No abstract

Projections and measurements of error rates in near-exascale and exascale systems suggest a dramatic growth, due to extreme scale (10 9 cores), concurrency, software complexity, and deep submicron transistor scaling. Such a growth makes resilience a critical concern, and may increase the incidence of errors that "escape", silently corrupting application state. Such errors can often be revealed by application software tests but with long latencies, and thus are known as latent errors. We explore how to efficiently recover from latent errors, with an approach called application-based focused recovery (ABFR). Specifically we present a case study of stencil computations, a widely useful computational structure, showing how ABFR focuses recovery effort where needed, using intelligent testing and pruning to reduce recovery effort, and enables recovery effort to be overlapped with application computation. We analyze and characterize the ABFR approach on stencils, creating a performance model parameterized by error rate and detection interval (latency). We compare projections from the model to experimental results with the Chombo stencil application, validating the model and showing that ABFR on stencil can achieve a significant reductions in error recovery cost (up to 400x) and recovery latency (up to 4x). Such reductions enable efficient execution at scale with high latent error rates.

We consider the use of non-volatile memories in the form of burst buffers for resilience in supercomputers. Their cost and limited lifetime demand effective use and appropriate provisioning. We develop an analytic model for the behavior of workloads on systems with burst buffers, and use it to explore questions of cost-effective provisioning, and missiondirected allocation of burst-buffer (SSD) lifetime.First, our results show that system efficiency can be increased by as much as 14% by considering a global perspective (workload mix, job size) for SSD lifetime allocation. Second, with size-based and system-efficiency based lifetime allocation, large jobs suffer as much as 40% job efficiency loss; job-efficiency based allocation must increase their allocations by 50% to eliminate this disparity. Finally, further results suggest that underprovisioning SSD lifetime (only 10-20% of the "optimum" as defined by per-job requirements without resource constraint) is sufficient to produce 90% system efficiency at failure rates three times that of current systems.

No abstract

Exascale studies project reliability challenges for future HPC systems. We present the Global View Resilience (GVR) system, a library for portable resilience. GVR begins with a subset of the Global Arrays interface, and adds new capabilities to create versions, name versions, and compute on version data. Applications can focus versioning where and when it is most productive, and customize for each application structure independently. This control is portable, and its embedding in application source makes it natural to express and easy to maintain. The ability to name multiple versions and ''partially materialize'' them efficiently makes ambitious forward-recovery based on ''data slices'' across versions or data structures both easy to express and efficient. Using several large applications (OpenMC, preconditioned conjugate gradient (PCG) solver, ddcMD, and Chombo), we evaluate the programming effort to add resilience. The required changes are small (\ 2% lines of code (LOC)), localized and machine-independent, and perhaps most important, require no software architecture changes. We also measure the overhead of adding GVR versioning and show that overheads \ 2% are generally achieved. This overhead suggests that GVR can be implemented in large-scale codes and support portable error recovery with modest investment and runtime impact. Our results are drawn from both IBM BG/Q and Cray XC30 experiments, demonstrating portability. We also present two case studies of flexible error recovery, illustrating how GVR can be used for multi-version rollback recovery, and several different forward-recovery schemes. GVR's multi-version enables applications to survive latent errors (silent data corruption) with significant detection latency, and forward recovery can make that recovery extremely efficient. Our results suggest that GVR is scalable, portable, and efficient. GVR interfaces are flexible, supporting a variety of recovery schemes, and altogether GVR embodies a gentleslope path to tolerate growing error rates in future extreme-scale systems.

No abstract

No abstract

No abstract