Archives for February 2009

Dave Linthicum is up now.  He says his purpose is to be a curmudgeon.  So, we’ll see how that goes.  He’s talking about the relationships between enterprise architecture, service-oriented architecture and cloud computing.

He warns us that this presentation is relevant for this instance in time.  In six months, this will change because cloud computing is rapidly evolving.

SOA & Cloud Computing

“The trick is to determine which services, information, and processes are good candidates to reside in the clouds, as well as which cloud services should be abstracted within the existing or emerging SOA.”

Basic idea is that SOA resides in the enterprise and leverages cloud resources (amazon, xignite, salesforce etc).  Leverage solutions you don’t own or host.  Morphing the lines between on-premise and off-premise solutions.  Dave sees the future as a hybrid model, not all in the enterprise, not all out.

Dave walks through his cloud computing framework, originally posted here.

Enterprise IT is understandably skittish about cloud computing. However, many of the cloud computing resources out there will actually provide better service than on-premise.  Security & performance are still issues.

Always have to understand your business requirements and cloud offerings and match for best value for your organization. 

When cloud computing may be a fit:

– when processes, applications and data are largely independent

– when the points of integration are well defined

– when a lower level of security will work just fine

– when the core internal enterprise architecture is healthy (cloud doesn’t give healthy architecture to existing problem applications)

– when cost is an issue

– when the applications are new (porting costs of legacy applications is an issue

Dave’s advice:

1. start with architecture – understand: business drivers, information under management, existing services under management and core business processes

2. get ready: (a) accept the notion that it’s ok to leverage cloud hosted services in your soa (b) create a strategy for the consumption and management of cloud services (c) create a proof of concept now

In other words, play with cloud computing before jumping in.

Dave offers a 17 point stepping into the clouds plan, starting with accessing the business all the way through implementing operations. 

Other thoughts:

External cloud services should function like any other enterprise application or infrastructure resource.  In fact, should be more reliable than internal.

Cloud resources should appear native.

Consider private clouds if your requirements call for that (security, control, privacy)

Watch the hype!

David Bernstein, VP/GM Software Group, Cisco on fundamentals of cloud interoperability and standards.  Echoes Russ Daniels’ statement that “The cloud is the internet’s next stage of evolution”.  Says pay attention to what google and microsoft are saying and doing in cloud space.

What is Cisco doing?

Cisco doesn’t run a cloud.  Does run a SaaS application, WebEx.  Cisco has helped a lot of people build a lot of infrastructure.  Cisco is ‘arms dealer’ in cloud computing.

Cisco’s Cloud Strategy:

– Build right products (unified fabric, unified compute, virtualization aware)

– Technology (Enhanced IP core with tight coupling to Software)

– Referenced software – services-led cloud blueprints, reference software stacks

– Open Standards

– Multi-phased (stand alone to enterprise to intercloud)

See cloud as the 4th wave of application infrastructure: mainframe, microcomputer/client server, web/internet, cloud.  This next wave of applications will be highly connected, highly collaborative, media intensive applications.

Cloud adoption phases:

1. Standalone clouds – derivative pricing, external data centers.  Folks want better security, SLAs and control

2. Enterprise class clouds – key challenges federation, portability, market

3. Intercloud – dynamic workload migration, apps integrate across clouds, and more

The intercloud is the grand vision. 

Consider an intercloud example of dynamically moving a workload (VM) from one cloud to another.  Some of the steps to accomplish this: 

– Communication (XMPP)

– Naming, Presence (cloud discovery)

– Cloud Trust

– Cloud negotiation (policy, entitlement, security, metering)

– Cloud sets up another cloud (placement, deployment, format, motion)

– [couple of more steps, to move to cloud 2, then mask addressing so cloud 2 is found as cloud 1]

Another dynamic workload example, federation, generalized service access across clouds

– same steps 1-3, then Cloud 1 needs to query Cloud 2 (RDF/SPARQL, OWL)

– Cloud 1 selects, receives protocols, interface (web services, REST)

– Cloud 1 calls services in Cloud 2 (metering, SLAs)

These examples illustrate the technology challenges related to cloud interconnection.  How does this translate into what Cisco is doing?  Some specific Intercloud Projects that Cisco is involved in to bring the intercloud to life:

– addressing, IETF LISP

– virtual machines – DMTF OVF

– conversations –  XMPP.org

– Unified Cloud Interface (UCI) – W3C, Google Code

– Distributed Storage Acceleration – open cloud consortium, UDP based data transfer

David is making a call for participation for folks to help solve these technical challenges.  Not a call to help Cisco, but to help the industry.

The strong cloud computing summit line-up continues after lunch with Jinesh Varia, Technology Evangelist, Amazon Web Services, Amazon.

What do large enterprises like about Cloud Computing?

Bit of a Nick Carr segue, businesses used to produce own electricity but didn’t add value to end product.  Same can be said of computing power.  Don’t need to generate computing power, and can get better economics and scale with external provider. 

