Leveraging Cloud Computing:
Pie in the sky or the next big thing on the horizon? (June 12, 2009)
If Cloud Computing is indeed the next big thing, then it is certainly a high stakes game. Current investments undertaken by Microsoft, Amazon, and Google in cloud computing indicate these firms are gambling with a Royal Flush; however, at the conclusion of ITEE’s latest discussion session on cloud computing, the group concluded these firms might merely be gambling with three of a kind.
Conceptually, cloud computing is compared with a utility—the source of the power, data and processing of data, can be distant from the final user. Just as electric power generation moved away from individual firms and became centralized, cloud computing promises to move our computing infrastructure to a utility that can securely, efficiently and cost-effectively manage systems in such a way as to allow us to plug into these networks with our devices of choice. The concept and vision of the cloud is revolutionary in scope, but is it practical? Is it the “next big thing,” or the “next big buzzword”? Millions of firms, large and small, have already invested in their own computing infrastructures.
What Differentiates Cloud Computing from Prior Models? Is It New?
Although there remains no formal definition of cloud computing, it certainly means that computing power is “not here.” It implies capacity on demand such that the cloud can add or subtract servers as necessary to meet demand. Consequently, the user pays for only that capacity which is necessary. “Instant on,” “pay as you go” and “capacity on demand” are buzz phrases of this movement. CIOs do not have to commit up front to buying capacity in the hope that they are making the right choices; instead, they can buy what they need, when they need it. It is not so easy to scale down if you have invested in hardware (however, it was noted that HP has a new rental model that provides a variable cost above a certain level of usage and does, in fact, allow some meaningful down-scaling to take place).
It is possible that currently existing IT practices have been renamed cloud computing due to marketing initiatives by major firms. There are three models where “cloud” computing may already exist. These three models include application hosting, a centralized data center, and the ASP (Applications Services Provider) plain vanilla model. They move from one extreme of strong user control over most elements (hosting) to the other of very little control (ASP vanilla).
• Application Hosting: a third party provides the physical data center infrastructure and (possibly) the hardware. Beyond this the user is free to run the application however he/she sees fit.
• Centralized data center/Private Cloud: A company may customize its version of the application, but consolidates servers such that they run the application from a single location and serve it to everyone from there.
• Application Service Provider (ASP)/ Software as a Service (SaaS) Plain Vanilla Model: remote users utilize an application just as it is, with no modifications beyond individual data. Examples: Salesforce.com, Google apps.
Further layers may be defined, e.g. business process management. In the imagination of business writers such Peter Fingar, any layer becomes subject to outsourcing (his book, “Dot Cloud,” received a partially enthusiastic review).
It was suggested that the scale of the investment by these large firms is the main differentiator between cloud computing and the three existing models. Other firms in this business, such as rackspace.com, are not as big as Google!
The Kindle also provides a window into how cloud computing may be different, as Amazon can now bundle the download cost into the book cost and, in essence, give the connection away for free. A similar move in the cloud computing world would indeed open up great new possibilities.
Other qualitative difference may comprise advances in virtualization, hyper-layer scanning (e.g. where virus scanning is done for all the processes running on one server or one part of the cloud, obviating the need for each process to do it), and the installation of patches. But some of these can be done, and are being done, by individual firms using virtualization, and so it is not obvious exactly what are the unique advantages of cloud computing.
What do Users Get from Cloud Computing?
A few in attendance have had experience with cloud computing. Benefits include minimizing hardware and running applications off site with ownership and management still available. Other motivations involve cost savings of shared servers and minimal security issues. Potential drawbacks surface in data transfer and it becomes clear that firms remain tethered to their local computing power when the applications involved are sharing large amounts of data remotely (e.g. with a video or media server application). Doing .NET development remotely is entirely possible.
The primary utilization of cloud computing among some of our participants currently consists of e-mail. Investments of large firms such as Microsoft currently center on hardware, applications, and building data centers. Currently, applications only handle e-mail; however, in the future, opportunities to develop custom applications within the cloud may become available.
Perhaps the best candidates for cloud computing comprise firms experiencing drastic shifts in capacity demand. Examples given included elections and the Olympics because purchasing hardware creates excessive and unnecessary expense. E-commerce companies experiencing seasonal demand patterns and utilizing customized software applications might fair best in a cloud-computing environment.
Project based models like R&D that need to scale up quickly over short times are also best suited to cloud computing. This is the exception rather than the rule for most firms because it does not operate in a vanilla state. Future options include vanilla and customized models available for different needs. Pharmaceutical work could also be ideal for cloud computing, because demand varies so much depending on stages of drug research.
Can cloud computing give open source computing a boost? The panel agreed that migration by service providers and e-commerce firms looking into clouds provides evidence.
Viability and Maturity of the Cloud Model
Users’ Perspective: The first question under this topic is whether there is sufficient demand for cloud computing. Cloud computing will not make mainframe or super computers obsolete. The insurance industry, for example, relies on mainframes to manage its bulk of data. Current budgets on cloud computing within the panel are very small and less than $500,000. Large companies, currently heavily invested in their own IT, will not give it up immediately to move into clouds. A firm that was writing a lot of its own software, and needed peak capacity for things like stress testing, could find the cloud quite beneficial. For example, this model might be beneficial to a firm that writes its own ERP system; firms that are writing custom applications may be able to write them to specifically take advantage of the cloud’s ability to scale. However, today there are few firms writing their own ERP systems due to economies of scale. This case and the case of using vanilla applications like email are two strong candidates for cloud computing.
