Sched Authnt Security MgrMgr VC Invent Directory Handler VC Invent Mgr Tasks MgrMgr Directory Handler Tasks Mgr User/ Message Bus Model DB Inventory DB Account Billing Mgmt Substrate Inventory VC Inventory Validator Diff VC Listener Validator Diff Mgmt Listener ESX hosts Networking Running VM storage Offline Substrate ESX hosts Networking Running VM storage Offline storage storage 30 Wednesday, April 15, 2009

Architecture/Implementation vCloud Self vCloud API client VMRC Transfer Client Service UI Loadbalancer Blaze DS DS Rest dispatcher Blaze Rest dispatcher vCloud Task vRPvRP VC VC IP Pool VMRC VMRC Network Image Task IP Pool Catalg Cloner Network Image Network MgrMgr MgrMgr Control Catalg Cloner MgrMgr OSGI Proxy Network srvcsrvc xfer xfer Netwrk Control Proxy srvcNetwrk srvc srvcsrvc vCloud Mgmt vCloud Application Application Cell Network Storage Sched Authnt Security Network Storage Sched Authnt Security MgrMgr VC Invent Directory Handler VC Invent Mgr Tasks MgrMgr Directory Handler Tasks Mgr User/ Message Bus Model DB Inventory DB Account Billing Mgmt Substrate Inventory VC Validator Diff Mgmt Listener Substrate ESX hosts Networking Running VM storage Offline storage 30 Wednesday, April 15, 2009

Architecture/Implementation vCloud Self vCloud API client VMRC Transfer Client Service UI Loadbalancer Blaze DS Rest dispatcher vCloud Task vRP VC IP Pool VMRC Network Image Catalg Cloner OSGI Network Mgr Mgr Control Mgr Proxy srvc xfer Netwrk srvc Mgmt srvc vCloud Application Cell Network Storage Sched Authnt VC Invent Directory Handler Mgr Tasks Mgr User/ Message Bus Model DB Inventory DB Account Billing Mgmt Substrate Inventory VC Validator Diff Mgmt Listener Substrate ESX hosts Networking Running VM storage Offline storage 30 Wednesday, April 15, 2009

Research directions • New operating systems: • e.g., for Java, security services, clustering services, HPC/ multi-core, ... • library OS, multi-core, accelerator to OS, sharing • Fungible computing: • performance, cost to user, cost to SP • Grid/HPC/batch scheduling, e.g., snowflock • Federation: • long tail for data de-duplication, encryption/security for data, trading floor/futures market, vmotion to user • What changes when your desktop is in the cloud? • Mom’s sysadmin, Fusion on steroids, disaster recovery... 31 Wednesday, April 15, 2009

Concluding reports • Cloud computing is going to be transformative to our industry • VMware building a platform so that anyone can play • There are a wealth of research opportunities • We will be providing the SW to universities this year • This is just the start... 32 Wednesday, April 15, 2009

http://cloudscaling.com	261
http://neotactics.com	80
http://www.slideshare.net	47
http://www.cloudscaling.com	43
http://www.vmworld.com	2
http://current.com	2
http://www.lmodules.com	1
http://web.archive.org	1

Orran Krieger: VMware vCloud Deck

by Randy Bias on Apr 26, 2009

Accessibility

Categories

Tags

Upload Details

Usage Rights

8 Embeds 437

Statistics

Orran Krieger: VMware vCloud Deck — Presentation Transcript