Closing the Gap between Cloud Providers and Scientific Users - Cloud Computing with e-Science Applications - page 120

Some important developments for private cloud infrastructures and scientific
workflow integration, such as Opal2 [14] and SciCumulus [15], have been made.
Under this approach, a researcher is responsible for wrapping a scientific appli-
cation in a preconfigured virtual machine (VM) or script. Although packaged
VMs can be deployed automatically according to user requirements, admini-
stration problems arise right away when the number of supported applica-
tions increases and constant updates are necessary. This can be the case, for
example, with the Scientific Computing as a Service (SCaaS) project [16].
There is also some work on an infrastructural level. Infrastructure-as-a-
service (IaaS) solutions like OpenNebula [17, Eucalyptus [18], PiCloud [19],
and Nimbus [20] offer a configuration environment especially designed for
common scientific requirements. In this case, scientists who want to use
these kinds of solutions need to be able to properly install and configure
their own applications. There are also some upper-level commercial offer-
ings like Cyclone [21] or SBGenomics [22], for which users can have access,
in a SaaS model, to some commonly used applications like Hmmer [23],
BLAST (Basic Local Alignment Search Tool) [24], or Gromacs [25]. These
projects were built with some general needs in mind, making customization
a complex process that depends on personal contact with the suppliers.
Different projects have focused on benchmarking conventional scientific
solutions and workflows in both private and public cloud environments.
Some studies have shown, for example, that a typical configuration in an
IaaS provider like Amazon EC2 can be significantly slower than a modern
high-performance system, especially when it comes to communication [2].
Despite this, it has been shown that research teams will adopt cloud com-
puting over the next few years; in the meanwhile, cloud providers will likely
improve their offering over important factors like costs, networking, admin-
istration, and elasticity [26].
Finally, it is worth mentioning some SaaS solutions that developed inter-
esting models at an enterprise level. Among the most important ones, sales-
force [27], ZOHO [28], and SuccessFactors [29] allow a wide variety of users
to access complete business functionalities with low effort and at minimum
costs. This way, small- and medium-size companies can benefit from solid
solutions that fit their budget. The e-Clouds proposal is based on an integra-
tion of the ideas developed under some of these projects to meet the scientific
requirements mentioned previously.
6.4 e-Clouds Architecture
6.4.1 Overview
e-Clouds is an effort to create an easy-to-use SaaS marketplace for scientific
applications. As part of the initial proposal, the e-Clouds team will be in