Conference Proceedings: 2012

KEYNOTE SESSION: Web Performance, The Big Picture
Pat Meenan, Google, Inc.

An overview on web performance from the end user's perspective. What it is, how to measure it and changes in the browsers and standards that you should be aware of.

Download handouts

Introduction to Wavelets and their Application for Computer Performance Trend and Anomaly Detection
Dima Seliverstov, BMC Software

In this session, I will present a technique to identify trends and anomalies in Performance data using wavelets. I will answer the following questions: Why use Wavelets? What are Wavelets? How do I use them?

Download paper
Download handouts

Webhosting in the Cloud. Cheaper and Smarter.
Peter van Eijk

Cloud services give us a lot of web hosting opportunities these days. We will discuss experiences with a number of alternative hosting approaches in the cloud. This includes measurements of their uptime, performance and scalability. We will also present some cost models for cloud based webhosting. If time permits we can have a look at how cloud providers work to make webhosting so affordable, allowing us to paraphrase Dolly Parton’s famous line: 'It takes a lot of money to make webhosting this cheap'.

Download paper
Download handouts

CMG-T: Getting Started in (z/OS) Capacity Planning
Ray Wicks

This two part session is an introduction to the concepts required in performance analysis and Capacity Planning. It is designed for professionals entering or just getting started in the field. Emphasis is placed on large processor systems and examples will be largely drawn from z/OS but the concepts apply to all operating systems and hardware. The tutorial is organized to review the architecture where appropriate (albeit briefly), discuss any appropriate Performance Analysis principles and then delve into CP principles and practices.

Download paper

Batches in Bunches: Finding the Critical Path of Large Batch Windows
Chris Papineau, Oracle

Large companies are complex to run, and the Enterprise Resource Planning (ERP) software packages they implement can entail running several thousand batch programs in a single day. Analysis and optimization of these batch windows present a significant performance analysis challenge. This session presents a case study in which a batch window of over 1000 individual jobs is analyzed for its critical path using some very simple techniques to build Gantt Bars which allow a dynamic, visual analysis of the window. A custom application is presented, which automates the building the Gantt Bars.

Download paper
Download handouts

Significance of Replacement Queue in Sequential Prefetching
Elizabeth Varki, University of New Hampshire

The performance of a prefetch cache is dependent on both the prefetch technique and the cache replacement policy. Both of these algorithms execute independently of each other, but they share a data structure - the cache replacement queue. The replacement queue captures the impact of prefetching and caching. This session shows that even with a simple prefetch technique, there is an increase in hit rate when the LRU replacement queue is split into two queues.

Download paper

Hybrid Performance Modeling using Layered Queueing Networks.
Greg Franks, Carleton University
Lianhua Li, Carleton University

The layered queueing network model has been very useful for solving performance models of distributed systems with client-server interactions. Fast analytic solutions exist for a broad range of models. However, models which contain elements which cannot be solved using approximate Mean Value Analysis (MVA) must instead be solved using simulation or another approach. In this session, a hybrid solution strategy is used where MVA is used for part of the solution, and another solver, in this case simulation, is used for the remainder.

Download paper

CMG-T: Getting Started in (z/OS) Capacity Planning
Ray Wicks

This two part session is an introduction to the concepts required in performance analysis and Capacity Planning. It is designed for professionals entering or just getting started in the field. Emphasis is placed on large processor systems and examples will be largely drawn from z/OS but the concepts apply to all operating systems and hardware. The tutorial is organized to review the architecture where appropriate (albeit briefly), discuss any appropriate Performance Analysis principles and then delve into CP principles and practices.

Download paper

Storm Chasing - Troubleshooting in the Cloud
Stanford White III, Siemens Medical Solutions

Take a storm chasing expedition in the cloud, and experience the often unappreciated complexities of moving from Kansas to Oz , while circumventing the dangers lurking along the Yellow Brick Road to success. This user story, from a cloud provider’s perspective, will take the reader on a journey of discovery through the design and implementation of a newly hosted application, as well as, a perilous race to find an elusive and dangerous creature hidden in the cloud. Always pay attention to the man behind the curtain!

Download paper
Download handouts

Introduction into Enterprise Applications Capacity Planning Using Queuing Models
Leonid Grinshpan, Oracle, Corp

The session demonstrates the basics of a methodology of enterprise application capacity planning field tested over many years. The methodology minimizes customer’s investment in hardware and at the same time ensures sufficiency of infrastructure capacity to support acceptable application performance. The majority of time will be dedicated to live demonstration of queuing models building and solving by using a commercial model solver hosted on presenter’s laptop. A free open source model solver will also be demonstrated.

