Tuesday, 12 December 2017

Session 12: Performance Modelling and Prediction II

Impact of Data Locality on Garbage Collection in SSDS: A General Analytical Study

Authors:

Yongkun Li (University of Science and Technology of China)
Patrick P. C. Lee (The Chinese University of Hong Kong)
John C. S. Lui (The Chinese University of Hong Kong)
Yinlong Xu (University of Science and Technology of China)

Abstract:

Solid-state drives (SSDs) necessitate garbage collection (GC) to erase data blocks and reclaim the space of invalidated data, and GC inevitably introduces additional writes due to data relocation. The performance of GC, which is quantified by cleaning cost or write amplification, is critical to the overall performance of SSDs. However, characterizing GC performance is complicated by the general implementations of GC algorithms and the complex data locality characteristics of real-world workloads. This paper presents a general analytical study to characterize the performance impact of data locality on a general family of GC algorithms. We develop probabilistic models to address two fundamental issues: (1) What is the impact of data locality on the performance of locality-oblivious GC? (2) How can data locality be leveraged to improve the performance in locality-aware GC? We further conduct extensive trace-driven simulations on real-world workloads to validate the findings of our models.

DOI: 10.1145/2668930.2688036

Full text: PDF

[#][]

A Framework for Emulating Non-Volatile Memory Systems with Different Performance Characteristics

Authors:

Dipanjan Sengupta (Hewlett-Packard Labs & Georgia Institute of Technology)
Qi Wang (Hewlett-Packard Labs & The George Washington University)
Haris Volos (Hewlett-Packard Labs)
Ludmila Cherkasova (Hewlett-Packard Labs)
Jun Li (Hewlett-Packard Labs)
Guilherme Magalhaes (Hewlett-Packard)
Karsten Schwan (Georgia Institute of Technology)

Abstract:

Exponential increase of online data and a corresponding growth of data-centric applications (Big Data analytics) forces system architects to revisit assumptions and requirements of the future system design. New non-volatile memory (NVM) technologies, such as Phase-Change Memory (PCM) and HP Memristor offer significantly improved latency and power efficiency compared to flash and hard drives. Many future systems are expected to have both DRAM and NVM. This can radically change system and software design, and enable new style of Big Data processing applications. However, the commercial unavailability of new NVMs technologies and uncertainty of their performance characteristics make it difficult to assess new system software stacks and to study their performance impact on future workloads. To bridge this gap and encourage an early design phase, we are building a DRAM-based performance emulation platform1 , called NVMpro, that leverages features available in commodity hardware, to emulate different latency and bandwidth characteristics of future NVM technologies. NVMpro enables an efficient and accurate emulation of a wide range of NVM latencies and bandwidth characteristics for performance evaluation of emerging byte-addressable NVMs and their impact on applications performance without modifying or instrumenting their source code.

DOI: 10.1145/2668930.2695529

Full text: PDF

[#][]

Towards a Performance Model Management Repository for Component-based Enterprise Applications

Authors:

Andreas Brunnert (fortiss GmbH)
Alexandru Danciu (fortiss GmbH)
Helmut Krcmar (Technische Universität München)

Abstract:

This work introduces a Performance Model Management Repository (PMMR) for component-based enterprise applications. A PMMR is a central server that allows managing performance model components in corporate environments. A key challenge when using performance models in such environments is to distribute, update and maintain them. Especially, when software components represented in performance models are under the control of different teams in an organization. Additional problems arise as soon as release cycles for their components are not synchronized. A PMMR helps to address these challenges by introducing a central repository in which different performance model component versions can be managed and maintained. Such capabilities support the collaboration of distributed teams as they can manage their performance model components independently from each other. Performance models of specific component versions can be combined into one performance model as required for the current performance evaluation. We propose to build such a PMMR using the capabilities provided by the Palladio Component Model (PCM) as meta-model and the EMFStore as underlying versioning repository.

DOI: 10.1145/2668930.2695526

Full text: PDF

[#][]

Automated Reliability Classification of Queueing Models for Streaming Computation

Authors:

Jonathan C. Beard (Washington University in St. Louis)
Cooper Epstein (Washington University in St. Louis)
Roger D. Chamberlain (Washington University in St. Louis)

Abstract:

When do you trust a model? More specifically, when can a model be used for a specific application? This question often takes years of experience and specialized knowledge to answer correctly. Once this knowledge is acquired it must be applied to each application. This involves instrumentation, data collection and finally interpretation. We propose the use of a trained Support Vector Machine (SVM) to give an automated system the ability to make an educated guess as to model applicability. We demonstrate a proof-of-concept which trains a SVM to correctly determine if a particular queueing model is suitable for a specific queue within a streaming system. The SVM is demonstrated using a micro-benchmark to simulate a wide variety of queueing conditions.

DOI: 10.1145/2668930.2695531

Full text: PDF

[#][]