In this first of a three part series on the state of cloud computing, we summarize the key messages from the Cloud Leadership Forum, held June 14-15th at the Santa Clara Convention Center. Future articles will look at the network infrastructure side of clouds and the ITU-T Cloud Focus Group activities (including a report of their June 14-16th meeting in Geneva). Note that this is the third cloud computing conference this author has attended in as many months. The differentiator this time were the number of high level IT and computing vendor executives present The basic theme was how to view cloud computing a part of a company's business portfolio strategy, with emphasis on migration of applications from in house servers to the cloud. Technology gaps and critical issues were also identified and discussed. The Cloud Leadership Forum event was co-produced by the research firm IDC and IDG Enterprises (both owned by IDG).
John Gallant of IDG Enterprise kicked off the conference by welcoming the attendees and articulating the conference themes. He stated that "cloud computing will gradually grow to encompass all commodity IT services." While we don't disagree with that statement, we are unsure about the timing, i.e. how long is "gradually?" A couple of years or a decade or more?
Presentations and Round Table Discussion
Frank Gens, IDC Sr VP and Chief Analyst, delivered the opening keynote: "Bringing Cloud into the Enterprise … and the Enterprise into the Cloud." He noted that there were a mix of early adopters, key decision makers and "tribal elders" in the audience and that there was wide global awareness of cloud computing But what do CIOs think about it? Frank said that cloud services were a delivery and consumption model, applicable to many business and consumer offerings. He listed several important benefits and attributes of cloud computing:
- Cost efficient: Shared, standard service
- Simpler to adopt: Solution packaged
- Faster to deploy: Self-service
- Scalable to need: Elastic scaling
- Priced to use: "pay for what you Use" pricing
- More accessible: via the Internet; Standard User Interface technologies
- Solution ecosystems: published service interface/API
Cloud users are looking for better cost efficiency, reduced complexity, simpler and faster adoption, scalable performance based on user demand, priced to use, and widely accessable cloud services.
Frank then reviewed the well accepted three layers of cloud computing:
- Application (or Software) as a Service– to run application programs in the cloud.
- Platform as a Service– to develop applications and other software in the cloud
- Infrastructure as a Service– to use compute servers, storage, and networking in the cloud.
[Note: there are many definitions of the above terms. We believe the NIST definitions of cloud computing are the most recognized]
Layered on top of those is "Business as a Service," where the organization uses cloud services to extend its reach into more markets, collaborate with partners and customers and reduce the cost structure of operations..
The three well known cloud deployment models were then reviewed:
Private Cloud (what most enterprises are using today)
- Designed for, and accessed by a single enterprise (or extended enterprise)
- An internal shared resource,not a commercial offering
- IT Organization is the “vendor” of the shared/standard service to its users
- Customization for the enterprise is acceptable
- Could be on site OR off site
- Could be managed by a 3rd party OR in house IT staff
Private cloud was described as an evolution to consolidate, virtualize and automate IT operations. Up/down scalability, self service provisioning, pay per use model and simplified/integrated packages were all seen as important private cloud attributes.
Public Cloud (offers the most cost efficiency but security, privacy and reliabity issues loom large)
- Designed for a market, not a single enterprise
- Open to a largely unrestricted universe of potential users
- Resources shared – with privacy and security among a large group of enterprises
- Customization is limited
Hybrid Cloud (mix of public and private cloud- compute jobs/apps partitioned to run on one or other; bursting/ overflow from private to public)
- Enterprise’s cloud services portfolio includes both private and public cloud services
- Some specific services are delivered through a combination of public and private cloud models
- Static allocation between public and private, or private cloud “bursting to” a public cloud service
IDC asked CIOs what's appealing about cloud computing? The top three benefits, each gathering over 75% vote from respondents, were: pay only for what you use, easy/fast to deploy to end users, and pay as you go model.
The top challenges/concerns were: security, availability, performance, may cost more, lack of interoperability standards, bringing back in-house may be difficult. And these six worries were all ahead of any potential benefit to be realized from clouds!
