View
1
Download
0
Category
Preview:
Citation preview
© 2008 Hewlett-Packard Development Company, L.P.The inf ormation contained herein is subject to change without notice.
Massive Scale-Out and the Emergence of Cloud Computing
Alanna Dwyer
Scalable Compute & Infrastructure
223 September 2008
Agenda
• Background & evolution
• Cloud-enabling the data center
• Cloud services from HP
• Research activity
323 September 2008
Agenda
• Background & evolution
• Cloud-enabling the data center
• Cloud services from HP
• Research activity
423 September 2008
Evolution to the Cloud:The First Generation
reach
19701960 1980 20001990 2010
t
2005 2020
TheInternet
connectivity Killer App: Mail
523 September 2008
The Second Generation
reach
19701960 1980 20001990 2010
t
2005 2020
TheInternet
The
Web
connectivity
information &
e-commerce
Killer App: Mail
Killer App: Search, e-
commerce
623 September 2008
The Third Generationreach
1970 1980 20001990 2010time
2005 2020
The
Internet
The
Web
The
Cloud
connectivity
information &
e-commerce
virtualized
services
Killer App: Mail
Killer App:
Search, e-
commerce
723 September 2008
IT as a Service, Delivered by the Cloud
7 21 August 2008 HP Restricted | © 2008 Hewlett-Packard Development Company, L.P.
Search
ProductivityApps
Social
Networking Infrastructure
on Demand
Backup
Media sharing Business Apps
Management Apps
Mobile Services
Location-BasedServices
Storage onDemand
Platform
on Demand
823 September 2008
And ...
• At massive scale
− Millions of users
• With unprecedented flexibility
− Mash-ups, aggregation, enhancing services, flexing up and down, ...
• Offering evolving APIs to exploit and extend
• At breakthrough cost levels
− Economies of scale
− New revenue models
− Eliminating old sources of cost
923 September 2008
Barriers to Enterprise Adoption
• Trust in the service vendor− Service levels
− Stability
− Security
• ISV support not widespread
• Few have taken the plunge in a big way
• Customizability of service offerings for specific needs of each enterprise
• Concerns about lock-in, lack of multi-vendor options
• Data: privacy, locality
• Regulatory concerns
• Challenge of migrating from in-house (or outsourced) apps
9 21 August 2008 HP Restricted | © 2008 Hewlett-Packard Development Company, L.P.
1023 September 2008
Cloud: Hype versus reality?
• Economics are a compelling driver!
− Look what we see in HPC
• Commodity technology + access + standards
• Volume enablers
− Virtualization – one physical platform, multiple applications/users
− Internet/web: global scope
− Affordable for individual and SMB
• Risk versus reward
− Security and trust – a matter of time?
1123 September 2008
Agenda
• Background & evolution
• Cloud-enabling the data center
• Cloud services from HP
• Future
1223 September 2008
Scale out and the new IT
CommodityProprietary
1,000s of nodes100s of nodes
SW resiliencyHW resiliency
Multi-siteSingle site
Revenue-CenterCost-Center
Power costsFacility costs
Replicated storageShared storage
DynamicStatic
Grids/CloudClusters
Multi-coreSingle core
Scale-OutTraditional IT
1323 September 200813
Challenges of scale-out and cloud enablement
… all fitting into your current budget
Resource adaptability
Power & cooling
Capacity and capability
Provide rapid response to service demands
Scale and density pushing physical limits and costs
Scale will push limits of manageability and QoS
1423 September 2008
• ‘EKA’, the world’s 8th fastest supercomputer, delivering 175 Tflops peak performance, used for:
− Advanced modeling and simulation in computational sciences
− Cutting edge R&D in HPC, Petascale design
− Cloud computing research
• Based on highly scalable, high efficiency HP core technologies
− HP Cluster Platform 3000BL
• 1822 HP ProLiant BL460c server blades
• 140 terabytes storage, HP SFS parallel file system
− HP Smart Cooling data center services
14 23 September 2008
EKA in CRL’s Rajiv Gandhi InfoTech Park HPC Facility, Hinjewadi, Pune
“Computational Research Laboratories (CRL), a wholly-owned company of the Tata Group, India’s largest
conglomerate, used HP blades, cluster OS and technical assistance to … help CRL design and deploy
EKA’s supercomputing infrastructure for cloud computing research with Yahoo. This is the world's largest
system used for research in cloud computing.”
S. Ramadorai, Chairman of CRL
1523 September 2008
Adaptive IT Infrastructure for Cloud Computing – some customer requirements
• Fast and flexible access to computing resources, enabled by virtualization
• Self-service web site
• Multi-vendor, multi OS
• Policy-driven Task Automation
• Single point of management and user access
• Billing and chargeback for use
• Platform for sharing IT services across the enterprise (multiple data centers)
• Reduce space & environment requirements
• Manage power and cooling
• Audit and verification
• Security, quality of service
1623 September 2008 For HP and Channel Partner Use Only
Platform as a Service
• Create virtual private resource pools
• Pre-defined infrastructure templates (includes HW + software and application stacks)
− Defined by administrators
− Customizable by users
• Architecture-independent
− Support bare-metal and virtualization based provisioning
• Secure web-portal
• Integrated with accounting
1723 September 2008 For HP and Channel Partner Use Only
Monitor / Configure“HP Site Scope, etc.”
