Page History

...

Abstracts

Paul Gibbon

...

Energy-efficient scheduling

Jean Utke

Designing and implementing a tool-indedendent, adjoinable MPI wrapper library

The efficient computation of gradients by the "adjoint-mode" of algorithmic differentiation (AD) entails the inversion of MPI communication graphs. The logic to be implemented for adjoining non-blocking communication patterns is sufficiently complex to warrant  a design of components that is independent of the algorithmic differentiation tool that provides the context in which the adjoint communication is to take place. We discuss (i) how we account for the different data models  implied by the AD  tool as well as the target language, (ii) the implementation choices among the possible adjoint communications, and (iii) the currently known limitations of our approach. We hope for feedback from the community regarding this design particularly  with respect to performance and current developments in the MPI standard.

Laurent Hascoet

In this talk, I will survey recent works on energy-efficient scheduling. The goal is to minimize the energy consumption of a schedule, given some performance constraints, for instance a bound on the total execution time. I will first revisit the greedy algorithm for independent tasks in this context. Then I will present problems accounting for the reliability of a schedule: if a failure may occur, then replication or checkpoint is used to achieve a reliable schedule. The goal remains the same, i.e., minimize the energy consumption under performance constraints

Jean Utke

Designing and implementing a tool-indedendent, adjoinable MPI wrapper library

The efficient computation of gradients by the "adjoint-mode" of algorithmic differentiation (AD) entails the inversion of MPI communication graphs. The logic to be implemented for adjoining non-blocking communication patterns is sufficiently complex to warrant  a design of components that is independent of the algorithmic differentiation tool that provides the context in which the adjoint communication is to take place. We discuss (i) how we account for the different data models  implied by the AD  tool as well as the target language, (ii) the implementation choices among the possible adjoint communications, and (iii) the currently known limitations of our approach. We hope for feedback from the community regarding this design particularly  with respect to performance and current developments in the MPI standard.

Laurent Hascoet

The adjoint of MPI one-sided communications
Computing gradients of numerical models by the adjoint mode of algorithmic differentiation is a crucial ingredient for model optimization, sensitivity analysis, and uncertainty quantification of many large-scale science and engineering applications. The adjoint mode implies a reversal of the data The adjoint of MPI one-sided communications
Computing gradients of numerical models by the adjoint mode of algorithmic differentiation is a crucial ingredient for model optimization, sensitivity analysis, and uncertainty quantification of many large-scale science and engineering applications. The adjoint mode implies a reversal of the data dependencies and consequently a reversal of communications in parallelized models. Building on previous studies regarding the adjoining of MPI two-sided communications, we investigate the construction of adjoints for certain one-sided MPI communications

...

Investigating the probability distribution of false negative failure alerts in HPC systems

As large parallel systems increase in size and complexity, failures are inevitable and exhibit complex space and time dynamics. Several key results have demonstrated that recent advances in event log analysis can provide precise failure prediction. The state-of-the-art in failure prediction provides a ratio of correctly identified failures to the number of all predicted failures of over 90\% and its able to discover around 50\% of all failures in a system. However large part of failures are not predicted and considered as false negative alerts. Therefore, developing  efficient fault tolerance strategies to tolerate failures requires a good  perception and understanding of failure prediction properties and characteristics. In order to study and understand the properties and characteristics of the false negative alerts, we conduct in this paper a statistical analysis to discover the probability distribution of such alerts and their impact on fault tolerance techniques. To this end we study failures logs from different HPC production systems. We show that: (i) surprisingly the false negative distribution has the same nature as the failure distribution; (ii) after adding failure prediction we were able to infer statistical models that describes the inter arrival time between false negative alerts and so current fault tolerance can be applied on these systems; (iii) the current failures traces contain a high amount of correlation between the failure inter arrival time that can be used to improve the failure prediction mechanism. Another important result is that checkpoint intervals can still be computed from existing first order formula when failure distribution is purely random.

...

Adaptation algorithms for HPC applications can improve performance but their implementation is often costly in terms of development and maintenance. Component models such as Gluon++, which is built on top of Charm++, propose to separate the business code, encapsulated incomponents, and the application structure, expressed through a component assembly. Adaptation of component-based HPC applications can be achieved through the optimization of the assembly. We have studied such an approach with the adaptation to network topology and data size of a gluon++ 2D FFT application. In this talk, we present our work thus far and comment preliminary experimental results on the Grid'5000 platform.

