Description
A number of European projects are collaborating to better understand
the value cloud computing could bring to today's researchers. To
succeed with this there is a need to learn, by testing, what is
possible to do today, and the implications these services has on
economy, security, vendor lock-in, and last but not least the
usability and quality.
To succeed with this objective, international collaboration is key,
and in this session six collaborating projects share their findings
and roadmaps.
Topics in this session covers ongoing studies of the use of open
source softwares as well as commercial products, and their
combinations. One common goal is to identify the value in cloud
computing to HPC users; end-user interfaces provided by cloud
technologies and identify possible added value to today's grid and
standard HPC interfaces. Concretely this amounts to e.g.
- Investigate if a machine provisioning interface like that of
Amazon's EC2 coupled with a library of scientific appliances is of
value to European academic research communities. - Investigate how to build a cloud storage infrastructure that
supports the use of clouds for scientific applications. - Identify the economic value of cloud computing
- Understand the economic implications of moving parts of HPC onto
cloud infrastructures. More notably by:- Estimate the fraction of current HPC compute jobs that could
realistically be executed on clouds - Estimate the cost of building and operating a cloud facility
of the same capacity as the above percentage of the total current
resources - Estimate possible cost reductions achievable by scaling out
peak loads to commercial clouds - Compare cost of a private cloud facility with a commercial cloud
- Estimate savings in scenarios with
- Commercial clouds when aiming at direct peering (data transfer cost)
- Allocating their spare capacity at a reduced price
- Estimate the fraction of current HPC compute jobs that could
During the OGF28 a 2 hr 25 min ECEE workshop was conducted, with a
common guiding document, and collaboration web site, was produced
(web: www.scientific-cloud.org).
This guiding document will be updated during the workshop. Outline of
current document:
1.Roadmaps
a.Sharing roadmaps
b.Avoiding ‘built in’ interoperability problems later on
-
Use cases
a.Gap analysis
b.“Market analysis” – today’s users, tomorrow’s
c.Guidelines – best practices, quick start one-pager + checklist
(cloud, grid, hpc, something else?), rules of thumb -
Offering
a.Compared to existing (and evolving) public services – what do/can we offer?
b.Service levels – how to handle, in practice
c.Some mentioned last mile, customer support (application specific
needs), platforms for specific users (like rendering example from
BalticCloud) -
Focus areas
a.Security – access and ID mgmt in line with local requirements
b.Metering (what), Accounting, Billing, Business models
c.Federation of clouds (cf grids)
d.Network
e.Licenses
f.Scheduling, load balancing (resource sharing, application correlation)
g.Technical solutions studied –list of tested solutions, their pros and cons -
Trends of small/medium DCIs and cloud computing
a.Will they change their mode of operation?
6.GAP analysis
a.Identified overlaps
i. E.g. compare project’s current deliverables
b.Identified gaps
i.What projects did not include into the plan
ii.What is posing the main problem for the projects. E.g. lack of
user communities, grid legacy
c.Identify already existing policies, best practices and
recommendations, and technology from the grid community – to be used
as is, or modified
