Hello,
A change in allocation has been requested:
Requester: balaji (Pavan Balaji)
Project: radix
Title: Scalable Parallel System Software
Description: This is a Computer Science request for 49,500 cpu-hours on Fusion for
development and evaluation of a family of tools and libraries with a
direct benefit to all users of high-performance computing resources.
The Radix group in Argonnne's Mathematics and Computer Science
division conducts research and development on facilities to better
enable efficient Parallel Programming, I/O, and Scientific
Understanding.
The Fusion system represents an attractive milestone for our software
development: projects start out on our breadboard cluster and can
further explore scalability in the presence of a high speed network on
Fusion before making the jump to BlueGene-scale parallelism.
Radix experiments often are concerned with latency. The infiniband
network on Fusion provides an excellent low-latency environment for
message passing and storage management research while also providing
ample bandwidth for I/O and message bandwidth testing.
For FY2013 our Messaging middleware experiments include the usual
MPICH2 development. We will study the continued improvement of various MPI
features including scalability and memory usage, performance tuning,
high-level libraries on top of MPI, fault tolerance capabilities, one-sided
communication, and collective operations. This allocation will also be used
for studying MPI extensions including active messages, and compiler support
for MPI refactoring.
We also plan for developing application-oriented libraries like ADLB, a
load-balancing library. We will port the ADLB load-balancing library to
Fusion and test its ability to function with multi-core nodes at intermediate
scale, taking advantage of the multiple cores to replace MPI communication
with local shared memory.
The NWChem library will continue to improve the performance of an
important scientific application that is widely used for science by
Argonne researchers. Specifically, we will develop inspect-executor
load-balancing techniques that will reduce network contention and
improve the overall performance and scalability of an important family
of methods (coupled-cluster methods). Furthermore, we will explore
hybrid programming models (that is, adding threading to the existing
GA/ARMCI/MPI model), particularly in the TCE module, which has been
used extensively by researchers in Argonne MSD and CSE. All the
aforementioned developments will be deployed on Fusion in a production
fashion and made available to all users. This allocation will enable
Jeff Hammond to maintain the NWChem installations on Fusion, as he has
done since the machine went live.
We work closely with application scientists, and develop application tools
under the 'radix' allocation. For one example, the KMI project defines a
high-level standardized interface for computational biology applications
relying on distributed search and match semantics for biological reads. We
will study efficient data management and data movement techniques to allow
scaling to large number of processes.
Fusion hosts two quality parallel file systems, GPFS and PVFS. The
wider HPC community often asks the Radix file system researchers to
compare these file systems head-to-head. Fusion provides us the
opportunity to do just that.
For FY 2013 our file system research extends into data storage
approaches suitable for exascale. Our research into object storage protocols,
including high performance access and replication, fits well on Fusion, with
the large amount of local storage on compute nodes. Our Exascale storage
efforts will explore scalable algorithms for use in next generation HPC
storage systems. The Fusion allocation will be used to evaluate programming
models, fault detection algorithms, synchronization primitives, and fault
tolerance strategies for a prototype object storage system.
Our NoLoss project explores how to integrate in-system storage in the
I/O software stack. Under the NoLoss project, an abstraction layer for
in-system storage was developed. The SCR checkpointing library was
modified to store checkpoints locally using this newly developed
abstraction layer. The fusion allocation will be used to explore the
performance characteristics of this new approach.
Hosting two parallel file systems allows us to do more than just
drag-race (so to speak). The two file systems have distinct
characteristics, and act differently in the broader role of the I/O
software stack. On Fusion, we can utilize a single software stack
(application, parallel-netcdf, ROMIO), and compare the impact of a
change in the underlying file system on overall application behavior.
We know from other research, for example, that the two-phase
optimization in ROMIO needs to adapt to the underlying file system,
and that aligning variables to file system boundaries can yield
performance improvements for parallel-netcdf.
In FY 2013 we are continuing research into new high level I/O
libraries. We anticipate Fusion and Surveyor will be our two main test
platforms.
Fusion compute nodes actually look fairly attractive to active storage
research. These nodes have enough disk space to store non-trivial
datasets, and powerful enough processors to carry out complex
computations without much impact on storage performance.
Radix projects also focus on deriving insight both about application
behavior as well as scientific insight. 8 core nodes make efforts
like in-situ visualization, where for example an application renders a
frame of a movie, ever more feasible. Even without fancy graphics
accelerators, the processing power on Fusion nodes makes such analysis
and visualization feasible, especially if the visualization can take
advantage of multiple cores and multiple nodes. Naturally, the radix
visualization tools have demonstrated scalability to Fusion sizes and
beyond on Blue Gene, so making full utilization of Fusion should not be
a concern.
As should be evident from the variety of planned experiments, the
Radix group plans to make full use of Fusion.
With Argonne employees and student collaborators, we expect around 20
members in FY2013.
We plan on about 0% of our jobs being single core jobs. It's crazy
that you even have to ask!
While Fusion may contain a modest number of compute nodes, the machine
still represents an attractive platform for testing: very fast, low
latency interconnect; 8 cores per node; .25 TB of storage per compute
node. The testing and experiments the Radix group can carry out on
Fusion ensures the tools the group develops today will remain relevant
even as supercomputers grow. Quite a few radix-developed projects are
part of the Fusion software stack: CPU hours for this project yield
improvements and benefits not just for the radix group, but for all
users of Fusion and indeed users of high-end computational resources
worldwide.
Current: undetermined amount
Justification:
Requested: 150000
A specific reason has been given:
We have initiated some new projects within the runtime system stacks that are requiring more core-hours than what we originally estimated. We are currently at a -63K balance, and we plan to have some reservations for paper deadlines as well (which cost double the core-hours).
This needs to be approved and the final allocation amount decided upon.
Thank You,
The LCRC Accounts System