Download paper

Application Invariants: Finding Constants Amidst All the Change
Richard Gimarc, CA Technologies
Adrian Johnson, CA Technologies

This session presents a method for deriving and utilizing Application Invariants. An Application Invariant is a metric that quantifies the behavior or performance of an application in such a way that its value is immune to changes in workload volume. Several sample Application Invariants are developed and presented. One of the primary benefits of an Application Invariant is that it provides a simple (flat) shape that can readily be used to track changes in application performance or behavior in an automated manner.

Download paper

Social Media for the Performance Professional - or Why Blog and Tweet and How Not to get in Trouble
Elisabeth Stahl

Social Media for the Performance Professional - or why blog and tweet and how not to get in trouble doing it. This session will highlight social media types, producers and consumers, tips for getting started, and social media rules of thumb.

Download paper
Download handouts

Better Information from Better Data Visualization
Nicole Arksey, INETCO Systems Ltd
Scott Chapman, American Electric Power

As performance analysis and capacity planners, we collect data to share with co-workers and management in order to explain and justify our technical capacity and performance recommendations. To display this data, we often create the easiest graphs and don’t consider how they are perceived. This session discusses ways to visualize your data and describes scenarios in which these visualizations should be used to transform data into useful information. We discuss common ‘mistakes’ and ways to improve your charts to better get your point across and better present your intended message.

Download paper
Download handouts

Late Breaking: vSphere vs. Hyper-V
Dale Feiste, Metron-Athene

A high level discussion is provided for the vSphere and Hyper-V architectures. The metrics available for each are discussed, and the results of benchmarks compared.

Download paper
Download handouts

Time-Series: Forecasting and Regression: 'And' or 'Or'?
Alex Gilgur
Josep Ferrandiz, PayPal
Matthew Beason, PayPal

At CMG’11, I had a fascinating discussion with Dr. Igor Trubin. We talked about Uncertainty, Second Law of Thermodynamics, and other high matters in relation to IT. That discussion prompted this session. We propose a method to get better predictions when we have a forecast of independent variable and a regression. It works for any scenario where performance can be linked with business metrics. A real-world example is worked through that demonstrates how this technique works to improve the performance metric prediction and highlight trends that would have been overlooked otherwise.

Download paper
Download handouts

Late Breaking: A Note on Knee Detection
Josef Ferrandiz
Alex Gilgur

The discussion of and around the knee is ongoing in CMG, as well as in other forums. Opinions range from complete denial to admission that there is a knee, but the transition is so smooth that is no practical way to detect it. This paper demonstrates that the truth is in between: there is a knee, and there is a way to detect its low end both in the post-mortem analysis and in the live system, when it has just begun (assuming there are enough data to do so).

Download paper

CMG-T: CMG-T z/OS Tuning Basics (Part 1): Understanding Your WLM Service Policy
Glenn Anderson, IBM

This basic session works through the elements of a WLM Service Policy. How do the four goal types - Average Response Time, Percentile Response Time, Velocity and Discretionary - operate, and how do you set goals for your workloads? What exactly is Importance, and why is it a key to an effective policy? What is the purpose of period migration? How do you classify work into Service Classes? Finally, how does WLM use the Performance Index (PI) to distribute resources to your loved ones based on the goals you set? The answers to all these questions and more, as this session helps you discover how to manage z/OS workloads with the power of WLM.

Download handouts

Risk Economics: How Quantifying Operational Risk Leads to Better IT Investments
Dennis Wenk, Hitachi Data Systems

We live in a technology driven world. IT is deeply embedded in the operating fabric of the modern organization and is now highly dependent on IT. Simultaneously, and quite unintentionally, IT introduced exposures that have seeped into every layer of the organization. Organizations are more likely to experience a large loss from IT interruption than from any disaster or ‘black swan’ event. Risk economics provides the cause-and-effect linkage that is deficient in the traditional approaches and improves business decisions relative to investments that reduce the risk of an IT interruption.

Download paper
Download handouts

Implementing Capacity Management According to ISO20000 Standards
Will Otten, De Nederlandsche Bank NV

This session deals with how we implement CAPACITY MANAGEMENT within our organization, according to ISO20000 standards. The idea for this presentation was born at the CMG'11 while talking to Adam Grummitt. This session explains how the standards helps us to realize our goal, why we choose ISO20000 instead of ITIL. The examples cover storage, shared storage, hardware and VMware.

