The idea of cloud computing usually suggests vast server and storage resources delivered by external providers. The benefits, we are told, are legion: you pay only for what you use, you scale up to what you need, procurement costs are minimal, operational overhead reduced, and so on. But many organisations remain concerned about …
Even those that can...
Cloud makes more and more sense every day. Increasingly we see organisations that have distributed workers (multiple offices, telecommuters and travelling workers).
Cloud services are clearly a really easy way to set up infrastructure for small orgainisations that lack the IT skills to run their own IT. But it even makes sense for those that do have the skills. Paying for a cloud service is often a lot cheaper than burning up employee time keeping the cloud running, do the backups etc.
Self-clouding probably makes sense for large organisations, but as it becomes more mainstream we're likely to see that farmed out like other services (building maintenance, cleaning, canteen etc).
Re: Even those that can...
Private Cloud == Potentially-Good
Public Cloud == (for most purposes) Bad
* If it breaks, you and your staff can not (are not allowed to) fix it, but you will probably be held responsible.
* Your data-in-the-cloud are vulnerable to sniffing, copying, and denial-of-service.
- Incompetent/inexperienced/underpaid-and-indifferent cloud-provider IT staff (minimum-wage monkeys who don't get active when they see Joe Backhoe tearing up the street outside their atacenter)
- Corrupt cloud-provider IT staff ("Hey, Mr. Cloud-Provider IT Guy/Gal! We'll give you $10,000.00 if you give us a copy of Company X's data.")
- Corrupt gov't in the place where your data are physically stored
- Armed business or gov't conflict in the place where your data are physically stored
- What are the data-protection laws (or lack thereof) in the island-nation of Vuanatu, where your data may be stored?
- Untested-by-you-and-your-staff backup/restore procedures of public cloud provider
* Costs appear to be MORE, not less than doing it yourself.
- Amazon EBS currently charges $0.10/month/GB, so 1TB == $100.00/mo.
- Current Newegg price for a Seagate Constellation ES ST31000426SS 1TB 7200 RPM SAS 6Gb/s 3.5" Internal Hard Drive with FIPS 140-2 Secure Encryption is approximately $290.00 with shipping.
- Three months of 1 TB Amazon EBS storage costs more than the drive itself!
- This ignores your cost of electricity, but ALSO EXCLUDES Amazon's charges of $0.10 per million IOPs, and monthly CPU rental, which varies depending on CPU power and instance size.
* Choosing to use public clouds is one way management says, "We hate / don't trust our IT department. We'd rather spend a pile of money on an external company than spend some money on training our IT staffers in cloud technology." (This same thought applies to outsourcing.)
* Can be *a* way for your company to afford get by if it's capital-constrained ("I can't afford to BUY a BMW, but I can afford to LEASE one.").
In light of the disadvantages listed above, and the likely-financially-shaky condition of your company if public cloud services are the only thing it can afford, you should probably look for a job elsewhere.
Re: Even those that can...
I think that is perhaps a little too simplistic; the simple truth is that different cloud models are best suited for different solutions.
For example, "your staff not being able to fix it" is countered by "we don't have the staff to do this task sufficiently to begin with (let alone the rest of the physical prerequisites)" so outsourcing it to someone who 'does it right' can make a lot of sense.
(Public, Private, Hybrid is one piece of the conversation, along with Internally vs. externally hosted.)
Not saying "the cloud" is an automatic panacea, there are concerns with any possible solution - just understand the risks, consider the benefits and make the best decision for you.
Must try it.
I just bought two Xeon-powered rackmount that work was decommissioning. I may mount them at home and load OpenStack on them to play around with.
Private Clouds are Getting small enough for all
Private Clouds are starting to make more sense as Moore's law and automation combine to make a potent combination. The recent CriKit Desktop Private Cloud is the first example of cool small Private Cloud solutions and more vendors will follow. CriKit is a compact, energy sipping Private Cloud platform that can run 32 virtual machines + on 4 nodes. That is enough compute power for a wide swath of the SMB market and it is expandable and cheap. If I can run Private Clouds on a desk in my organization and keep my data in my org to reduce risk, why would I even consider public cloud solutions? The truth is both will coexist for a wide variety of implementation requirements, but the key here is that Private Clouds no longer need to be expensive undertakings ... search on " CriKit Desktop Private Cloud " and see what I mean ..
