Now, the individual sessions, I picked Jim Blakely’s Cloud Transformation of Business Models, Architectures, and Usage. The room is packed. Mostly vendors and cloud operators, but a good number of hands went up for enterprise IT folks.
Jim opens with story about making the ASP (application service provider) rounds. Met with potential customer in IHOP to talk about $150 million data center, made him think this thing “ASP” didn’t have legs. It didn’t, it fell apart 6 months later.
Now, with 10 years of experience, believes it is real.
Three things:
1. Not just one cloud
2. Dispel myth of commodity computing
3. The balanced data center <or cloud>.
Ontology Storm Continues: has a slide up of different cloud definitions and taxonomies. There won’t be one model. Important thing to look at is the real dimensions:
The Stack Model
- The Web
- SaaS
- PaaS
- IaaS
- Data center operation
Other models:
The application model – what is the application you are going to run in the cloud: legacy, new, Web 2.0?
Targeting model – specialist clouds vs. generalist clouds (Amazon)
Provisioning model – customer integration, self-service
Tenant model – single tenant (private clouds), SLA multi-tenant
Jim calls out single thing that defines cloud is elasticity
Examples:
Cleveland Clinic, Enterprise Web 2.0 in the cloud; patient records. [I saw this as example of SOA last week…no surprise, services, service orientation are important cloud constructs]
Obviously, there is a lot of compliance (HIPAA) and privacy concerns. Might lead to special types of clouds. [Me, I’d like the secure my data cloud please]
Myth of commodity computing
The reason cloud computing is successful is not (as some folks say) because computing is commodity
The reasons are disruptive technology and Volume Economics and Metcalfe’s Law
Re: disruptive technology, Jim prefers term “Extreme sustaining technology” as these technologies come into play, cloud becomes more important. One reason, multi-core processors force applications to be broken up into smaller units. (This is Intel after all)
Other extreme sustaining technology candidates: energy proportional computing, web 2.0 and cloud software, solid state storage, virtualization, wireless broadband access, many & diverse broadband access clients
Examples: Rapid Software Tech Innovation
Yahoo Hadoop Clusters and Eucalyptus Architecture: WS-Cloud; eucalyptus is open service version of amazon services model
Now, Jim is talking of some hardware innovation related to launch of Intel Nehalem. Innovations related to power management, SSD and more.
Jim’s question: Do you need a data center and a cloud stack to win. Big internet providers run their own data centers. Data center management and application development/delivery are very different skills.
Concept of the Balanced Cloud
The balanced cloud data center “Do you have the right architecture and mix of compute, storage, networking and power for the workloads that run in the cloud?”
The balance is breaking, compute up 30x, network up 10x, hard drive IO up 1.5x over time. Everything is growing, but not in balance. Could be problematic.
SSD wear out overtime, after being written and rewritten. Research going on in this area. Flash wears out.
Flatter data center hierarchy better for application architectures like SOA that have many interconnections.
