Characterizing Application Memory Error Vulnerability to Optimize Datacenter Cost via Heterogeneous-Reliability Memory

Abstract

Abstract-Memory devices represent a key component of datacenter total cost of ownership (TCO), and techniques used to reduce errors that occur on these devices increase this cost. Existing approaches to providing reliability for memory devices pessimistically treat all data as equally vulnerable to memory errors. Our key insight is that there exists a diverse spectrum of tolerance to memory errors in new data-intensive applications, and that traditional one-size-fits-all memory reliability techniques are inefficient in terms of cost. For example, we found that while traditional error protection increases memory system cost by 12.5%, some applications can achieve 99.00% availability on a single server with a large number of memory errors without any error protection. This presents an opportunity to greatly reduce server hardware cost by provisioning the right amount of memory reliability for different applications.Toward this end, in this paper, we make three main contributions to enable highly-reliable servers at low datacenter cost. First, we develop a new methodology to quantify the tolerance of applications to memory errors. Second, using our methodology, we perform a case study of three new dataintensive workloads (an interactive web search application, an in-memory key-value store, and a graph mining framework) to identify new insights into the nature of application memory error vulnerability. Third, based on our insights, we propose several new hardware/software heterogeneous-reliability memory system designs to lower datacenter cost while achieving high reliability and discuss their trade-offs. We show that our new techniques can reduce server hardware cost by 4.7% while achieving 99.90% single server availability.

A Runtime Heuristic to Selectively Replicate Tasks for Application-Specific Reliability Targets

“…On the other hand, there is the growing body of evidence showing that selective fault-tolerance support is of key-importance to decrease the resource costs while providing the required level of reliability. For instance, Luo et al [24] and Fang et al [16] find that different applications and different phases in applications (in our case tasks) exhibit different vulnerabilities. Although neither of these works state it explicitly, it follows that selective fault-tolerance is a natural fit to achieve a reasonable trade-off between costs and the required level of reliability for different applications.…”

“…However complete task replication may be prohibitive due to the high resource cost and in fact might be excessive due to the uneven susceptibility of the different program parts to SDCs [24]. Therefore effective and efficient techniques are needed to selectively replicate tasks.…”

Designing and Modelling Selective Replication for Fault-Tolerant HPC Applications

“…On the other hand, there is the growing body of evidence showing that selective fault-tolerance support is of keyimportance to decrease the resource costs while providing the required level of reliability. For instance, Luo et al [10] and Fang et al [7] find that different applications and different phases in applications exhibit different vulnerabilities. Although neither of these works state it explicitly, it follows that selective fault-tolerance is a natural fit to achieve a reasonable trade-off between costs and the required level of reliability for different applications.…”

“…However complete replication may be prohibitive due to the high resource cost and in fact might be excessive due to the uneven susceptibility of the different application phases to SDCs [10]. Therefore effective and efficient techniques are needed to selectively replicate tasks.…”

ReSense: A Unified Framework for Improving Performance and Reliability in Multicore Architectures

“…Luo et al analyzes the memory error vulnerability for data-center applications by quantifying the applications' tolerance to soft errors in memory and proposes heterogeneous memory systems [53]. Ma et al characterizes an application based on its vulnerability to soft errors in caches at di erent level of the memory hierarchy using a fault injection methodology [54].…”