Passions about cloud computing vary from the ho-hum to the evangelical.
The ho-hum crowd sees cloud computing as Back to the Future computing. Ride the DeLorean back to the 70s and return to the heyday of "timesharing" - i.e., renting computing resources on more powerful remote computers. Customers purchased resources by CPU hour, connect hour, and disk space and were surcharged for specialized software like Model 204, a high performance data base management system then known for its bit-mapped indexing technology. (Model 204 is still commercially available almost 40 years later - a rarity.) Technical challenges, the Ho-Hummers argue, aren't that different from today; the issues in distributed computing were well understood then. The benefits to users then were pretty much the same as they are today. Timesharing users had little or no responsibility for the data centers housing the remote high performance computers. They gained access to large scale resources with smaller-scale budgets or staffing commitments.
The evangelists - let's call them Cloudistas - argue that then was then, and now is very different. Today, they insist, bandwidth is vastly greater, customers operate their own networks, and considerable local processing can be performed on personal computers and handheld mobile devices. Further, the user communities are vastly different. In the 70s customers were confined to government and fairly well-heeled enterprises - probably numbering in the thousands. By contrast, one poster child for the 2011 version of the remote mainframe - iTunes - has as many as 200 million users. One in four U.S. adults are already using mobile applications - and many of these apps transparently rely on cloud services. The consumerization of cloud computing, say Cloudistas, makes possible economies of scale and performance that were unthinkable in that distant computing era.
NIST's Cloud Computing Program statement put it this way: "The Cloud Computing model offers the promise of massive cost savings combined with increased IT agility." Whether this proves to be the case or not, NIST aptly identifies "fast wide-area networks" as the first of three key enabling technologies (along with fast servers, and high-performance virtualization for commodity hardware). NIST's view represents a consensus of sorts, and, perhaps, indirect marching orders for network engineering.
Apple's recent iCloud announcement, like Google's Chromebook, ensures that, sooner or later, enterprises will need to come to terms with cloud computing. Just which facets of the cloud trend matter most is a matter of some dispute, so the list provided here is intended to identify trends of interest to network specialists.#1 WAN feature set
There's no avoiding it; sooner or later there will be one or more cloud services that some user or department in the enterprise will want to connect with. Many shops will need to pay more attention to WAN pipes than in the past - and not only bandwidth.#2 Location-independent computing
The idea of location-independent computing isn't new, but now millions in the workforce will have firsthand experience with it. It will be natural for them to expect applications they use, purchase and develop to be able to operate wherever workers find themselves. For multinational enterprises, this will be a particular challenge.#3 Increased scale and reliability expectations
The notion of synchronizing content isn't a new one. It's been part of personal information manager (PIM) software for a couple decades. What is new are scale and expectations. The content to be synchronized won't be limited to a few hundred lines of text (e.g., just contacts), and those potential terabytes being slewed around will be expected to work dependably. Perceived improvements in up-time performance and functionality for cloud resources may, in some quarters, be difficult to match with locally managed resources, which could affect the careers of some IT professionals.#4 Additional points of failure
Updated network diagrams for cloud-reliant operations may reveal both additional points of failure, and failures that affect many more users.#5 Reworked DR and risk assessments
Dependencies on cloud computing will lead to revisions of disaster recovery plans as well as updated risk assessments. Some risks may be reduced; e.g., cloud backups for modest volumes provide a level of immediate offsite security that is not easily matched with on-premises services. Other risks, such as having entire departments or even entire enterprises down during outage periods may not have been considered; pre-cloud mitigation strategies could be rendered infeasible.#6 Synchronization is nontrivial
While file-level synchronization, such as that offered by iCloud and similar services, is at least seen as relatively straightforward, other kinds of synchronization may not be at all trivial. Synchronization failures can degrade the integrity of backups, and errors introduced by such failures could go undetected for days or even months.#7 Lockstep upgrades and migrations
While some cloud service providers may choose to give customers some control over configuration management of services customers have subscribed to, others may not. As a result, internal project schedules, test plans, and training budgets may be impacted by unavoidable version changes mandated by providers.#8 Balance sheet influences
Rented services have a different profile in a company's financial statements. The profile will vary from firm to firm, but the shift from IT capital assets to rented services may have implications not fully understood.
Cloudistas may think they're promoting a comparatively carefree model for computing, but, as with other transformations within IT, it's more likely that one set of problems will be traded off against another set. Network professionals accustomed to managing internal flows may turn their eyes toward the perimeter more often than before, but concerns such as security and reliability are pretty much the same old fare off the same old menu.
Can you think of other cloud consequences for network specialists that aren't listed above?
Mark Underwood ("knowlengr") works for a small, agile R&D firm. He thinly spreads interests (network manageability, AI, BI, psychoacoustics, poetry, cognition, software quality, literary fiction, transparency) and activations (www.knowlengr.com) from a large island near NYC.