Cloud computing is about using and providing computing resources in new ways. The session will be about some success stories in using cloud computing.
Relying on cloud computing in companies will go from virtually 0% (2007) to 20% in 2012. But it will not be dominant and it will take time to being accepted. There are lot of places where you already use cloud computing (e.g. WebEx).
A small private bank in Canada used to run its services on an ASP model and is turning to cloud computing to remain competitive against larger (huge) banks.
David gives the Gartner definition for Cloud Computing (see last year's sessions). The cloud enables companies to focus on their business.
Three critical issues in this presentation:
- How have companies successfully used cloud system infrastructure services? (get rid of servers)
- How have companies successfully used cloud application infrastructure services? (get rid of middleware)
- How have companies successfully used cloud application services? (get rid of applications)
Classically, companies try to transfer their internal applications to an external host: that just introduces security issues and does not give a cost advantage. There is a right balance between internal and cloud applications.
Example 1: Eli Lilly
This is a SMB, which are the pionneers in cloud computing and should be examples for larger enterprises.
The need was to reduce the fixed IT costs and their infrastructure was inhibiting their business (they need to be extremely reactive when delivering new drugs). When researchers needed new servers, they just went to Amazon and had it running in a few minutes and it cost less than 100 USD (for 1 day of computing power).
They have expanded to other providers (including Google Apps for collaboration). Reduced delivery from 7.5 weeks to 3 minutes for a new server, 8 weeks to 5 minutes for a new collaborative environment and from 12 weeks to 5 minutes for a 64-nodes Linux cluster (for scientific applications). It does not replace their HPC infrastructure but it augments it. Time delivery is the major benefit.
They still have security concerns: leaving the data in the cloud, issues with identification and SSO. Then there is still a vendor lock-in.
Example 2: large energy company
They had been looking at the web for long and were predisposed to Cloud Computing. They have strict security policies but some applications have lower requirements.
For setting up a wiki & blog site, the team was told to wait for 3 months to set up the environment. They setup an environment on Amazon in one afternoon and it hen went trough a thorough security check Everything was in production in 2 months.
The team invstigated an open source ECM pilot on Amazon. An intern created an Alfresco on Fedora image for evaluation. The whole project took 2 months, then the image was retired (but kept as a reference for further use)
The benefit here is agility: respond quickly to new business needs.
The same company had an internet-facing application with lower security requirements. Deployment on the cloud lowered costs and provided agility.
They are also using it for development and testing, especially on SAP projects. Another benefit here is simpler processes.
A lesson learned is not to understimate the cultural and political issues. For example, not involving security teams early creates tensions.
Recommendation: setup a "cloud team".
Example 3: Razorfish
They wanted to respond quicker to their customer demands. They used external resources but this did not address all their requirements (e.g. high volume short run campaigns). They switched to Rackspace cloud infrastructure. It cost them about 25% of their previous infrastructure (which was already hosted).
Key lesson: look at your process model first.
Example 4: Wipro
They setup a private cloud, with a self-service portal for dynamic provisioning. It addressed excessive provisionning times for new projects.
Benefits: 46 days to 5 minutes for provisioning a server, utilization rate went from 10 to 40% and reduced capital expenditures (30%). But the cloud can add management complexity. It requires to build a highly standardized environment.
Example 5: Japan Post
Recently privatized, JP manages financial services as well as post. The privatization window was short and fixed, they developped a new application in 3 or 4 months, for 57,000 users. It was built on Salesforce.com and Force.com, with matching development processes. The results include high user satisfaction (on functions and performance).
Example 6: METI (Japan)
The Ministry needed to build a new application for the public, to be deployed very rapidly. They used Force.com and built the application in 3 weeks. The key lesson is scalability (the platform handles peak accesses with million of transactions).
Example 7: Presidio Health
To help with rapid growth, they used Appistry (coming from the grid market), processing internally stored data (for regulatory reasons) on the (external) cloud.
Example 8: Federal Express
They have complex algorithms and they did not have the required processing resources. They broke up the batch processes to make them parallelizable and deployed them on an internal grid (built with Appistry). The batch run time has gone from 4 hours to 20 minutes and development is 60% faster.
The hardest thing was to adjust the mindset of developers for componentized parallel execution.
Examples with smaller companies
Author Solutions automated the end-to-end self-publishing workflow with an hybrid cloud (Amazon, Salesforce) and integration with on-premises systems.
SaaS vendors, such as HR AnswerLink or Coda, also use the (platform) cloud...
Fulcrum integrates Salesforce and Quickbooks (internal), with Pervasive DataSynch.
JohnsonDiversey adopted Google Apps to replace e-mail and augment MS Office. Investment paid back in 14 monhs.
Lessons learned
- Cloud is good for: web application, that do not have much interaction with the back-end systems, short-term use, rapid ramp-up required.
- Issues remain with security, data location, governance...
No comments:
Post a Comment