For this month's data center advisory board question, we asked our panel if data center managers should consider cloud computing for their operations, and under what scenarios.
The panelists offered very specific advice on where to use cloud computing, and provided a checklist of cloud computing risks to consider, while others offered healthy skepticism about the entire concept.
The advisory board is made up of a diverse group of data center professionals -- facilities executives, mainframe programmers, CIOs and systems administration experts. You can read their full bios on the Data Center Advisory board page.
Get smart: IT managers need to be cloud computing subject matter experts
Cloud computing is more than just a buzzword. Cloud computing is real, it is out there and IT managers cannot ignore its existence. In fact, given the broad range of definitions and the various forms cloud computing can take, I would even say it is incumbent on the IT manager to maintain a current understanding and be conversant in the technology serving as the corporation's subject matter expert.
The core promise of processing and storage capacities on-demand that can flex dynamically with varying loads without owning a single piece of hardware is intoxicating. The devil lives in the details and those details include pricing models, international issues, service interruptions, compatibility of integrated applications on the cloud platform, and the platform provider's customer service.
With a solid understanding of the associated issues, strategies can be crafted mitigating any perceived risks and only then can specific organic processing/storage requirements be considered for migration to this cloud. Such a migration should only be considered when the corporation has embraced all the upside benefits having considered all the risks and that will only happen when a trusted source is fully competent in the cloud computing environment. The IT manager should be that fully competent trusted source.
--Ben Stewart, vice president of facilities engineering at data center colocation company Terremark Worldwide Inc.,
Cloud can work for horizontally scaling apps on commodity hardware
Cloud computing is not a cure-all for IT department growth problems. There are many data centers that simply won't benefit much from cloud computing. That having been said, there are a few areas in particular where IT managers should definitely consider cloud computing options. Cloud computing works particularly well if your applications run on commodity hardware and scale well horizontally.
If you can increase your capacity simply by adding another twenty nodes to your infrastructure (such as with standard clustered LAMP deployments) you should try putting a few nodes in a cloud for a month and see if it works for you. In addition if you have a web service that serves large files (such as image or video sites) and are having trouble keeping up with the storage and bandwidth demands, it might be worth considering using cloud storage for at least those large files.
Another great use for cloud computing is for temporary increases in capacity. If your site traffic is mostly flat but spikes in short, periodic, predictable ways (such as at the end of the month or quarter, for instance), it might be more cost-effective to supplement your current infrastructure with services in the cloud since many of them allow you to increase or shrink capacity on demand, and only pay for what you use.
Finally if you need to increase your capacity but are having a hard time getting the budget for a large capital investment, putting some services in the cloud could be a good way to address your growth needs temporarily while you wait for the budget for a more permanent solution. Who knows, after a month or two you might find that the cloud is a better long-term solution as well.
--Kyle Rankin, senior systems architect at Quinstreet and Linux author
Cloud computing considerations and risks
I think cloud computing presents great opportunities to businesses of all sizes. Smaller companies can get the latest and greatest software without paying staff to maintain it. Large companies can also avoid huge "per seat" license charges and get upgrades without time consuming rollouts to thousands of desktops.
For larger companies, however, I would make a distinction between office tools and business applications. For standard tools it makes sense to rent them especially from a total cost of ownership (TCO) perspective. On the other hand, applications run the business and most companies like to include custom features designed to provide competitive advantages.
Hosting mission-critical applications offsite has a couple unpleasant side effects. First, the application is subject to the limits of the hosting company's tools and staff. Second, some host companies build applications on a proprietary platform, thus limiting your ability to comprehend the code, let alone move to another provider.
The decision about cloud computing must rest on these criteria:
- Security: How important is the confidentiality of your customer or corporate data? Does the provider support the appropriate levels of security? How well can the hosting company match your security model?
- Cost: Your company must weigh the monthly rental fees against the cost of self-hosting. Some companies may reach the break-even point sooner than others.
- Capacity and Service Levels Agreements (SLA's): Does a provider have enough capacity to support your company at an appropriate level of service? Are there any penalties if you need to grow beyond the original agreement? What will the provider do if it misses the SLA?
- A backup plan: Somewhere there's a backhoe waiting to cut through your fiber bundle. What happens if you lose contact with your provider? How would you plan and implement backup routes?
- Disaster Recovery: What will happen if you or your provider has a disaster?
- Control: Does your company need to control software versions, and who gets what, when?