Download paper
Download handouts

CMG-T z/OS Tuning Basics (Part 2): Analyzing Key RMF Performance Data
Glenn Anderson, IBM

Interesting and useful performance data abounds throughout that pile of RMF reports sitting on your desk! The secret it knowing where to find it, and understanding what it is telling you. This session will focus on several of the key reports generated by the RMF Post Processor, such as the CPU Activity Report and the Workload Activity Report, combing through tons of data fields to find the messages that your z/OS system is sending to you.

Download handouts

Choice of Critical Flows : Enable Effective Risk Based Testing
Abhijit Ranjekar, Infosys

Enterprise solutions today have multiple applications cascading further into huge numbers of individual transactions. IT managers need to select 'critical entities' to focus the testing efforts upon – in particular for performance tests. This session describes a novel methodology to facilitate the decision of choosing the right set of applications-flows for testing. The methods as described lead to selection being based on scientific evaluation rather than rules of thumb or perception of individuals, allows flexibility and transparency to involve business stakeholders in deciding thresholds.

Download paper

Using Capacity Modeling to Support Server Consolidation: A Case Study
Richard Gimarc, CA Technologies
Byron Dorman, Moody's Shared Services, Inc.

This session describes a case study where capacity modeling was used to support the consolidation of a number of AIX LPARs to x86 servers. Capacity modeling was used as the first step in the consolidation process to determine feasible target environments. This session will walk through the process of profiling the original LPARs and then using what if modeling scenarios to evaluate a feasible target infrastructure. It will be shown that capacity modeling provides valuable initial insight that can help to guide, confirm and solidify decisions made during the server consolidation process.

Download paper

Big Transaction Data - Accelerating Business Growth
Mike Hammond, OpTier

Your goal is to quantify how IT impacts the business so that you can optimize resources and improve profitability. To do this, you capture and store massive amounts of data from applications that run the company. But because these applications are complex and distributed, and without a real sense of each other, the data cleansing that you must conduct often becomes a bottleneck - requiring data warehousing teams and IT impact analysis - and will often force you to abandon the project and proceed on gut instinct. Could you simplify the entire process? With Big Transaction Data you can!

Download handouts

Real World Enterprise Workload Modeling- How Much of a Representation is Really Enough?
Priyanka Arora

The debate between Performance teams and Stakeholders over requirements is as old as the performance discipline itself. While technically, the benefit from load testing every transaction does not justify the effort; for a business, even a small risk of one transaction crashing the system is a concern. This session discusses how to strike a balance by creating holistic workload models that closely represent real-world enterprise load patterns and analyzing risks and mitigation of selective modeling early in the project. It also lists considerations that go into defining the performance test scope.

Download paper

CMG-T z/OS Tuning Basics (Part 3): Exploring the World of zEnterprise Hybrid, Blades, and the zManager
Glenn Anderson, IBM

The IBM zEnterprise hybrid system introduces the Unified Resource Manager, allowing customers to manage a collection of one or more zEnterprise nodes including any optionally attached BladeCenter Extensions (zBX) as a single logical virtualized system using a Hardware Management Console (HMC). For the typical mainframer, that sentence is quite a mouthful! The HMC now provides a single point of control through which data center personnel can deploy, configure, monitor, manage and maintain the integrated System z and zBX blades based on heterogeneous architectures in a unified manner. Let’s explore this new world of blades and virtual servers and how to manage them, in this exciting session!

Download handouts

Virtualization and Capacity and Performance Management Techniques
Ron Kaminski

Is Capacity Planning and Application Performance Management even needed any more in a virtualized firm? Does virtualization hurt application performance? What are the optimum ways to deliver great application performance on modern shared infrastructures? How to take advantage of the virtualization process to make changes that improve application performance?

Download paper
Download handouts

Software Performance Antipatterns for Identifying and Correcting Performance Problems
Connie Smith, Performance Engineering Services
Lloyd Williams, PerfX

Performance antipatterns document common software performance problems as well as their solutions. These problems are often introduced during the architectural or design phases of software development, but not detected until later in testing or deployment. Solutions usually require software changes as opposed to system tuning changes. This session covers five performance antipatterns and gives examples to illustrate them. These antipatterns will help developers and performance engineers avoid common performance problems.

