Critical Enterprise 2020: A Research Agenda for Distributed IT
Infrastructures

Karsten Schwan
CERCS Research Center/College of Computing
Georgia Institute of Technology

Today's enterprise systems and applications implement functionality that
is critical to the ability of society to function. An example are the
operational infrastructures used by large corporations like our nation's
airlines, where 24/7 uptime requirements are coupled with the need to
respond to surges in demand, the need for live updates or upgrades of
the distributed programs and systems, and the need to protect data from
disasters. The complexities of applications, technologies, and systems
inherent in such large distributed systems are leading us to propose the
`Critical Enterprise 2020' Grand Challenge. `Critical Enterprise 2020'
will methodically address criticality with the full set of technologies
Computer Science can bring to bear, ranging from topics in program
verification and testing, to hardware-based methods for component
protection or isolation, to research in operating systems that seeks to
better protect or manage critical system assets, to middleware that
combines the best of today's flexibility or agility inherent in
enterprise software infrastructures with the lessons learned from years
of developments in high performance and grid computing, to trust models
and mechanisms applied to information management to focus
systems on critical vs. non-critical actions and data.

Academic research should explore topics not open to industry, which is
forced to be standards-compliant or compatible with existing hardware or
applications: (1) '2020' should focus on `vertical' transparency rather
than horizontal layering. That is, future critical distributed systems
should not only permit but encourage the integrated and application-specific
use of hardware, operating systems, and middleware. Sample functionality
includes the dynamic extension of network infrastructures with enterprise
knowledge (e.g., business rules for data routing and distribution) and
the ability of middleware to be `aware' of underlying resources, their
current abilities and constraints. The desired outcome is the
application-specific management of system resources. (2) '2020' also
indicates that humans will not be able to cope with the innate
complexities of future distributed systems. Self-management,
self-explanation, and similar system properties are therefore key
elements of future solution approaches. (3) `Criticality' provides the
metrics by which research should be evaluated, focusing on the ability to
meet changeable end user constraints instead of merely improving
system-level parameters.

To summarize, the challenge is to make methodical improvements across
the entire `vertical' stack of hardware to applications. Toward these
ends, research should be complemented by community efforts that create
an open source ecosystem future researchers can use to study realistic,
complex applications and systems. This `Critical Enterprise Facility'
(CEF) could leverage proven technologies like EmuLab and PlanetLab for
providing external access and use, but its additional focus should be on
applications that must operate with strong quality of service or end
user utility requirements, running on systems representative of future
enterprise-scale hardware. Another unique characteristic of CEF should
be its end-to-end nature, reaching all the way from server systems to
the mobile end devices that will be routine in future complex IT
infrastructures.