- Plan B: It's never too early to think about how you might bring the processing back in-house or move to another provider.
--Robert Crawford, lead systems programmer, mainframe columnist
Examine all low-priority, intranet apps for cloud computing potential
IT managers should be looking at converting their on-premise infrastructure to what we recently called "Software-as-a-Service" and now the bucket of "cloud computing." If your email isn't in the cloud already, there should be a fantastically good reason, like regulations that prevent off-premises email.
Can you host your instant messaging in the cloud? How about file sharing and basic intranet functions? Even things like SharePoint look attractive. Essentially, you want to inventory all of the low-priority items you have on your intranet and ask if it's cheaper to move them off-premise.
In looking across your portfolio for things to move to the cloud, you still need to ask why you should do it. Top of the list tends to be cost (both up-front and ongoing, especially when it comes to upgrading and maintenance) but also flexibility and new functionality that come with cloud-based applications. If there's no reason to move off-premise, however, don't get caught up in doing it because it seems like the right thing to do: make sure you can justify it.
--Michael Coté, analyst, RedMonk
Do a full risk analysis before entering the cloud
A prudent IT manager should consider it but they really have to look hard at the security issues – and that's much more than is my data safe. What happens if there is a breach? What happens if the company goes out of business? You really have to do a full risk analysis.
Will it work for some people? Sure. But there are many others that, despite the vendor claims, it really is not a good solution.
--Robert Rosen, CIO and past president of the mainframe user group Share
Turn to virtualization, not cloud computing, to battle hardware demand
I don't see a current demand for cloud computing other than a small number of migrations to online office and email suites (e.g. Google Apps and GMail). Most IT folks I know prefer to directly manage the majority of their infrastructure in order to provide guaranteed levels of service, reliability, and performance. Rather than outsourcing to the "cloud", services are more likely to be kept in-house and moved to virtualized hardware. This decreases costs for power and system maintenance, but at the same time retains direct control and accountability.
--Bill Bradford, senior systems administrator, SunHELP.ORG
Cloud computing: Same as it ever was
IT managers should seriously consider every option available to them. Every organization is different, and every organization changes enough over time that re-evaluation of how deploy technology has to happen regularly.
That said, I think that cloud computing users will likely always reside at either end of a bell curve. Very small organizations with no ability to create and maintain infrastructure can utilize clouds to accomplish things that would otherwise be impossible for them. Larger organizations can use clouds to expand, prototype, and scale up or down running systems.
I can not see any scenario except perhaps for very small, virtual companies who will cease running their own computing infrastructure entirely, and rely 100% on cloud resources. The risks are just too large. When you look very hard at cloud computing, it is really just a way to hide the cost of infrastructure by sharing it with others and gaining economies of scale.
Cloud resources are subject to all the same risks of any shared resource. The name is new, but the concept is as old as computing itself. Timesharing. Network computing. Shared hosting. Application service provider. Cloud computing. The technologies and names may change, but the concept remains, as does the primary risk: You have no idea what other processes, users, bandwidth, etc can impact you processes, users, bandwidth, etc.
IT managers are almost universally control freaks. Letting go of all control, which a 100% commitment to cloud computing requires, is a leap of faith that very few IT managers are going to be willing to take. However, I can see how cloud computing could be used to extend, replicate, and make resilient a core of any organizations systems, and that makes perfect sense.
The IT manager will still have their object of control: Their core servers, which could be in-house or housed at a colocation facility. Then using cloud resources to distribute load, grow or constrict on demand, geographically replicate content, etc. Essentially allow growth without capital expenditure beyond the core. That core will always have to remain within the IT manager's grasp. How else can they ensure that they can survive a storm in their choice of cloud?
The old ball and chain of legacy systems, many of which will never be able to migrate to a cloud environment, will keep that vast middle of the bell curve out of the cloud as well. Add to that list sensitive data that can not be trusted to "leave the building" be it due to regulatory, privacy, or legal liability concerns. When the data involved can cost you your job, do you really want it "out there somewhere"... or on a device you can point to when the auditor arrives?
Cloud computing provides some unique opportunities and is an interesting tool. It is not the revolution it is made out to be though. It remains a well-worn concept at the most fundamental levels, and does not represent a dramatic shift in the future of computing. It is just another option for cost deferment and mitigation.
--Chuck Goolsbee, data center executive, Digital Forest
What did you think of this feature? Write to SearchDataCenter.com's Matt Stansberry about your data center concerns at firstname.lastname@example.org.