CLOUD COMPUTING > – It's about the data Dr. Jim Baty Distinguished Engineer Chief Architect, VP / CTO Global Sales & Services, Sun Microsystems CClloouudd CCoommppuuttiinngg –– –– iitt''ss aabboouutt nnootthhiinngg nneeww –– iitt cchhaannggeess eevveerryytthhiinngg
New York Times, June 14, 2006 Jan. 22, 2009 Sun Microsystems 2009 2 Cloud Computing – “... data is mostly resident on servers 'somewhere on the Internet' and the application runs on both the 'cloud servers' and the
user's browser.” Eric Schmidt
The Information Factories • George Gilder • Wired 14.10 2006 • The desktop is dead. Welcome to the Internet cloud...
Jan. 22, 2009 Sun Microsystems 2009 3 chmod 777 s
e – petascale-content, tribal-knowledge: driving future infrastructure c i v e
d Web 4.0
/ The Intelligent Web s t n
e Web 3.0 g
a The Semantic Web
A.I.
(read-write-execute)
e.g. Cycorp Semantics
Web 2.0
e.g. Metaweb The Social Web Tagging s
n (read-write) e.g. Technorati Linguistics a e.g. Attensity m
u Web 1.0 Statistics h
The World Wide Web
t e.g. Google
s (read-only) u j just 'data' social content / execution
Web 1-4.0 ala Nova Spivak Internet pop. – 2000=360M, 2008=1.2B, 2010=2B they won't use PCs, they will be 'programmers'
Jan. 22, 2009 Sun Microsystems 2009 4 'Pay by the Drink'(N. Carr, the Big Switch) – But how are you going to measure, control, pay?
As in .... app • SaaS – SW as a service (Salesforce, GoogleApps) platform svc
S containers
a
a
• PaaS – Platform as a Service P guest (Google App Engine) OS
• IaaS – Infrastructure as a Service host OS
S
(AWS - EC2/S3) a
a
I CPU disk
Jan. 22, 2009 Sun Microsystems 2009 5 Business models & Use cases – Set clear, focussed, few priorities
Rent or Build? Use How? • Use the cloud, > don't own hardware, > start-ups, development / research projects • Leverage the Cloud > temporary on-demand load > functional off-load • Build an 'internal' cloud > enterprise infrastructure grid (internal 'AWS') > standardized dev env / services (internal 'Google Apps) • Be the cloud > new businesses offering clouds or cloud services
Jan. 22, 2009 Sun Microsystems 2009 6 How / Why do clouds form? – Balancing 'Infrastructure' and 'Platform'
● Minimize costs – ● convert IT costs from capex to opex ● Cloud computing begins as a way to improve infrastructure resource deployment and utilization
● Maximize return – ● IT as a competitive weapon ● Cloud computing begins as a way to transform application development
Jan. 22, 2009 Sun Microsystems 2009 7 Key Cloud Tech - Virtualization
Type 1 Hypervisor •runs on bare metal •e.g. VMware ESX, xVM
xVM – a Xen-based type 1 hypervisor – built in a Solaris container – providing unique capabilities Type 2 hypervisor multi-threaded CPUs, • runs inside an OS 10GbE links & • e.g. VMware Server, Virtual Box QoS control
Jan. 22, 2009 Sun Microsystems 2009 8 Cloud Services / Developer Continuum – lock-in vs. choice decisions – generic API ??
Google App Engine Developer Developer lock-in independence
Google App Engine Sun Project Caroline AWS EC2 – AMIs Compute Model HTTP Request Handler Process Linux OS Instance
Service Creation Java, Ruby, PHP, Perl, any (including native Python (no native code) Language Python,(no native code) code)
network ZFS S3, network block Storage: Files local read-only local tempFS storage, local tempFS Storage: DB BigTable PostgreSQL SimpleDB Network Hidden Private Network Shared Network Net. Protocols HTTP/S, SMTP TCP, UDP TCP, UDP NAT (static & dynamic) Internet Hidden VPN, direct NAT (static) Connectivity L4 & L7 Load Bal. DNS static dynamic fixed
Jan. 22, 2009 Sun Microsystems 2009 9 What's running in the Cloud? – A shifting development model but which do I write to?
1998 2008 ?
Jan. 22, 2009 Sun Microsystems 2009 10 Cloud Space Ecosytem – it's about the partners & services
added value on-ramps, services n o i t c a r t s b A
f o
l e v e L
core infrastructure providers
Developer vs Enterprise
Jan. 22, 2009 Sun Microsystems 2009 11 Super-scale Cloud Architectures – It's about 'components' & patterns
Load Balancing Moving to open sourcing Unstructured enterprise patterns Sessions Data ● L7 HW – hi perf but expensive Keep thin for scalability Moving from local to ● Local Sessions (PHP, APC) == bad global FS ● Centralized Sessions (DB) good ● replication is hard ● No sessions == awesome
Structured Data Caching Scale out while Application server, watching database server or complexity both ● tough to scale / ● modPerl – vertical is best invalidation hard ● typical apps 80% ● query cache – read Data Set Intensive low performance ● ● design schema to Computing memcached – needs app logic avoid collision Exploding area ● becoming the new 'value'
Derived from Cal Henderson's 'Building Scalable Web Sites' Jan. 22, 2009 Sun Microsystems 2009 12 What's New / Next ? – Data Intensive Computing FOSS Analytics Semi Structured 'Data'
The End of Science • Chris Anderson • Wired 16.07 2008 • The quest for knowledge now begins with massive amounts of data.
Jan. 22, 2009 Sun Microsystems 2009 13 the Data 'Tsunami' – ingest, archive, distributed, fast, open..... MORE
• 281 exabytes • 45GB /person • 10X growth in 5 years • 50% thrown away • excess 'halo' effect ➢ 1.1 MB email to 4 people = 51.5 MB
Data From IDC 'Digital Universe'
Jan. 22, 2009 Sun Microsystems 2009 14 Refactoring Storage as a Service – Critical for Data Intensive computing ?
General Purpose Storage 'Servers' – Combine server with disk & networking – Specialized software enables general purpose systems designs to provide high performance data services
Sun's Open Storage Platform direction • Sun Fire X4500/4540 – Thumper/Thor 2 dual-core Opteron, 64GB, 48 hot-plug SATA drives, 48TB in 4 RU
• Sun Storage 7000 Unified Storage System Hybrid ready architecture High performance analytics
data functionality moves to the infrastructure
Jan. 22, 2009 Sun Microsystems 2009 15 Three Product Units – What's our 'Cloud' strategy? –The Network is the Computer Cloud Computing & Developer Platforms • DDeevveelloopp cclloouudd--eennaabblliinngg tteecchhnnoollooggiieess && sseerrvviicceess • HHeellpp ccuussttoommeerrss bbuuiilldd && Application Platform Software ooppeerraattee cclloouuddss > ffoorr oowwnn uussee oorr aass aa sseerrvviiccee • OOppeerraattee aa ppuubblliicc cclloouudd sseerrvviiccee Systems Platforms > ssttaarrtt wwiitthh IIaaaaSS && mmoovvee uupp • DDrriivvee ooppeenn ssttaannddaarrddss aanndd xVM ccoommmmuunniittiieess
Jan. 22, 2009 Sun Microsystems 2009 16 Which is the “cloud computer”?
Jan. 22, 2009 Sun Microsystems 2009 17 > THANK YOU