Download paper

Best Inhouse Techniques to Optimize Memory Footprint
Tejaswini Rentala, CA Technologies
Muruganadam Somasundaram, Ca Technologies
Venkata Krishna Bezawada, CA Technologies

Optimization of memory footprint has always been a fascinating proposition. Despite rapid strides in technology resulting in lower memory costs, it continues to be a limiting resource in any computer. While good software programming skills can't be substituted, this session will share a number of easily applied ''in-house'' techniques to optimize the memory footprint.

Download paper

Top 10 Mantras for a CDE (Catalog Driven Engagement) Model for IT Service Management
Arvind Raman, Infosys Ltd.
Thothathri V, Infosys Ltd.

In the search for dramatic transformation, IT organizations world wide are virtualizing and consolidating their datacenters. While this exercise receives good attention, there is little change in the day-to-day delivery of services, resulting in continued poor experience for users. (PARA)During this session, the speakers will share best practices, or 'mantras', for driving a Service Catalog Driven Engagement (CDE) Model that has helped organizations achieve up to 30% productivity improvement. The speakers will also share checklists, how-tos and guides on setting up and achieving this.

Download paper
Download handouts

AIX frame and LPAR level Capacity Planning. User Case for Online Application
Igor Trubin, IBM

The session shares some challenges the Online Banking Capacity Management team had and overcame during the Solaris to AIX migration. The raw capacity estimation model was built to estimate AIX frames capacity needs. The Capacity Planning process was adjusted to a virtualized environment. The essential system, middleware and database metrics to monitor capacity were identified; business driver correlated forecast reports were built to proactively tune entitlements; IT-Control Charts were created to establish dynamic thresholds for Physical Processors and IO usage. A Capacity Council was established.

Download paper
Download handouts

CMG-T: Java Performance Analysis and Tuning
Peter Johnson, Unisys

Attendees at these CMG-T sessions will benefit from my many years of doing Java performance tuning, both in our lab where we run industry standard benchmarks, and in our application excellence centers where I have helped tuning our customer’s real-world applications. I will cover the following topics: 1. Analyzing the garbage collector (GC). a) Understanding how the GC works; b)Gathering GC data (there are three different formats for this data); c)Graphing the GC data and understanding what the graphs mean; d) Examining some real-world examples, what the GC graphs showed, what tuning was done, what the results were (hint: significantly improved performance). 2. Survey of various GC algorithms; a) Understanding how the GC works; b) Gathering GC data (there are three different formats for this data); c) Graphing the GC data and understanding what the graphs mean; d) Examining some real-world examples, what the GC graphs showed, what tuning was done, what the results were (hint: significantly improved performance). 3. Miscellaneous JVM tuning.

Download paper

What Does CMG Want to be when it Grows Up, Now its Fourth Age is Underway?
Adam Grummitt, Metron

CMG has been around a long time, but not many seem sure of its purpose in life. Members ask ‘What does CMG want to be when it grows up?’ or 'Is CMG a Membership organization or a Conference organizer' or 'What are the real benefits of being a member?' All find little objective detail in the replies. Assuming that CMG may have a lifespan of three score years and ten, it is now well into its ‘fourth age’ of seven ages of man as described by William Shakespeare, so its purpose and objectives should be clear. This paper considers the issues that affect CMG and the options for change.

Download paper

Measuring the Visual Experience of Website Performance
Patrick Meenan, Google, Inc.

There are a lot of technical measurements for measuring the performance of a website but they are not very good at expressing the actual user experience. This session will introduce a visual measurement technique that we have been using at Google that better represents the user experience as well as several use cases and real-world examples.

Download paper
Download handouts

SSDs as a type of Cache Memory
Bruce McNutt, IBM
Lee LaFrese, Intellimagic

Since Spring 2010, a number of storage vendors have introduced mechanisms by which it is possible to deploy SSDs as part of a tiered structure. In these offerings, the storage system provides automated, dynamic relocation of data, based upon its current observed level of I/O demand. The data relocation process is not, however, driven directly by individual I/O operations. By contrast, this paper explores the use of SSDs as a true ''on demand'' cache. We explore a specific approach, called FLash Enabled Architecture (FLEA), and its performance characteristics.

Download paper
Download handouts

A More Robust Regression Methodology to Estimate the Parameters of Super-serial Scalability Law
Jayanta Choudhury

A methodology to estimate the parameters of Super-serial Scalability Law (SSL) was first proposed in 2001 by Dr. Neil Gunther. SSL was substituted by Universal Scalability Law (USL) later. SSL has displayed better performance when USL fails. Some theoretical errors of the first methodology along with a new methodology circumventing the theoretical errors were reported recently by Dr. Jayanta Choudhury. More improvements to that methodology to perform with noisy input are reported here, and effectively applied to example noisy performance data.

