Managing Tail Latency in Datacenter-Scale File Systems Under Production Constraints

Misra, Pulkit A.; Borge, María F.; Goiri, Íñigo; Lebeck, Alvin R.; Zwaenepoel, Willy; Bianchini, Ricardo

doi:10.1145/3302424.3303973

Cited by 32 publications

(7 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Parameter z is a function of the workload and it will be explained shortly. The global function f : R N → R is the sum of the cost function (7) of each node v i . The main goal of the nodes is to allocate the jobs in order to minimize the cost function in a distributed fashion, by communicating with their neighbors only.…”

Section: Optimization Problemmentioning

confidence: 99%

“…Solving a scheduling optimization problem in such a large-scale system is challenging due to the size of the network and the dynamic nature of resource requirements of incoming and existing workloads. Furthermore, due to unexpected cluster changes as nodes randomly fail and/or abnormal runtime behaviors due to software or configuration faults and resource contention, latency variability is introduced into the network [1], [7]. To this end, we posit a novel scheme that takes in account these potential latency variations in the form of explicit delays in the communication links during planning, while still remaining a asynchronous in its operation and we guarantee that it will converge in finite-time.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

CPU Scheduling in Data Centers Using Asynchronous Finite-Time Distributed Coordination Mechanisms

Grammenos,

Charalambous,

Kalyvianaki

2021

Preprint

View full text Add to dashboard Cite

We propose an asynchronous iterative scheme which allows a set of interconnected nodes to distributively reach an agreement to within a pre-specified bound in a finite number of steps. While this scheme could be adopted in a wide variety of applications, we discuss it within the context of task scheduling for data centers. In this context, the algorithm is guaranteed to approximately converge to the optimal scheduling plan, given the available resources, in a finite number of steps. Furthermore, being asynchronous, the proposed scheme is able to take in account the uncertainty that can be introduced from straggler nodes or communication issues in the form of latency variability while still converging to the target objective. In addition, by using extensive empirical evaluation through simulations we show that the proposed method exhibits state-of-the-art performance.

show abstract

Section: Optimization Problemmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

CPU Scheduling in Data Centers Using Asynchronous Finite-Time Distributed Coordination Mechanisms

Grammenos,

Charalambous,

Kalyvianaki

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…The main reason for tail latency in RAID-enabled SSDs is that diferent RAID components (e.g. SSD channels) have uneven busyness on I/Os and GCs while running user applications [18]. Especially with respect to the issue of mitigating the negative efects of garbage collection that is the heaviest operation in SSDs, I/O requests on the GC target channels will be fulilled by reading the data on other channels of the same stripe with certain XOR computations [15,19,20].…”

Section: Introductionmentioning

confidence: 99%

Degraded Mode-benefited I/O Scheduling to Ensure I/O Responsiveness in RAID-enabled SSDs

Sha

Cai

et al. 2022

ACM Trans. Des. Autom. Electron. Syst.

View full text Add to dashboard Cite

RAID-enabled SSDs commonly have unbalanced I/O workloads on their components (e.g. SSD channels), as the data/parity chunks in the same stripe may have varied access frequency, which greatly impacts I/O responsiveness. This paper proposes a I/O scheduling scheme by resorting to the degraded read mode and the read-modify-write mode, to reduce the long-tail latency of I/O requests in RAID-enabled SSDs. The basic idea is to avoid scheduling read or update requests to the heavily congested but targeted RAID components. Such requests are satisfied by accessing other relevant RAID components by certain XOR computations (we call the degraded modes ). Specially, we build a queuing overhead assessment model on the top of factors of data redundancy and the current blocked I/O traffics on SSD channels, to precisely dispatch incoming I/O requests to be fulfilled with the degraded mode or not. The trace-driven experiments illustrate that the proposed scheme can reduce the long-tail latency of read requests by 23.1% on average at the 99.99th percentile, in contrast to state-of-the-art scheduling methods.

show abstract

“…The development of new flash memories such as 3D-stacked charge-trap (CT)-based ones largely benefits the storage density of modern SSDs. Meanwhile, they show some new physical characteristics, e.g., the increased block size and layer speed variation, the effect of which on performance have not been fully investigated [ 9 ].…”

Section: Introductionmentioning

confidence: 99%

Observation and Optimization on Garbage Collection of Flash Memories: The View in Performance Cliff

Liu

Gao

et al. 2021

Micromachines

View full text Add to dashboard Cite

The recent development of 3D flash memories has promoted the widespread application of SSDs in modern storage systems by providing large storage capacity and low cost. Garbage collection (GC) as a time-consuming but necessary operation in flash memories largely affects the performance. In this paper, we perform a comprehensive experimental study on how garbage collection impacts the performance of flash-based SSDs, in the view of performance cliff that closely relates to Quality of Service (QoS). According to the study results using real-world workloads, we first observe that GC occasionally causes response time spikes, which we call the performance cliff problem. Then, we find that 3D SSDs exacerbate the situation by inducing a much higher number of page migrations during GC. To relieve the performance cliff problem, we propose PreGC to assist normal GC. The key idea is to distribute the page migrations into the period before normal GC, thus leading to a reduction in page migrations during the GC period. Comprehensive experiments with real-world workloads have been performed on the SSDsim simulator. Experimental results show that PreGC can efficiently relieve the performance cliff by reducing the tail latency from the 90th to 99.99th percentiles while inducing a little extra write amplification.

show abstract

Managing Tail Latency in Datacenter-Scale File Systems Under Production Constraints

Cited by 32 publications

References 32 publications

CPU Scheduling in Data Centers Using Asynchronous Finite-Time Distributed Coordination Mechanisms

CPU Scheduling in Data Centers Using Asynchronous Finite-Time Distributed Coordination Mechanisms

Degraded Mode-benefited I/O Scheduling to Ensure I/O Responsiveness in RAID-enabled SSDs

Observation and Optimization on Garbage Collection of Flash Memories: The View in Performance Cliff

Contact Info

Product

Resources

About