Setefan Wild

Stefan Wild

Loud computations? Noise in iterative solvers

 The adjoint of MPI one-sided communications
Roundoff errors, discretizations, numerical solutions to systems of equations, and adaptive techniques can destroy the smoothness of processes underlying computations at scale. Such computational noise complicates optimization, sensitivity analysis, and other applications that depend on the simulation output. We present a method for analyzing computational noise and illustrate the insights it enables on a collection of problems based on Krylov solvers.

Guillaume Aupy

On the Combination of Silent Error Detection and Checkpointing 

In this talk, we revisit traditional checkpointing and rollback recovery strategies, with a focus on silent data corruption errors. Contrarily to fail-stop failures, such latent errors cannot be detected immediately, and a mechanism to detect them must be provided. We consider two models: (i) errors are detected after some delays following a probability distribution (typically, an Exponential distribution); (ii) errors are detected through some verification mechanism. In both cases, we compute the optimal period in order to minimize the waste, i.e., the fraction of time where nodes do not perform useful computations. In practice, only a fixed number of checkpoints can be kept in memory, and the first model may lead to an irrecoverable failure. In this case, we compute the minimum period required for an acceptable risk. For the second model, there is no risk of irrecoverable failure, owing to the verification mechanism, but the corresponding overhead is included in the waste. Finally, both models are instantiated using realistic scenarios and application/architecture parameters.

Dries Kimpe

Triton: Exascale Storage

In this talk, I will present a status update of our work on Triton, a newly designed exascale era storage system.  In addition to Triton specific information, the presentation will also include a brief discussion about the tools and techniques that help us in implementing and designing Triton. One such tool is the use of discrete event simulation to quickly evaluate algorithms at scale before implementing them in Triton.

Tatiana Martsinkevich

On the feasibility of message logging in hybrid hierarchical FT protocols

abstract: Hybrid hierarchical fault tolerance protocols are a promising solution for providing fault tolerance on large scale. A hybrid hierarchical protocol combines coordinated checkpointing and message logging. A lot of work has been done on more efficient implementation of checkpointing protocols, however there are some questions that stay not fully studied with regards to message logging. Message logging requires some portion of the process memory and logged data is flushed to a safe storage together with the checkpoint. There are several possible strategies to take in case where there is not enough memory to log messages between two checkpoints. Each of the strategies will be discussed.with regards to message logging. Message logging requires some portion of the process memory and logged data is flushed to a safe storage together with the checkpoint. There are several possible strategies to take in case where there is not enough memory to log messages between two checkpoints. Each of the strategies will be discussed.

Di Sheng

Optimization of Google Cloud Task Processing with Checkpoint-Restart Mechanism

Programming multicore or manycore architectures is a hard challenge particularly if one wants to fully take advantage of their computing power. Moreover, a hierarchical topology implies that communication performance is heterogeneous and this characteristic should also be exploited. We developed two load balancers for Charm++ that take into account both aspects depending on the fact that the application is compute-bound or communication-bound. This work is based on our TreeMatch library that computes process placement in order to reduce an application communication cost based on the hardware topology. We show that the proposed load-balancing scheme manages to improve the execution times for the two classes of parallel applications.

Guillaume Mercier

Topology Management and MPI Implementations Improvements

Modern hardware architectures featuring multicores and a complex memory hierarchy raise challenges that need to be addressed by parallel applications programmers. It is therefore tempting to adapt an application communication pattern to the characteristics of the underlying hardware. The MPI standard features several functions that allow the ranks of MPI processes to be reordered according to a graph attached to a newly created communicator. In this talk, we explain how the MPI implementation of the MPI_Dist_graph_create function was  modified to reorder the MPI process ranks to create a match between the application communication pattern and the hardware topology. The experimental results on a multicore cluster show that improvements can be achieved as long as the application communication pattern is expressed by a relevant metric. We also show several areas in MPI implementations where similar techniques can be beneficial.