Monitor / Configure“HP SIM”
Ne
two
rk
Prov isioning Manager“HP BSA”
StorageN
etw
ork
Request New Resources
Workload & Capacity Management
Adaptive Env. Orchestration
Access Portal End UsersMonitor, Diagnose,
Manage
Admin
Allocation Management
Monitor / Configure“HP OVO, Supermon”
Dynam
ically
Provi
sion
Manager
Review Service Levels, Capacity
Planning ,etc.
Report Current State & Provision Resources
Compute Resources
Update Allocation, Usage, Charges
Monitor, Maintain
Report State/Key E
ventsMaintain Usage
Credits & Charges
Provisioning, Updates, etc.
Intelligent Orchestration
“Adaptive Computing Portal”
Schedule Service Levels
Applications….
MiddlewareWeb Tools, DB, ….
Operating SystemSLES, RHEL, Windows
Storage
Interfaces
Information Monitoring Adaptation Tools
Cluster Manager + HP
Tools”
Platform as a Service: Instantly Create Project and Test Environments
1823 September 2008 For HP and Channel Partner Use Only
HP Business Service Automation (BSA)Automate management of client, server, software, network, storage & process
Help desk/ change requests
ITSM
Monitoring events &
alerts
BSM
Service Automation Reporter
Universal CMDB
Operations Orchestration
IT Teams IT Systems
• Ser vers & network devices
• Software
• Process-enabled
storage
provisioning
• Audit
• Remediati on
• Reporting
• All applications &
components
• Ser ver, network, storage
• Applications
• Configurations
• Patching
• Software
distribution
• Discover ser vers, software networ k
devices , &
storage
• Interdependencies
• Configurations
DiscoveryProvisioning &
Scaling
Business Service Automation
Server Automation
NetworkAutomation
Service Automation Visualizer
Client Automation
Storage Automation
Live Network
1923 September 2008
Defense Information Systems Agency (DISA)
19
Objective Approach Results
• Rapid access to computing resources
• Eliminate the need to procure physical infrastructure
• Self service portal through a single, secure interface
• User configurable server environments
• Automated provisioning
• Flexible billing options
• Meet DoD security
requirements
• HP C&I development of shared
services utility for Rapid Access
Computing Environment (RACE)
• HP Server Automation and HP
Operations Orchestration for
provisioning and configuration management
• HP Operations Manager for
monitoring and control
• HP Service Manager to automate incident & problem management
• HP Systems Insight Manager and
HP ProLiant Essentials
• Cluster Resources Moab for intelligent orchestration and Gold
for billing
Business outcomes
• Reduced costs
• Consolidated simplified processes
• Shortened time to delivery
IT improvements
• Flexible development platforms for Web, Application or Database
• User can allocate own resources
through Web interface
• Can provision a server in a few minutes
• CPU, memory, storage, virtual
environment provided in one simple solution
US Department of DefenseCloud Computing Infrastructure
23 September 2008
2023 September 2008
Holistic Solutions for the IT Power & Cooling Chain
Optimizing fromchip to chiller
Dynamic Smart Cooling: up to 45% cooling cost savings w/Mapping
HP Services: Thermal Mapping over 10% cooling cost reduction
Storage Thin Provisioning / Dynamic Capacity Mgt saves up to 45%
Virtualization/Consolidation: up to 40% reduction in power cost for data centers
Insight Power Manager & iLO 2: 10% power reduction w/regulator
Power Distribution Rack & MCS15% - 20% savings on power & cooling
BladeSystem & Thermal Logic: 25% cost savings to power & cool
Power Optimized ProLiant Servers: 18% less power
SFF Drives: 2.5” 9 watts vs 18watts for 3.5”
Power Supplies: 90% efficient supplies
Low Power processors: up to half the power consumption
Energy Saving Solutions from the Server Chip to the Data Center Air
Chillers and everything in-between
Up to 40% power
savings
2123 September 2008
Comprehensive HP Services for Next-Generation Data Center
Manage, Automate and protect DC operations
Rationalize, Modernize and
Migrate your applications
Consolidate and Virtualize
Design, consolidate and/or modernize
aging facilities
2223 September 2008
Recent additions to HP’s Scale-Out portfolio
ProLiant BL2x220c
• Double compute density
• Infrastructure costs in half
• Maximum power & cooling
efficiency
Performance Optimized Datacenter (POD)
• Industry-standard Flexibility
• Best-in-class Density
• Fast & easy deployment
• Energy Effectiveness
StorageWorks ExDS9100
• Extreme scalability
• Unified manageability
• Extreme affordability
AnnouncedMay 28, 2008
AnnouncedJuly 16, 2008
AnnouncedMay 6, 2008
2323 September 2008
Agenda
• Background & evolution
• Cloud-enabling the data center
• Cloud services from HP
• Future
2423 September 2008
HP Flexible Computing Services
Customers choose how to plug in
Infrastructure Provisioning Service
IPS
IPS + Scheduling, GRID, Tools
IPS+
Application Provisi oning Service
APS
Basic infrastructure service. HP provides & manages hardware infrastructure and OS.
