The Large Hadron
Collider at CERN(the European Centre for
Nuclear Research) in Geneva,which was switched back onon the 21
Novemberfollowing an electrical fault which saw the particle
collider switched off in September 2008,
isdesigned to create hundreds of millions of collisions
between subatomic particles every second, in
an attempt to advance knowledge of energy and matter, but most
significantly to piece together the events immediately after the
big bang.
Four experiments spaced along an underground tunnel the size of
the London city line will produce 15 petabytes of data per year,
creating one of the greatest challenges ever in the history of
computing, requiring that data be collected, formatted, stored,
shared and secured on a massive scale.
On site is a staggering concentration of heavy-duty computing
power, which includes 5,700 systems, 36,600 processing cores,
41,500 disk drives, 45,000 tape cartridges and 160 tape drives. But
this only begins to touch on the effort needed to support the
project.
When
the Large Hadron Collider accelerator at CERN is running at
full strength, access to experimental data needs to be provided for
over 10,000 scientists in several hundred research institutes and
universities worldwide participating in the Large Hadron Collider
(LHC) experiment. All the data need to be available over the
15-year estimated lifetime of the LHC. Analysis of the data,
including comparison with theoretical simulations, needs about
100,000 CPUs at current measures of processing power.
CERN's IT department head, Frédéric Hemmer is helping to drive
an ambitious project to develop huge grid computing networks that
would support the LHC and many other important research
initiatives throughout Europe and rest of the world.
Frédéric Hemmer's former colleague,
Tim
Berners-Lee, of course invented the web. Now his and
CERN's work on grids is itself leading to powerful and interesting
innovations leading to major changes in computing and
communications; most notably the move towards the cloud.
The Worldwide Large Hadron Collider Grid
(WLCG) is a collaboration
between more than 170 computing centres in 34 countries. It uses
both the Open Science
Grid (OSG) in the US as well as the
Enabling Grids for E-Science
(EGEE), the largest grid in Europe.
As well as particle physics, these grids support research into
fields as diverse as climate change, medicine and computer
graphics. Hemmer and a number of his peers are presently pushing
for a more formalised grid organisation that would link most of the
universities and research institutes throughout Europe.
This is one of the goals of the European Grid Initiative
(EGI) which in addition to
fostering technical collaborations is also trying to help member
countries to resolve legal differences and agree on frameworks and
standards for privacy and security.
"We are now trying to have something more permanent that would
rely more on national efforts," Hemmer explains. Key to this has
been the creation of independent National Grid Institutes (NGIs)
which represent the research groups and universities of each member
country. NGIs are entities with a public mission aiming to
integrate funding resources at national level for the provision of
grid-based services. They are designed to be a one-stop-shop for a
number of common, grid-based services for national research
communities.
As of 2007 the EGI had 36 supporting countries. By the time the
organisation officially comes into effect next year, it is hoped
that it will play a central role in ensuring a sustainable future
for the e-infrastructure developed under the series of EGEE
projects.
The massive task of data management and resources allocation for
CERN and the many other organisations linked to it through various
grid architectures has demanded some innovative solutions on the
middleware front. In collaboration with EGEE, CERN and other groups
have worked to develop what is seen as the next generation of
middleware for grid computing. Born from the collaborative efforts
of more than 80 people in 12 different academic and industrial
research centres as part of the EGEE Project,
gLite provides an
open-source framework for building grid applications tapping into
the power of distributed computing and storage resources across the
Internet.
Glite effectively offers researchers two tiers of service. Grid
Foundation middleware covers things like security infrastructure
information, monitoring and accounting systems, and access to
computing and storage resources with the aim of delivering a
consistent and reliable production infrastructure. Higher-level
grid middleware covers services like job management, data
catalogues and data replication.
It is expected that these sorts of open source middleware
solutions emerging in the academic realm will increasingly inform
the development of applications in the corporate environment
running on Linux and other platforms in the future.
Security
Naturally security is a major concern with grids, especially, as
is often the case, where breaches of sensitive research data could
be very embarrassing or damaging.
In the cloud environment, many make the argument that service
providers are at pains to deliver better security, lest their own
commercial reputations suffer. In the grid environment, on the
other hand, the drivers are somewhat different, as well as
constituting a different approach to the management and
distribution of information
"When we look at reliability and security of grids, the problem
is not the grid itself; the difficulty is in securing the sites,"
says Hemmer. To address these concerns, the EGEE is conducting a
wide ranging security trial across several of its members' sites in
an effort to tighten the system. "The efforts we are putting in
will ensure that the sites are secure and able to react to security
incidents."
Strong security is of course essential for there to be the sort
of widespread collaboration envisaged by the EGI.
In the cloud environment, on the other hand, it's not the
sharing information that causes concern, rather it's the fact that
businesses usually don't know where their data is being kept.
A key difference between cloud and grid computing is that grids
are designed to foster collaboration between users while users in
the cloud are invisible to each other.
The way in which resources are allocated is also different, with
cloud services structured much like utilities whereby the user
simply pays for how much they use. Users of grids on the other hand
typically make sporadic, yet, very large requests for
resources.
However, analysts note that the lines between the two are
beginning to blur as researchers look to bring more flexibility to
grid infrastructures, while in the cloud businesses are beginning
to discuss opportunities for collaboration. "Now cloud computing
has come into focus it's starting to blur the difference between
grid and cloud computing," explains
Vuk Trifkovic,
senior analyst with
Datamonitor. "Some of the
cloud platforms in the future will be built in a similar way to
grid computing, like consortia. In time, with cloud computing we'll
see a much higher degree of collaboration."
He adds: "As the technology stabilises through improved
international standards and reliability, grids are gaining
acceptance in mainstream business and science communities." CERN's
Hemmer observes a trend towards industry-specific cloud platforms
which operate much the same way as grids, only that the users still
essentially rent the infrastructure and applications from a service
provider. The benefit of this is that things like capacity and
security can be more easily guaranteed.
However, it is expected that the pricing and business models
currently in use in the cloud will look very different in years to
come, as more industry- and application-specific services emerge
and companies demand more flexibility.
Conversely, various grid architectures around the world have
begun to incorporate certain features of cloud computing. In fact
CERN and others are investigating the feasibility of layering cloud
services from Amazon and
others over the grid.
Hemmer explains however that there are considerable
sensitivities associated with doing this,
particularly with regard to data being held in dispersed or unknown
locations. "This is something that some funding agencies for
science do not like; they prefer to install in their own
country."
Nevertheless as CERN and thousands of researches around the
world prepare for the relaunch of the massive LHC project, it's
somewhat ironic that the vast grid networks of which so much of its
success depends may at times find themselves leaning on the very
cloud services they essentially gave birth to.