Download paper
Download handouts

Driving Capacity Management up the 'Value' Chain
Ramapantula Udaya Shankar, Tata Consultancy Services Limited

In a difficult economy that has imposed severe restrictions on budgets, IT is increasingly being asked to do more with less. This creates a platform for the organization's capacity management function to demonstrate 'value' from all the data and analysis that has been happening to support the business. The author describes how the various business, application and infrastructure metrics that are collected by monitoring probes in the IT landscape can be used to identify opportunities for optimization thus reducing the TCO and improve ROEI on the existing infrastructure.

Download paper

CMG-T: Model and Forecasting Basics: Modeling and Forecasting Basics
Michael Salsburg, Unisys

Although most computing environments are heterogeneous, computer system modeling is, in most ways, platform neutral. The same techniques and tools can be used to model zSeries, Unix / Linux, and Windows. At the heart of these models is the essential queueing network. This course provides the details of the essential queueing network, including the necessary statistics that need to be collected from the system, as well as various modeling techniques that yield insights that cannot be gleaned from observing the actual computer system. Once the model is validated, it can be used to explore 'what-if' scenarios where either the workload or the underlying configuration can be changed in the model so that the resulting service levels can be observed. If time permits, an additional section on the subject of time series estimation and forecasting will be presented. This course will not teach you everything you need, but it will give you a full survey of the various approaches with a full bibliography for future reference. Session 1 - This is the first of three sessions. Computer performance modeling is mainly focused on understanding how business activity and the infrastructure can be analyzed to understand the impact on IT services. The key to this activity is to understand how requests for service queue for usage of resources. This first session provides basic definitions, the history of modeling queueing systems and some basic analytical queueing models

Download paper

Building Calendars with Proc Report
Frank Bereznay
Raff Rushton

Web based calendars are a very intuitive way to organize and present large sets of chronologically related reporting objects. Each day or period of time becomes a link to the related content. Cell attribution, such as background colors (traffic lighting) and or pop-up values, can also be used to enhance the information content for a calendar cell. This session will show how SAS Proc Report can be used to create these types of calendars.

Download paper
Download handouts

SEDS-Lite: Using Open Source Tools (R, BIRT, MySQL) to Report and Analyze Performance Data
Igor Trubin, IBM

Statistical Exception Detection (SEDS) is one of the variations of learning behavior based performance analysis methodology previously developed, and implemented. This session takes the output of the main SEDS tools – IT-Control Chart and Exceptions (Anomalies) Detector - and shows how they could be built by Open Source type of BI tools (such as R, BIRT and MySQL) or just by spreadsheet. The session includes source codes, tool screen-shots and report input/output examples to allow reader building/developing a light version of SEDS.

Download paper
Download handouts

Quantitative Evaluation of SPEC Component-Level Benchmarks
Jozo Dujmovic, San Francisco State University
Wilson Kwok, Model N

In this session, we present a systematic evaluation of the most popular SPEC benchmark suites. Our goal is to show that SPEC component-level benchmarks need to be evaluated from the standpoint of redundancy and completeness, and that such procedures can be efficiently used for tuning of existing benchmark suites and reducing the cost of benchmarking.

Download paper
Download handouts

Late Breaking: Level of Service Based Capacity Sizing
Josef Ferrandiz
Alex Gilgur

The role of a capacity planner boils down to solving optimization problems: provisioning just enough to survive whatever is coming at the system, with some safety margin. One of the many interesting problems in capacity planning is that of circuit sizing. It is relevant in today’s IT environments since we often need to decide how many Apache processes, worker threads, etc. to configure an application with. We show an economical solution that we are now using; identify potential pitfalls to avoid; and share some thoughts on how many other capacity-related problems can be solved using the same approach.

Download paper
Download handouts

Queuing Network Models to Predict the Completion Time of the Map Phase of MapReduce Jobs
Daniel Menasce, George Mason University
Shouvik Bardhan, George Mason University

This session presents analytic models for Hadoop/MapReduce environments and illustrates how simple models that neglect resource contention in compute notes significantly understimate the completion time of the map phase of MapReduce jobs. Our models capture contention at compute nodes and parallelism gains due to increased number of slots available to map tasks. Experiments validated the model on a single as well as a 2-node Hadoop environment. We ran experiments for different input split sizes and different map slot sizes to validate our model.