Eddy Caron

Seed4C: Secured Embedded Element and Data privacy for Cloud Federation

In this talk we introduce the design of a secure federated cloud from end to end. We discussed the core of this platform based on a High Performance Computing middleware that uses federated clouds and other virtual resources as well classic HPC resources. We propose an architecture to ensure a high level of security from personal devices to the targeted virtual machine. The Seed4C platform improved security in each layer. With DIET Cloud, we are able to deploy a large-scale, distributed and secure HPC  platform that spans across a large pool of resources aggregated from different providers through a secure way

Jonathan Rouzaud-Cornabas

SimGrid Cloud Broker: Simulation of Public and Private Clouds

Before migrating an application to public Clouds, it is required to evaluate its performance. Doing so on a real Cloud has a time and money cost. By using simulation, it is  possible to evaluate the migration without money cost and with a reduced time cost. Furthermore, it ables to test different resource reservation and application allocation  algorithms without paying for these resources. The same is true when using private and/or  hybrid Clouds. SimGrid Cloud Broker (SGCB) ables to easily simulate a whole Cloud public and/or private  to evaluate an application. Moreover, due to high modularity, SGCB can also be used to evaluate the  inner working of a Cloud middleware. Accordingly, it is possible to evaluate the impact of new VM to PM placement algorithms and virtual machine image deployment policies. To conclude, SGCB is a general  purpose simulator for evaluating applications that run on public and private Clouds and their compositions.

Frederic Hecht

FreeFem++, a user language to solve PDE.

I will make a small overview of the capability of FreeFem++. and I will focus on four  computer science problems: 

-the design of the language:

--- from store mathematical formulation (the weak form of PDE) to Data Structure (DS),  

---  from DS  to matrix and right hand side.

-the way to  use lots of third party software : like : MUMPS, IPOPT, TETGEN, MKL, ....

-the use of mesh adapted

-the parallelization with MPI.

Ian Foster

Compiler optimization for distributed dynamic data flow programs

Distributed, dynamic data flow is an execution model well-suited for many large-scale parallel applications, particularly scientific simulations and analysis pipelines running on large, distributed-memory clusters. In this paper we describe compiler optimization techniques and an intermediate representation for distributed dynamic data flow programs. These techniques are applied to Swift/T, a high-level declarative language that allows flexible data flow composition of functions written in other programming languages such a C or Fortran. We show that compiler optimization can reduce communication overhead by 70-93% on distributed memory systems, making the high-level language competitive with hand-coded coordination logic for certain common application styles

Christian Perez

On Component Models to Deploy Application on Clouds

Clouds have become a complex ecosystem, providing many kinds of virtual machines (with different capabilities), of usage (on demand, spot instances, reservation), of data storage, etc. Moreover, some clouds provides worldwide "regions", enabling large scale distributed applications. Users also have very different requirements, potentially from execution to another such as minimizing execution time, respecting budget constraints, etc. Therefore, automatically and efficiently deciding how to map an application to a set of VM is a difficult challenge. This talk will discuss how the European PaaSage project as well as the French ANR MapReduce are using component models to describe and map an application structure, independently of anycloud, to an actual cloud

Kate Keahey

Research Topics and Collaboration Opportunities in the Nimbus Team

The advent of IaaS cloud computing promises acquisition and management of customized on-demand resources. What is the best way to leverage those resources? What new applications are emerging in this context? How will they change our work patterns? What new technical approaches need to be developed to support them? What new opportunities will they lead to? In this talk, I will describe tools the Nimbus team is developing, among others, in the context of the Ocean Observatory Initiative project, that focus on answering these questions. I will describe  our approach and tools, the problems we are trying to address, as well as the interaction patterns associated with scientific applications currently driving our approach.

Abdou Guermouche

Towards resilient parallel linear Krylov solvers

The advent of exascale machines will require the use of parallel resources at an unprecedented scale, probably leading to a high rate of hardware faults. High Performance Computing (HPC) applications that aim at exploiting all these resources will thus need to be resilient, i.e., being able to still compute a correct solution even in presence of faults. In this work, we investigate possible remedies in the framework of the solution of large sparse linear systems that is often the inner most numerical kernel in many scientific and engineering applications and also one of the most time consuming part. More precisely, we present recovery followed by restarting strategies in the framework of Krylov subspace solvers where lost entries of the iterate are interpolated to define a new initial guess before restarting. In particular, we consider two interpolation policies that preserve key numerical properties  of well-known solvers. We assess the impact of the recovery method, the fault rate and the number of processors on the robustness of the resulting linear solvers. We consider experiments with CG, GMRES and Bi-CGStab.