Biomedical Engineering Reference
In-Depth Information
The phrase “cloud computing” is often applied to both remote network storage
and computer virtualization. Cloud computing can help to overcome some of
the barriers that have inhibited the collaborative sharing of tools and methods
in bioinformatics. Although many bioinformatics projects are open-source, the
diffi culties in replicating analysis platforms often prevent the use of these open
source resources.
Many academic bioinformatics projects suffer from a common problem:
Support for many academic bioinformatic projects is often sporadic. A project
may be undertaken under the fi xed term funding of a grant or as part of
the training program of a graduate student or postdoctoral fellow. During the
grant or training period, there are resources to develop and maintain the
resource. This often culminates in the release of the resource to the research
community. Over time, the research community discovers the utility of the
resource and comes to incorporate it into their research workfl ow and thus
becomes dependent on it. Meanwhile, one or more of the following may occur:
the grant funding ends, the graduate student completes his or her degree or
the postdoc obtains a job, and the project ends while the resources that it
generated become orphaned. Often projects are left to coast until software
incompatibility creeps up or hardware issues or budget/space constraints lead
to the server hosting the resource to be taken offl ine. At that point there is
often a lack of will and/or resources to repair or re-create the tool and the
research community is left without access to what has become an important
part of their workfl ow.
This has been one of the strong arguments for the release of these types of
tools as open-source projects. In theory this means that the end users could
download the source and replicate the resource on their own. Functionally this
is not always easy. Many of these tools are not a simple executable fi le but
rather a core of code enmeshed in a web of Web servers, databases, and other
system resources. Although these dependencies usually are also open source
as well, the integration of them is often diffi cult and not well described or
documented. It is sometimes the case that tools require specifi c versions of
these components and will not work when the components are upgraded. If
the project is still active and with the detailed knowledge of the person who
developed the system still present, these problems can be overcome. In con-
trast, rebuilding this “house of cards” from the ground up can be nearly impos-
sible even for individuals with signifi cant IT backgrounds. A possible solution
to maintaining analysis infrastructure is to use cloud computing systems such
as Amazon's Web Services (AWS) that allow users to save snapshots of virtual
computers as Amazon machine images (AMIs). These AMIs can be made
publicly available and can be used by anyone with an Amazon AWS account.
The end user can call the preconfi gured computer into existence for the
amount of time required and shut it off when done. Like a fl y caught in amber,
the AMI is a snapshot of the working system and remains in a static state and
is never tied to a particular piece of hardware. Since the effort required to
generate an AMI is not prohibitive and the storage costs for these images are
Search WWH ::
Custom Search