Download paper
Download handouts

Mythbuster for The Guerillas
Jayanta Choudhury

A methodology to estimate the constant parameters of the Universal Scalability Law (USL) has been proposed by Dr.Gunther. Certain theoretical inadequacies of the subject methodology are found when applied to some typical example data sets. An improved methodology is proposed to circumvent the problems. USL is also compared against the Super-serial Scalability Law (SSL) for accuracy. SSL is found to be more robust than USL in the case of noisy or insufficient input data.

Download paper
Download handouts

When Load Testing Large User Population Web Applications The Devil Is In the (Virtual) User Details
James Brady, State of Nevada

Many times load testing is dismissed as a waste of time and money because past results didn’t conform to real world experience when the application went live. Sometimes it’s because the test suite is too narrow but often it is due to the approach used to produce traffic and the way results are interpreted. This session focuses on the latter situation because a lack of testing scope is an obvious limitation, but poor quality traffic and improper analysis techniques are subtle shortcomings that impact test credibility in ways that aren’t always clear until the live application reveals them.

Download paper
Download handouts

Green IT Based Performance Measurement and Tuning
Vaibhav Bhatia, Infosys Limited
Mitesh Desai, Infosys Limited
Arvind Raman, Infosys Ltd.

The reliance of modern business on IT has lead to rapid growth in IT real estate. With rising energy costs and increasing global regulations, it is important for IT organizations to reduce their environmental impact. Measurements and improvements have to be made in these areas: data center, end user computing, people, IT services, and IT asset management. In this session the speakers will share their experiences in implementing a green IT optimization project based on their approach. This session provides best practices, along with detailed checklists and how-to instructions.

Download paper

Becoming an Intrapreneur
Denise Kalm, Kalm Kreative, Inc.

What would it be worth to you to be more excited about going to work each day? For too many of us, much of our work day is spent doing a series of tasks no more interesting than scrubbing out a sink or changing a tire. Some decide simply to change jobs to reconnect with the joy they once felt, only to discover that nothing has changed after the initial newness wears off. Learn how to connect your passion to your expertise and then market it to your management, to create a job you can truly love. There is no passion to be found playing small - in settling for a life that is less than the one you are capable of living. Nelson Mandela

Download paper

Towards Processing Capacity Prediction for the Cloud: Theoretical Models
André Cesta, Eldorado Research Institute
Geraldo Silva, Eldorado Research Institute
Mauro Storch, Eldorado Research Institute

Amdahl’s law and the universal scalability model are functions of the number of machine processors p, therefore only modeling a single machine with varying quantities of homogeneous processors. These models cannot directly predict software capacity for: heterogeneous machines (varying in their threads and clock speed); virtual machines (on the same server, migrating across servers or scaling-up). We address these shortcomings by: (1) generalizing and extending these models to cover heterogeneous machines and virtualization, (2) proposing model building methods for QA and production systems.

Download paper
Download handouts

Performance Engineering a Web 2.0 Application in an Agile World
Mohit Verma

Agile Methodology brings in new complexity and challenges to traditional waterfall-driven performance engineering approaches. We will discuss a Agile Performance Testing Lifecycle of a mission critical Web 2.0 app. We will discuss when as performance engineers, to get involved in the project lifecycle, understand the technical architecture, what are the NFRs,what tools to use, what value pairing with Developers and Test leads provides, how the rapid changing environment effects your test scripts, make tuning suggestions and how to communicate efficiently to the project teams.

Download paper
Download handouts

Converting From SAS/GRAPH(r) to ODS Graphics
Jim Horne, Lowe’s Companies, Inc.

As of SAS 9.3, SAS has moved ODS Graphics and the Statistical Graphics procedures from SAS/GRAPH® to Base SAS®. This provides an opportunity to eliminate SAS/GRAPH by converting SAS/GRAPH procedures to ODS Graphics procedures if we choose to do so. (PARA)This is an overview of some ways to convert basic graphs from SAS/GRAPH to ODS Graphics. It provides a brief introduction to ODS Graphics and the SAS statements required to run the ODS Graphics procedures. It then compares some basic SAS/GRAPH and ODS Graphics output along with the statements required to produce each graph.

Download paper
Download handouts

Late BreakingCross-Platform Application Performance Management Implementation on zEnterprise Hybrid
Glenn Anderson, IBM

The Platform Performance Management component of IBM zEnterprise Unified Resource Manager extends a goal oriented performance management capability to both traditional System z and BladeCenter applications. This session introduces the intersection of WLM, RMF, and Platform Performance Management, and helps you understand the leading-edge performance management capabilities of zEnterprise for cross-platform hybrid applications. The most recent announcements, including load balancing, API’s for data access, and statements of direction will all be covered.

