.: Panel Discussion
Green database computing: should we care?
moderator:
Stavros Harizopoulos (HP Labs)
panelists:
John Cieslewicz (Aster Data)
Jignesh Patel (Univ. Wisconsin)
Daniel Abadi (Yale University)
Costs related to power and cooling in large-scale data centers are
soon expected to overtake server hardware costs. Given the growing
popularity of cloud computing and recent trends of consolidation of IT
operations in large enterprises, it is reasonable to expect that a
large portion of database management installations will move away from
dedicated servers into large-scale data centers. Pricing of such
installations and provisioning of hardware resources will need to
change accordingly to reflect the new pricing models, which in turn
will reflect power costs, and the dynamic nature of hardware resource
allocation. Up until now, database software developers and
researchers, besides exploring new functionality in the various
offerings, have mostly focused on improving performance. Should the
new focus be on improving energy efficiency instead? Or is power best
managed at the data center and platform design level?
Undoubtedly, innovations at the hardware front and cooling methods
will continue. As the energy-efficiency of individual hardware
components improves, data-center designers and operators will strive
to achieve "energy-proportionality," a mode of operation where any
given workload results into energy consumption that is proportional to
the actual fraction of resources used. But what is the role of
database software inside a well-tuned and well-managed data center?
Specifically, what is the potential of a DBMS to significantly affect
energy use by dictating hardware resource use in all data management
operations? Besides obvious choices in database setup and
configuration, database software is well equipped to perform several
hardware resource-use tradeoffs due to physical data independence and
query optimization. Does this, however, warrant a thorough
re-examination of database software choices, and should the database
systems community steer their efforts away from pure performance to
energy-efficiency instead?
In this panel, we expect our panelists to present arguments covering
the full spectrum of reactions: from solely supporting hardware-based
techniques and innovations to supporting a drastic shift in the
research agenda of the database systems community, moving away from
performance considerations to just energy-related concerns.
|