Customer provides & manages applications.
Same as IPS, but HP installs & manages
scheduling, compiler, and/ or grid software.
Includes IPS+, and HP installs [to a default
configuration] application software
CPU, Storage, OS, Network, Maintenance
• A pay-per-use public computing utility, charged on a “cpu-hour” basis
• HP-owned assets housed in HP’s highly secure data centers, dedicated per customer for the time reserved
• Choice of standards-based technologies and operating systems
• Access to industry leading applications and middleware on a utility basis
2523 September 2008
ExchangeExchange SAPSAPCompute Intensive
(FCS)Compute Intensive
(FCS)
Automated delivery processes
Pre-built infrastructure delivered as a service
Virtualization AutomationSecurityManagementPower & Cooling
IT Systems& Services
Built on Adaptive Infrastructure Innovations
Quicker time to realize benefits of a next generation infrastructure
Extending HP FCS - A new approach to outsourcing
HP Adaptive Infrastructure as a Service
2623 September 2008
HP Upline: Storage-as-a-Service
2723 September 2008
Snapfish
• Number One online photo service
− @ 50 million users
− 2 billion photos stored online
2823 September 2008
Agenda
• Background & evolution
• Cloud-enabling the data center
• Cloud services from HP
• Research activities
2923 September 2008
Open Cloud Computing Research Testbed Driving Innovation to Enable the Grid
• A loose federation of “Centers of Excellence” around the globe
− UIUC, Singapore IDA, KIT: 3 initial CoE
− HP, Intel, Yahoo: 3 initial sponsors with CoE
− One common research theme
− Specific research germane to each region
• Sponsors
− HP Labs, Intel Research, Yahoo!
• Each center: 1000-4000 cores and up to PB storage
− With experimental cloud management services
3023 September 2008
Open Cloud Computing Research Testbed What is it?
• An open, scalable, secure, large scale global testbed for cloud computing research
− Collaborative research focused on data center management & cloud services
− Scalable, multi-continent, multi-datacenter, cloud computing system
− A broad base research community including HP, Intel, Yahoo, and other partners and academic researchers
3123 September 2008
Cloud Computing Research @ HPLFulfill the “Everything as a service” vision
• Cells as a Service: Service Cells are secure “containers” for virtual infrastructure elements. Each Cell can contain an arbitrary assembly of virtual machines, virtual storage volumes and virtual networks
• Exascale Data Center: Aims to radically redesign the future data center and its computing components, delivering significantly improved customer experience at dramatically lower costs
• Scalable Storage: A versatile automatic enterprise storage service that can scale, across geographically dispersed locations
• Service Lifecycle Management: A management system for converting any type of application into a service
• Sustainable Data Center: Focus on developing sustainable IT approaches that are more energy efficient and will reduce the environmental impact
3223 September 2008
Cells as a ServiceAutomated Infrastructure Lab
• An infrastructure-level Cloud service
• Delivering secure, isolated virtual infrastructures – Cells– to multiple customers
• Offering enterprise-grade properties
32 21 August 2008 HP Restricted | © 2008 Hewlett-Packard Development Company, L.P.
3323 September 200833 21 August 2008
Service Lifecycle Management (SLiM)Automated Infrastructure Lab
• Configuring, deploying and managing cloud services
• Models and model transforms for the complete service lifecycle
• From the ‘kit of possible parts’ through to the deployed service
• Multi-viewpoint, modular, parametric models
• Managing and predicting non-functional properties
• Lifecycle is round-trip, and includes service adaptation, service upgrade, etc.
• Delivered as a service (of course!)
Service Ma na gement Tools
M odel Repository
Cloud Infra structure
Model Information Flow
HP Restricted | © 2008 Hewlett-Packard Development Company, L.P.
3423 September 200834
• Enable “service marketplaces”where businesses can obtain services from a global ecosystem of service providers, rapidly, efficiently, and with high quality of service
• Research contributions− Business process specification,
service selection, instantiation, choreography and adaptation over independently created service components
− Service-oriented and model-based architectures, combined with Web 2.0, social networking, and semantic web mechanisms
MercadoMercado
Service Automation and Integration Lab
21 August 2008 HP Restricted | © 2008 Hewlett-Packard Development Company, L.P.
3523 September 2008
Other R&D for scale out capabilities
• Converged fabrics and virtual networks
• Multi-core programming tools
• Power management and workload allocation
• Dense computing alternatives
3623 September 2008
So…Is the Cloud Ready for the Enterprise?
• Yes, for some applications
− Crash and burn, dev/test, batch processing, peak-load capacity expansion for some tasks
− Some applications making in-roads
− LOTS of interest
• Not for the core mission critical stuff ... Yet
36 21 August 2008 HP Restricted | © 2008 Hewlett-Packard Development Company, L.P.
3723 September 2008
Visit our website
www.hp.com/go/enablethecloud
Recommended