Download paper
Download handouts

Follow The Evidence' - Web Application Performance Investigations.
Jack Woolley Jr.

This session shows the basic usage of a free browser tool to perform initial performance analysis. The tool shows how a client browser requests and receives web page objects. The tool also allows prioritizing the performance investigation, thereby maximizing the benefit for a given effort. The tool provides clarity on there a performance analyst should focus attention. This is called the Follow the Evidence strategy for web application performance investigations. Finally, the session presents several examples that demonstrate how this strategy was successfully implemented.

Download paper
Download handouts

Utilizing Benchmarks in Performance Models - Varying Software/Hardware Platforms Case Study
Nidhi Tiwari, Infosys Limited
Harish TS, Infosys Limited

To analyze the implications of hardware/software platform changes on application’s performance, either elaborate performance testing methodologies or economic scientific techniques like performance modeling are used. Using industry benchmarks, which provide comparative hardware performance ratings, a performance model can predict impacts of hardware changes but cannot predict impacts of software platform changes. Here we present our experience of using performance modeling with hardware/software benchmarks to predict system performance on varying platform combinations for a web application.

Download paper
Download handouts

A Notification Management Framework Integrating ITIL Processes and Network Monitoring
Lilian Nassif, Minas Gerais Public Ministry
Daniel Carnevalli
Frederico Passos

Information and Communication Technology (ICT) currently supports the business of all modern enterprises. This dependency demands a predefined information flow when an incident occurs. The success in notification management following ITIL, among other challenges, depends on its adaptability to operational tools of network monitoring. This session presents a notification framework that integrated ITIL management processes in an open source network tool. A case study demonstrates that the solution is viable and highlights the benefits and difficulties of notification management.

Download paper
Download handouts

Late Breaking:Don't Let the Cloud Fog Your Mind: An Update for the Mainframe Sysprog
Glenn Anderson, IBM

As a mainframe sysprog in 2012, what is your reaction to the term cloud computing? For many of us, it is simply confusion. What do vendors, or your boss, or the IT industry mean today when they say cloud computing. Is the term even relevant any more? Don’t let all the hype cloud your thinking. This presentation clears the current fog for you and nets out just what you need to understand about the current state of cloud computing.

Download paper
Download handouts

Performance Implications of Big Data
Yiping Ding, OPNET Technologies

Big data is a big opportunity as well as a challenge for businesses and IT organizations. Not all IT organizations are facing big data challenges. But if they are, a very different strategy for both hardware and software has to be taken. The enormous data volume has changed the way we acquire, organize, and analyze the data. In this session, we investigate typical big data solutions and identify their performance implications.

Download paper

Fear and Loathing in Automation-Land - Capacity and Performance in the 21st Century
Denise Kalm,

Remember when you only had a handful of servers or a single mainframe to manage? Back in those days, we also had teams of experts working on capacity and performance management. Now, one person is managing 1000’s of servers. It isn’t possible to manage as you have in the past – you need to change the game, eliminating manual processes wherever possible. Learn how to work smarter and have more fun on the job by letting go of the past and venturing boldly into the future of capacity planning. Build a vision and the management skills to create it.

Download paper
Download handouts

Load Testing: See a Bigger Picture
Alexander Podelko, Oracle

Load testing is an important part of the performance engineering process. It remains the main way to ensure appropriate performance and reliability in production. Still it is important to see a bigger picture beyond stereotypical last-moment load testing. There are different ways to create load; a single approach may not work in all situations. Many tools allow you to use different ways of recording/playback and programming. This session discusses pros and cons of each approach, when it can be used and what tool's features we need to support it.

Download paper
Download handouts

0 to 100 - The Supply Chain of Business Change
Arvind Raman, Infosys Ltd.
Aswin Kumar

Organizations world-wide are struggling to justify their IT spending. This session illustrates a case study where the speakers implemented an ITIL based Business Change Methodology to enable Portfolio Management Transformation and Value Realization of Business Change. This session also describes an approach to rollout a Business Change in an Enterprise from 0 (i.e. the time when a business change is conceptualized) to 100 (i.e. the time when value is realized), in 3 simple steps. The speakers will share their best practices, detailed approach, checklists and how-tos.