The top applications likely to be deployed in a public cloud are in a state of flux, according to Frank, with only two gaining more than a 50% vote (it is not clear how IDC came up with these percentages as respondents were asked to vote for each item on a scale of 1 to 5). The top ranking public cloud workloads (each gathering more than 40% of respondents) were: collaboration apps, email, IT help desk, personal apps, data backup/archive, web apps, server on demand, storage on demand, IT management
There was a lot more conviction among respondents for private cloud workloads, with the following high ranking items gaining more than a 62% vote: collaboration apps, email, storage on demand, server on demand, data backup/archive, Business Intelligence/Analytics, IT help desk, web apps,
IDC wanted to find out what stage various organizations were in adopting cloud. The results were not surprising to us. 20.5% of respondents had already implemented and were already using cloud. 15.5% were considering cloud for specific needs while 5.5% were planning to adopt cloud in 1 to 2 years But a whopping 25.6% had no interest in clouds, while 27.9% were educating themselves about cloud. [That implies that well over 50% of respondents won't deploy cloud within the next two years, and it might be quite a bit longer.]
Finally, IDC asked who was best suited to be a trusted cloud service vendor The answers were all over the map and indicated a "brawl" between many potential cloud vendors. The list included: pure play cloud vendor, large and medium size software vendor(s), large networking vendor, large diversified IT vendor, global telco, lage IT systems integrator, etc.
In summing up, Frank conjectured that we are at "the Chasm" [Geoffrey Moore, Crossing the Chasm] in the cloud adoption cycle. We are in an extended evaluation and testing period where the cloud industry moves from early to mainstream markets. It is a time for critical learning and planning as cloud computing crosses that chasm and becomes a huge revenue generating business for all participants and players
The panel session that followed was titled, "Decision Frameworks for Acquiring and Moving to Cloud-based Services." The participants included:
- Brian Boruff, Global Vice President, Emerging Technologies & Strategic Growth Markets, CSC
- James Harris, Managing Director, Cloud Computing, Accenture
- Captain Nicholas V. Buck, Served as Director, Ground Mission Framework and Services Program, National Reconnaissance Office, US Navy
We found the comments of Captain Buck quite illuminating. The Navy thinks clouds will enable them to better prepare for missions. "Getting ready to get ready" is what it's all about, said the Captain. Their emphasis is currently on private clouds and infastructure as a service. They are currently deploying private clouds for collaboration and email. The Navy is looking at whether other applications might benefit from cloud services. "Will it improve on: content, timeliness, access? If so, the business case (for clouds) will make compelling sense."
For example, getting twice the throughput at a 30% cost reduction. The Navy is looking at potential cloud applications on a case by case basis. Mission critical apps are questionable as they need 24/7 availability and extremely high reliability. 40% of their IT costs are in storage, which would imply that storage as a service or backup might be considered for the cloud. Potential cloud service users were advised to look at cloud vendor use practices, especially with respect to providing security. The top echelons within an organization must buy into clouds- not just the IT department. Lastly, single sign on is critically important when collaborating with other users/ departments.
Jimmy Harris of Accenture suggested that services that supports an organizaiton's business are good candidates for the cloud. Can the workload be sourced in a different way? If so, cloud may be a candidate for service delivery. But that might involve several cloud service providers working together, especially if multiple organizations are colloborating via cloud services (e.g. supply chain or collaborative design and development). Jimmy observed that large companies are moving towards standard solutions and clouds would be no exception
The panelist agreed that "legacy apps" should not be in the cloud. These were described as narrow consituency apps that have been around for years. The applications that were seen as ripe for the cloud included: collaboration, email, CRM, or "any application that was already virtualized."
Tim O'Brien, Microsoft Cloud Evangelist, shared his perspective in an excellent presentation titled, "The Economics of Cloud." While cost savings is a compelling reason to move to cloud, there are other benefits that may be just as important. Foremost among these are IT agility and refocusing on things that have higher value (to the organization). Cloud capabilities, like self service provisioning and elastic scaling, go a long way in helping to realize those benefits.
Tim argued that "scale matters." In particular, there are both supply and demand side economics at work in the cloud. On the supply side, the user doesn't need to build his own data center and worry about the cost of hardware, labor, and power consumption. For the demand side, the problem of inefficient server utilization is effectively outsourced to the cloud service provider (e.g Infrastructure as a Service). Variability of workloads can result from: random usage patterns, seasonality, and time of day. This results in peak work loads that are often much greater than average loads. In many cases, servers must be over over provisioned to accommodate those peak loads, which, in turn, leads to very poor server utilization.