Use only what you need.  Pay as you go model.

Value of time.  7-8 weeks from ordering to getting machine in place for development.  With the cloud, this can be accomplished in minutes.

On-demand provisioning through web services.  Software that creates and manages hardware.

AWS enabling Pluggable Cloud Architectures: compute (EC2), storage (S3 and SimpleDB), messaging, payment, distribution (caching at edge locations), scale (new services that sit on top of EC2 for load balancing, monitoring and auto scaling).

Is the cloud ready for the enterprise?  Essentially, Amazon cloud features list, including:

– elastic IP addresses

– availability zones

– new instance types (M1, C1)

– elastic block store and snapshots

– public datasets

– platforms – windows, open solaris, linux

– partners, ISVs, SIs etc

– security across offerings

– load balancing (new)

– monitoring (new)

– auto scaling (new)

So, of course “Yes”, the cloud is ready for the enterprise.

Some reference examples: Autodesk, Eli Lilly, ESPN (fan pages built on EC2, data on S3), another pharma doing protein calculations, NY Times and Washington Post.

“cloud computing is inevitable, it will get more mature in 2009, early move advantages to enterprises”.

Jinesh closes by asking “What would you like to do [in the cloud], but cannot?  Why?”  Sounds like an open call for requirements to me.

Now, Russ Daniels, VP & CTO, Cloud Services Strategy, Hewlett Packard.  Some Cloud capabilities have been previously discussed (in industry) as utility computing. 

More interesting, from Russ’s perspective: “The cloud is the internet’s next stage of evolution”.

The cloud is great for connecting.  It’s more than an alternative delivery system.  Connecting people to people, people to experiences, data to context & location, businesses to customers and businesses to businesses.

“Experience is a never ending stream of events.” 

The cloud removes the technology burden from individuals.

Providing an example of human life events managed from a cell phone via e-government cloud services that can be replicated for many emerging markets.

Now, a “self-serving” (his term) example of a print publishing ecosystem.  Create, Produce, Sell and Use.  Today, point to point connections, recreated for every pairing of publisher and service provider. 

Speaks to how the cloud allows publishers, other than giants like NY Times, to have fresh, daily, local publishing.  HP has Magcloud offering.

Cloud enablers/contributors: cloud computing (scale, commodity compute resources), SOA (service centric design), web 2.0 (rich user interface, focus on social computing, user is not afterthought, user is first class concern of design).

No longer an application centric view.  Service view.  Requires separation of data from consuming application.

Design drives cloud economics.  Traditional IT has distributed design, where the components constrain each other (hardware, middleware, OS, applications) and customer is responsible for integration.  Cloud has integrated design – compute, OS, middleware, applications — all in integrated system, with common management services.

Evolution of cloud service platforms: dedicated (purpose built, amazon bookstore), multi-purpose (support multiple applications, amazon retail), partner (support high touch business partners, amazon retail services), platform (platform as primary value (amazon retail platform (if amazon chooses to)).

The evolution of architecture: logical (structural description of a software system), enterprise (process-centric description of an enterprise), ecosystem (role/responsibility based description of an ecosystem, ecosystem in which an organization (or individual) exists).

In closing, this is a large and complex space, this talk just scratches the surface.

Now, Peter Coffee from salesforce.com is up.  He believes that “Enterprise cloud computing implies API leverage”.  Start conversation with the function you want to perform, not the underlying technology.

He has a quote up from William S. McNee, of Saugatuck Technology “On-premise computing is going to drop-off a cliff” (or something close to that, the slide changed)

Salesforce cloud began with CRM… Fundamental ideas: enterprise software accessible via web, web-based systems should be designed for global scale, everything not distinct to customer should be shared, everything distinct to customer should be customizable

Layered model:

application exchange

user interface as a service

logic as a service

integration as a service

database as a service

infrastructure as a service

Ok,Peter talks way too fast… plugged Python as something IBM Rexx programmers would like.

Enterprise wants secure, high available, low cost solutions.  Enterprise is not, anti-cloud.  If these characteristics exist, enterprise will move to cloud.  If it’s too difficult to move existing portfolio to cloud, then keep that in-house and go forward on cloud.

many tech trends oppose governance goals:

– processing availability is enemy of encryption

– connectivity provides attacker opportunities and tools

– storage, ever growing risk of larger data losses (thumb drives, laptops, email)

Service models offer greater leverage:

– granular management of roles & privileges

– data is in the cloud, not carried around and subject to physical loss

– [one more, sorry…]

Now, talking of developer productivity using force.com vs. java development; overall project cost 30-40% less

Points to charts and metrics from the economics of the cloud post on O’Reilly.

PaaS puts IT spending back in balance.  Conventional IT model front-loads capital expenditure on infrastructure.  PaaS enables preparation for upturn.  Build now without big up-front investment, scale at upturn via PaaS.

“Do not mistake the consumer Web for the enterprise cloud”.  It’s not like putting your G/L on facebook.