Cloud Computing Challenges

The challenges of cloud computing have been well documented over the past few years – below is a summary of these challenges as perceived by end users. Note that, while many of the challenges below are true technology barriers, others are simply perception issues that companies will need to address internally before a cloud solution can be adopted.

In any case, any viable cloud solution will need to address these Top Challenges:

• Ability to seamlessly transition from Private to Public Clouds

  Seamless transition from the machines in the corporate datacenter to the machines in the public cloud environment is the most important operational requirement for utilizing public/external clouds. When extending compute capacity to the public cloud, customers demand the same simplified setup, configuration and tear-down of virtual compute resources within the public cloud that they are used to in their corporate datacenters. Cloud vendors must therefore provide a clear path, both technically and from a business perspective, in getting from private to public cloud computing.

• Security

  • User Authentication and Authorization: In public clouds, safeguards must be placed on machines to ensure proper authentication and authorization. Within the private cloud environment one can track, pinpoint, control and manage users who try to access machines with improper credentials.
  • Data Security: In order to take advantage of public clouds a corporation must transfer the data and applications from their private cloud to the public cloud in order for their jobs to run successfully. Most companies are concerned about security of their corporate data in the public cloud environment. A key obstacle to cloud computing is the security of the data while being transported to the public cloud and while inside the cloud. Most public cloud providers do not guarantee the security of the data while being transported to the public cloud or while inside the cloud.
  • Data Transport: Due to the relative infancy of public clouds, most customers are not clear on how the data is transferred to the public cloud environment, between public and private clouds, or within the public clouds. Most companies have the following data transport concerns with extending their cluster into a public Cloud: Network speed/latency/isolation/quality; bandwidth charges for data transfer; and speed of the data transport from the private cloud to the public cloud.
  • Data Persistency: Another critical factor with-respect-to data security is data persistency. Companies need to contractually ensure that this corporate data is really deleted when it is deleted.

• Application Performance

  There are technical factors limiting the application performance in the public cloud environment. The most obvious performance-limiting factor is the virtualization penalty: Most applications and workloads in a cloud computing environment run 20 – 50% slower than the physical compute infrastructure. This performance toll varies depending upon a variety of factors including application type, workload type, cloud provider (virtualized infrastructure or similar infrastructure to your internal datacenter), and hardware (Intel's Nehalem processor offsets the performance penalty with significant performance gain over previous generation processors). This performance penalty can also be offset by pre-planning peak need against time value.

• Memory Limits

  Today, public cloud providers have a 15.5 GB limit on available memory on the machines within the public cloud environment. This limit is the by-product of virtualization. For customers used to running specific workload on large memory machines, this 15.5 GB memory limit could pose a problem. It is up to the virtualization vendors e.g. VM Ware and Citrix to increase this 15.5 GB memory limit in the near future.

• Software Licensing

  Customers cannot finalize software licensing agreements with large software vendors for the use of their software in the cloud computing environment. The fact remains that may ISVs still license their software on a per-node, per-CPU or per-core basis and are still wrestling with the software licensing in a virtual environment. In addition, system integration issues where the various services require different information and translation.