Mr. O'Brien stated that the server utilization problem is much more easily solved in a public cloud, where there are many more users, workloads, and applications running than in a public cloud. However, he noted that government entities/ agencies seem to be only interested in a private cloud.
Using a mid size company as an example, public cloud economics were quite good for several applications, including High Performance Computing (HPC), batch processing, development and test, web workloads, communications and collaboration. Customized business applications were seen to be a better fit for a private cloud. Best suited for the cloud are packaged applications common to all businesses, such as customer relationship management, enterprise resource planning, or e-mail. Less suitable for the cloud are customized apps created for a business process that is unique to a company's business, he added. Those applications are best left to the in-house IT department.
Microsoft expects a hybrid cloud environment will dominate for the intermediate future (five to 15 years). "I think the takeaway here is that we are looking at a hybrid type environment for the foreseeable future and I don't know what the foreseeable future is," O'Brien said. "The time frame is the great unknown." However, Microsoft expects a gradual shift to public clouds as some of the concerns (e.g. security, reliability, availability, privacy) are resolved favorably. Tim noted that some of those same concerns are common to private clouds, where a third party provides an off premises data center (in the cloud).
Cloud was seen as causing two distinct disruptions for users:
- Technology: shift from client/server computing to the cloud model (private, public, hybrid)
- Business model: licensing and maintenance to a subscription based model (for cloud services)
Microsoft provides on line calculators to assist potential cloud users in analyzing their cloud cost models. Total cost of ownership (TCO) as well as return on investment (ROI) tools are included.
In summary, Microsoft's cloud computing assessment is that:
- It’s about scale
- Economic forces driving the shift
- Pace of uptake increasing
- Hybrid clouds in the interim
- Consistency matters
And, of course, they're "all in."
There was great audience participation in a round table discussion on Interoperability, Integration and Management, which was led by Richard Villars, IDC's VP of Storage Systems and Executive Strategies. Richard asserted that cloud computing solutions address many business requirements:
- Not having to build a data center
- Lower cost of entry for start ups (SaaS, Web 2.0, departments, etc…)
- Efficient delivery of information/applications/content
- IT Infrastructure on-demand for intermittent, bursty or unpredictable workloads (e.g., reporting, analytics, media launch)
- Lower/shared cost solutions for long term archiving of digital assets (medical records, personal images, backups. eDiscovery archives)
Cloud was depicted as an IT delivery strategy with virtualization as the "gateway drug." The emergence of cloud accessories (or accelerators) could make the cloud faster, safer and more useful. Multiple environments could thus be navigated.
The audience participants identified several requirements for cloud Interoperability, Integration and Management:
- Identity and access management
- Data Integration
- Identity brokers (huge opportunity)
- QoS (parameters TBD)
- Commonality of SLAs and monitoring for compliance
- Audit for breach
- Measuring performance
Not surprisingly, one participant stated that large vendors don't want interoperability because they want to provide a turn key solution to the cloud customer (otherwise known as customer lock-in). Those large vendors maintain that open interfaces degrade performance. They say it is more important to have a good data architecture and know where data is going from a services perspective. A virtualization and security strategy is also needed. Nonetheless, the audience felt that the lack of universally accepted cloud standards was seen to be a critical issue to ensure interoperability, especially between private and public cloud providers.
Another participant asked, "Is there a good "boiler plate" (or checklist of attributes) available that covers key cloud services issues?" No one could identify one. "Who will monitor SLA compliance?" he asked. No one had a clue,
Attendees cited agility, new capabilities and cost as the main drivers for adopting cloud computing. Brokers who could match user requirements to cloud service providers were seen as an important step in enabling users to migrate from in house IT to cloud computing.
From the surveys, user concerns, open issues and cautiousness expressed by both users and vendors, we can safely say that cloud computing is not going to dominate the IT world overnight. While it has tremendous promise and potential, it'll be a while before the critical issues and perceived problems are resolved. The one that has not gotten nearly enough attention, in our opinion, is interoperability. Participants in that round table discussion definitely agreed.
Stay tuned for part II. in this series, where we'll summarize and evaluate Juniper Networks presentation for the new data center. Several other presentations and panels from this excellent conference will also be included in that article.