Re: Private Clouds are Getting small enough for all
It highlights the fact that "cloud technology", be it private or public should just be considered another hammer in the tool chest, and used when it is suitable for getting the work done. Unfortunately, the never-ending cloud hype makes it difficult for decision-makers and PHBs to see through the vapor.
In my last job, we deliberately chose not to build the new infrastructure with virtualization in mind. Then some bright spark decided to check out Hyper-V. Before we knew it we were happily running 65 virtual machines (on a pool of 72 blade servers) and suddenly had more spare hardware than we knew what to do with. Of course we had to re-configure the storage, buy more memory and so on, but it was far from being a dramatic wrenchng change and actually went quite smoothly
Fact is the tools are getting better, cheaper and easier to use. Cloud architecture may very well become the first choice for any new start up, and private cloud the option par excellence for people who are too worried about security and availabilty to trust their goolies to public cloud providers.
Re: Private Clouds are Getting small enough for all
We went a same route at my present employer, but it was partially driven by the management we had at the time, which was... resistant to virtualization. In the three years that we've changed the lead IT management, we've virtualized nearly everything that's not mission critical, demands a physical machine for whatever reason (timing sensitive functions, is a CPU hog, etc.) and a few systems that we can't virtualize because there's no route *to* virtualize them (the apps running on the small handful of mini-frames primarily)
We gone from a roughly 100 or so physical servers of various sizes (mostly 1 'pizza box' servers and 2U rackmounts) to a cluster of 5-6 servers with lots of cores, memory, and fast connections to our SAN appliance. (which, I might add, is more agile then most of the vendors we work with!) Those same servers which were physical are virutalized, and we've added about 50 more virtual servers for new apps and other roles. To us it still is another tool, but one that we've taken to understand very much what it can and cannot do. It allows us to make better decisions regarding if we need to buy another physical machine for something, or take the money for that server and buy another cluster node.
Well written, thoughtful, good analysis.
We started down the path of private cloud a couple of years ago. Still only really toe-in-the-water, but it has not been too expensive and surprisingly easy to manage; very little stress in the process. There are lots of reasons for us to continue, most of which were highlighted in the article.
However, we now have a marketing manager that wants to move everything out "into the cloud" and as far as he is concerned, that means externally. Going to have to do a BOFH on him!
Migrate best practices? ITIL adapted?
Nice article, BUT... ;)
I don't think you want to migrate your management best practices, but your management processes... and implement them according to best practice (whatever that happens to be this month)
As for the need to adapt ITIL and Cobit for the cloud/virtualised environment with dynamic provisioning... well, I think anyone using it would have "adapted" it anyway even before using a private cloud.
That's even the point of those frameworks: they give a broad outline of how things should work, leaving it to you to implement. Yes, your change management in a cloud shop will need to be more efficient, but the basic premises of managing change won't... change.
Actually reading through (an earlier version of) Cobit caused me to either cry or laugh convulsively: they tended to include everything and the kitchen sink, in various levels of abstraction in their "objectives"... Ah well, time for my pills now...
Apart from that, I like the way you present the article and slice through the hype regarding cloudsand presents a sensible strategy of starting a private one...
There more to the cloud than just the computing
Very interesting and clear article. One point however that seems to be missed in these discussions is the role of the network for most enterprise IT. Many enterprises use an outsourced MPLS infrastructure for their WAN connectivity which effectively creates a private domain to distribute their IT resources. If the provider of the network builds the cloud computing into the network then the 'transition' to the cloud is as simple as either adding compute resources to their existing WAN infrastructure. Moreover the smarter providers are taking the best parts of the public cloud, flexibility, utility and speed and combining it with the inherent security of the MPLS network. Most of the providers of these services also are more enterprise oriented so understand the issues over data sovereignty etc..
- Twitter: La la la, we have not heard of any NUDE JLaw, Upton SELFIES
- China: You, Microsoft. Office-Windows 'compatibility'. You have 20 days to explain
- Apple to devs: NO slurping users' HEALTH for sale to Dark Powers
- Is that a 64-bit ARM Warrior in your pocket? No, it's MIPS64
- Apple 'fesses up: Rejected from the App Store, dev? THIS is why