Thursday, 14 December 2017

Session 2: Performance Models and Techniques - Part 2

An Automatic Trace Based Performance Evaluation Model Building for Parallel Distributed Systems

Authors:

Ahmad Mizan (Carleton University)
Greg Franks (Carleton University)

Abstract:

Performance models can be built at early stages of software development cycle to aid software designers to assess design alternatives and identify fundamental design pitfalls before the implementation phase starts. These models are flexible for varying operational conditions and design alternatives; however, their creation is not trivial and requires considerable efforts. This paper addresses this problem by introducing automation in process of Layered Queuing Network (LQN) performance model creation for traces of events generated from instrumented software programs in the nodes of a distributed parallel software application. The event-traces are created based on a new timestamp format, which is independent of physical time and uses extremely low count elements. A set of post-mortem methodologies have been introduced to identify the interactions between the service nodes of the parallel distributed software application and determine their workload activities, while supporting concurrent executions in the nodes. It can capture Forward, Asynchronous, Synchronous and loops of Asynchronous or Forward interactions. The final result is a framework of methodologies, specifications and tools which is appropriate for model-based performance evaluation parallel distributed software applications.

DOI: 10.1145/1958746.1958760

Full text: PDF

[#][]

Hierarchical Performance Measurement and Modeling of the Linux File System

Authors:

Hai Nguyen (University of Arkansas)
Amy Apon (University of Arkansas)

Abstract:

File systems are very important components in a computer system. File system simulation can help to predict the performance of new system designs. It offers the advantages of the flexibility of modeling and the cost and time savings when utilizing simulation instead of full implementation. Being able to predict end-to-end file system performance against a pre-defined workload can help system designers to make decisions that could affect their entire product line, affecting several million dollars of investment. This paper presents a detailed simulation-based performance model of the Linux ext3 file system. The model is developed using Colored Petri Nets. A performance study using the model shows that the obtained results are close to the expected behavior of the real file system. The model shows that file system parameters have significant impact on the performance of the I/O when compared to the parameters of the disk subsystem.

DOI: 10.1145/1958746.1958761

Full text: PDF

[#][]

Real-World Performance Modelling of Enterprise Service Oriented Architectures: Delivering Business Value with Complexity and Constraints

Authors:

Paul Brebner (NICTA / ANU)

Abstract:

Performance and Scalability Modelling of real-world enterprise systems is challenging due to both the complexity and size of the system being modelled, and constraints imposed by real projects such as the need to provide business value, deadlines, and the accessibility, relevance, quality and quantity of available documentation and performance data. Our hypothesis is that enterprise Service Oriented Architectures (SOAs) are more amenable to performance modelling as services are more granular, visible, and measurable. Since 2007 we have developed, trialled and refined a method with model-driven tool support for directly modelling the performance and scalability of increasingly complex Service Oriented Architectures. This paper reports an illustrative experience modelling a large-scale production SOA Enterprise Service Bus (ESB) upgrade, focussing on lessons learnt related to the complexity and constraints of modelling in the real-world. The key observations are that model construction is a type of theory formation and therefore: (1) Models (functioning as theories) can be simple but powerful enough to model large complex SOAs within the boundaries of real project constraints; (2) Model formation can be incremental, starting with a simple model (as simple theories are easier to refute) and refining as required; (3) Building multiple competing models can be a useful approach if information is inadequate or ambiguous, as the rival models can be tested with the aim of discarding incorrect ones; (4) If insufficient information is available to build a single "über" model to answer all the performance questions, it is often possible to build multiple specialised models for different purposes.

DOI: 10.1145/1958746.1958762

Full text: PDF

[#][]

Correct Router Interface Modeling

Authors:

Krzysztof Rusek (AGH University of Science and Technology)
Lucjan Janowski (AGH University of Science and Technology)
ZdzisBaw Papir (AGH University of Science and Technology)

Abstract:

The aim of this paper is to determine how to model router interface in order to accurately predict packet drops. There is an enormous amount of research on traffic models reported, however, a model of router interface has not gained proper consideration yet.

Our experiments reveal that an incorrect model of the router interface can result in a significant disparity between drop probabilities measured on a physical interface and derived from a trace-driven simulation study. In this paper an accurate model of the router interface (Cisco IOS-based routers with a non-distributed architecture) was presented.

DOI: 10.1145/1958746.1958763

Full text: PDF

[#][]

A PMIF with Petri Net Building Blocks

Authors:

Catalina M. Lladó (Universitat de les Illes Balears)
Peter G. Harrison (Imperial College London)

Abstract:

Performance model interchange formats (PMIFs) support the portability of models and sharing of solutions amongst different tools. XML-based interchange formats have been defined for the interchange of queueing network and Petri net models, amongst others, but there is still scope to extend their application to multiple formalisms, in particular beyond queueing networks. We extend an existing PMIF to hybrid models by including a new type of node, called a "building block", defined as a certain class of Petri nets. The synchronisation primitives of these building blocks can be used to specify fork-join systems whilst, under certain conditions, retaining product-form solutions when embedded in queueing (or other) networks possessing this property already. When a product-form does not exist, the whole network is translated into a Petri net and solved either by simulation or direct solution of the underlying Markov chain by an existing analyser. Finally, we apply the extended PMIF to model a computer system with RAID storage.

DOI: 10.1145/1958746.1958764

Full text: PDF

[#][]