I'm working on getting some materials together that can be used to pitch the effectiveness of cloud computing to support trying to convince people to move to cloud based services.  Part of doing this is to know the problems that have been seen so far and to work out good answers / alternatives to arguments supporting those.  I've listed here what I've identified so far and have explained why they are problems, but not yet the ways to counter...

• It's not secure

This is a fundamental problem.  I work with trading systems and large banks, and I don’t see them feeling comfortable with moving their data off site.  It’s not that the data isn’t secure somewhere else, as it may actually be more secure than on site, but that certain regulations (such as HIPAA, SarBox, …) may require the customer to have data on site.

This still does leave the option for computing to be done within the cloud with data stored locally, and with secure communications between on site data storage and the cloud based computing.  But this may actually lead to additional problems such as latency and not enough throughput for high volume systems.

• It can't be logged

Tying into the issue of security is being able to log access to data within the cloud.  Much of what is done for compliance is “who and when” accesses the data.  There is nothing inherent in the cloud that actually supports this, and with systems being remote, access controls to systems may not be as secure as with on site.

• It's not platform agnostic

This come up a lot in my discussions with people.  It is true to a point, but you really need to look for what you need.  Historically Amazon services only ran Linux (and because of that LAMP applications), but they have started to offer windows servers within the cloud and that has opened things up a bit, as you can use EC2 not just for prepackaged applications, but now for a base server configuration (both Linux and Windows) that you can put your own applications within.

• Reliability is still an issue

The SLA of the cloud provider will be real important to selecting a service, as well as their history of operations.  Many people have moved to using Amazon S3 for data service, but also many got burnt badly with S3 went down in 2008.  Also, moving to the cloud not only includes the reliability of the cloud provider, but also of the Internet providers connecting you to that cloud.

• Portability isn't seamless

Many cloud application providers may store your data in formats that are not portable to other platforms should you decide to move.  This is a little bit less of an issue with EC2 or Azure services as with those in many cases your data is backed with common database systems that you can easily migrate full databases to another cloud provider, but other application services such as Google or Yahoo you may not  have direct access to the databases.

• It's not environmentally sustainable

One of the touted benefits of cloud computing is the supposed savings in energy by moving your infrastructure to a cloud services provider that, at least in theory, manage the electricity running the servers better than you do.  This will likely be true if you use the cloud based applications only when needed (instead of running them all the time onsite whether you use them or not), and if the cloud provider also either turns off that server when you are not using it or allocates the power no longer used by yourself to someone else for that time.

• It still has to exist on physical servers

There are a few things that tie into this.  First, has to do with the previous comments on power.  Simply moving your app to the cloud is simply moving power usage from you to another place, so if they cloud provider does not manage this better, or perhaps even worse, you’ve not done any good.  Second has to do with the initial comments on security; you may not actually know where your data is stored.  Amazon services allow you to specify a geographic region, but something to consider is that a particular cloud provider may not, and what would the ramifications be if you data ends up hosted in a foreign country where data privacy laws are not amenable to what you need?

• The need for speed

Although you may get enough horsepower out of the cloud (if not more, and for less $$), you are introducing latency in the various communications links between you and the cloud, which can be a problem if you have applications that are very sensitive to latency.  There are solutions being developed to help manage this, including caching and ability to route to local severs, but it is still early in the game.

• Some companies already have an internal cloud

And they’ve worked hard at it, and may have operational efficiencies similar to what is available by cloud providers, so there may not be a desire to move.  I do think that this will be a rare case, especially when pointing to the use of on-demand applications provided through virtualization, as it is too early to expect that many have the ability to provide that level of virtualization in house.

• Bureaucracy

Big companies may be very resistant to moving to the cloud, not because they don’t want to, but because they have “processes” to follow, and need to be sure that systems don’t break when moved.  I would expect that you will see many small companies move to the cloud quickly, where it is easily done due to a small amount of infrastructure to move and they cost advantages are easily quantifiable, but I would think that larger organizations will first “dabble” with the cloud trying a few new and non-critical systems first before moving on existing systems.

• Management / Operations

I still see operations of the systems from the various cloud providers to be deficient.  Sometimes just knowing if an application deployed to the cloud is not available (be it the server down, network outages, firewall issues, data corruption) is a real issue.  There are companies working on this (such as RightScale), but it will be a while until good infrastructure is in place to prevent you from staying up at night wondering if you app in the cloud is up or down.

• Licensing

Not all software in the cloud is free.  Much of it is, but I still think that there is confusion on things like licensing Windows nodes in Amazon EC2.  Suppose I need 100 instances for 5 minutes each (a total of 5 minutes execution time with each running at the same time) to do some heavy calcs, but only once a month?  Do I need 100 full licenses?  With Microsoft’s Azure services this won’t be a problem as this will be a part of the bill that MS sends you, but with others like Amazon it’s not totally clear.

• Billing

I still don’t think that proper billing is in place for usage of the systems.  The scenario I mention in the issues with licensing is an actual scenario, and although I get a bill from Amazon at the end of the month showing me how much CPU I’ve used over the month, I’d sure like to see on demand how much my usage for any instance is at my demand.  Perhaps this info is available but I don’t know how to get it yet, but if it isn’t it surely should be.