Several other points surfaced during the discussion concerning the viability of cloud computing in an environment where the majority of what a company runs is not compatible with the Software as a Service/Cloud model. First, most businesses modify vanilla programs (such as accounts payable) to suit their preferences. Second, using custom software to obtain a competitive advantage is a primary concern of most firms. Most firms desire to keep this strategic software in house for security purposes. Third, purchased commercial software does not currently sit well in the cloud environment because most businesses do customize the programs. If the cloud provider could serve the versions that had been customized to each individual firm, the model would be more attractive (note this is like saying that an electrical utility would provide different grades or types of electricity to different firms). How would the cloud provider know how to upgrade a package you had customized? It was asserted, therefore, that the bulk of the software that businesses currently run is not well suited for the cloud model.
A firm that wants to use customized software versions can use the Application Hosting model, but has to be responsible for taking care of all these details itself. And because infrastructure is so cheap these days, the hosting model does not seem to offer than much of an advantage to firms with existing infrastructure already.
Additional questions came forward such as how much control a firm desires over its hardware and renting space on a commercial server versus owning its own.
Other dangers in entering into cloud computing entail different departments utilizing different clouds independently. Corporate policy must dictate future procedures. And there seems to be some risks if the cloud should experience large failures.
Providers’ Perspective: A second set of questions involves the provider firms themselves. One possibility is that these big companies will now become experts at forecasting demand. The maturity of the cloud concept may develop when we see supply chain visibility such as e-commerce businesses exchanging forecast data with cloud providers.
Issues arising from cloud computing involve the need for the host companies to change both pricing and licensing structures due to the scalability factor. Currently, contract violations may exist if demand drops and the contracted capacity becomes unnecessary. So how do these firms scale down? The answer is that the big firms are gambling that customers will come despite the fact that scalability remains unclear.
Using the Cloud In House
Currently, many companies create their own private “clouds” to store data. The primary inhibitor of this is slow data transfer rates. By consolidating servers and adopting virtualization, many firms are already, to a certain extent, creating their own clouds. For companies that are thinking about implementing a “private cloud,” scalability is much easier for large firms possessing elevated internal IT investments. Resources re-deploy easily to other uses as necessary. An advantage exists for large firms with established IT resource investments.
Bottlenecks must be identified in order to allow the successful implementation of cloud computing. Applications must be able to scale according to demand. Current software packages may not take advantage of a scalable hosting platform. Some applications are still single-threaded, whereas with a four-core Intel processor with hyper-threading, there are sixteen processors available. Software written in house may require extensive re-tuning when changing hardware platforms. The existence of massive hardware capabilities, in a sense, guarantees that something else will be the bottleneck.
Economic factors must provide necessary incentives. Long and short run costs require comparisons similar to leasing a car. The long run question of whether or not to host in-house requires comparing the cost of internal inefficiency by doing it in-house, vs. the cost of loss of margin by paying for it “out of house.” The gap between the two provides the answer. Determining how all of the pieces work together becomes critical in the future of cloud computing viability. In the end, companies must resist the seduction of capacity or scalability benefits alone. E-mail is an example of something that firms are becoming more and more hard pressed to do more efficiently in house than can be done by an outside provider.
Security of the Cloud
What are the advantages and disadvantages concerning security? The group cited advantages regarding patches and updates performed on one cloud benefitting multiple companies. More than one company may use the same physical server, so all benefit when patches are applied. Disadvantages included the possibility of a hacker gaining access to a cloud and gaining entrance into multiple companies. One panelist jokingly remarked that hacking clouds might become a topic of instruction at the college! It was noted, however, that the argument about hackers has been around for some time with respect to virtualization. It has to be reviewed and addressed, but such instances are not being seen, at least not yet.
Server confiscations and other legal disputes also raised concerns among the group.
Another question concerned how to handle outages in a global environment. Some large firms locate their clouds as close as possible to the customers. Theoretically, technology might provide zero outages; however, applications must develop to eliminate downtime.
In addition, Sarbanes Oxley auditing concerns might pose problems especially with global companies where a cloud might be located in one country and the operations in another country. Will auditors need to know who is in the cloud? A member of the panel indicated that these concerns are unfounded because this type of auditing occurs currently without issue.
Further concerns surfaced about possible locations of data centers. I a firm receives a cheap price in Burma or Myanmar does it need to worry about a rebel overthrow and data falling into the wrong hands? From a security standpoint, does a firm care that someone might be physically on site in the cloud facility?
Does cloud computing comprise a new un-regulated frontier? Are there any companies currently facing regulatory issues hosting in another country? Is privacy an issue, especially concerning credit card transactions? The panel agreed that all these concerns deserve investigation.
The Bottom Line (For Now)
In sum, the complexity of the entire IT industry is rapidly advancing. Industry CIO’s face a mosaic of both long and short-term choices entailing critical decisions. It is entirely possible that the term “cloud computing” is a mere buzzword to drive increased revenues. Or firms are just on the cusp of shedding their own computing capacities; once data transfers for really large amounts of data become possible, the source of the computing utility may shift. Only time will tell if the gamble of the large IT firms in cloud computing infrastructure will pay off. The “three of a kind” models already exist, but perhaps cloud computing will prove to be a “Royal Flush.”