Download paper

Virtual Memory Subsystem Performance: Benchmarking and a Case Study
Manoj Nambiar

When a program need for memory cannot be satisfied by the available RAM the operating system resorts to using virtual memory. In doing so there is a certain performance penalty as secondary storage access is involved. In this session the authors present the implementation of a microbenchmark for measuring virtual memory performance. Also presented are results of this microbenchmark executed on two different virtual memory configurations. The results being not so obvious are followed up with detailed performance analysis.

Download paper

Performance Prediction for Processors and External Resources
André Cesta, Eldorado Research Institute
Geraldo Silva, Eldorado Research Institute
Mauro Storch, Eldorado Research Institute

Regular forms of Amdahl’s law and the universal scalability model are functions of the number of processors p alone, therefore modeling only data sets with a single machine varying in its processor count. This session demonstrates how these performance models can be expressed in terms of intrinsic machine characteristics such as: processor clock; external resource access speed; total threads; etc. Our new models broaden all previous applications including: sensitivity analysis, forecasting, scaling; while providing new ones such as evaluations for: migration, purchase, upgrading, etc.

Download paper
Download handouts

The APM Challenge
Andrea Glaser, Marriott International

Performance—responsiveness and scalability—is a make-or-break quality for applications. Many projects consider performance objectives as an afterthought. As a consequence, the application may fail to respond rapidly enough to end user actions or may be unable to handle the number of business operations required. This is the Application Performance Management (APM) challenge and how an IT organization responds to it will have a profound impact on its position within the corporation. Meeting this challenge will reduce overall project costs by minimizing production break fix which is costly, let alone risky and stressful. This paper provides helpful information to successfully guide your organization’s management through the APM challenge.

Download paper
Download handouts

Performance Requirements: the Backbone of the Performance Engineering Process
Alexander Podelko, Oracle

Performance requirements should to be tracked from system's inception through its whole lifecycle including design, development, testing, operations, and maintenance. They are the backbone of the performance engineering process. However different groups of people are involved in each stage and they use their own vision, terminology, metrics, and tools that makes the subject confusing when you go into details. The session discusses existing issues and approaches in their relationship with the performance engineering process.

Download paper
Download handouts

Connecting the Dots - Business Unit Drivers for z/OS Capacity Planning
Gregory Caliri, BMC Software, Inc.

Over the years, z/OS performance specialists, capacity planners, and cost analysts have had access to a sea of activity measurement data, from which they attempt to determine if goals are being met, or how continued growth will impact resource availability. One dilemma they face is identifying and quantifying the activity that is driven by distinct business units. It is possible to correctly categorize a large portion of activity, provided certain steps are taken.

Download paper

Improve Performance Testing by Employing the Scientific Method
Larry Watson, An-de-le Computing Company
Coburn Watson

The outline of this session is based on the steps of the scientific method applied to performance testing: (1) formulate a hypothesis, (2) design and then (3) execute a series of experiments (tests) to either prove or disprove the hypothesis. Through this iterative process, the hypothesis is refined, or the systems-under-test are reconfigured, until the hypothesis is proven. Either reconfiguration of the systems-under-test, or the load generation capacity, can be accomplished within minutes in many of today’s cloud-based infrastructures. Common mistakes in configuration will be addressed.

Download paper
Download handouts

Capacity Planning and Performance Software Procurement Strategies
David Halbig, First Data Corporation

Every organization has faced the prospect of buying tools for capacity planning and performance management. The effort can be time-consuming and contentious, with vendors sometimes driving the process. The author proposes a solution-agnostic method based on Use Case Methodology that has proven very effective over the 5 times he has either driven or advised its use. The session is divided into a) requirements gathering, b) requirements prioritization, c) conversion to Use Cases and d) managing the vendor review and selection process.

Download paper
Download handouts

Late Breaking:Understanding, Exploiting, and Enjoying the Latest z/OS Performance Management Capabilities
Glenn Anderson, IBM

Confused with all the new functionality in recent z/OS releases and zEnterprise processors? Let one of your favorite IBM instructors walk you step by step through a number of cool new capabilities that are now part of your z/OS performance management world. Let’s slow down and spend some time on line items you have heard about, but not taken the time to really understand. Topics will include WLM and the DS8000 I/O Priority Manager, WLM updates to transaction management and LDAP management, workload promotion and what’s reported in RMF, new RMF report data and monitoring functions, hiperdispatch, etc.

Download paper
Download handouts

Taking a trip down VMware vSphere Memory Lane
Jamie Baker, Metron Technology Ltd

Within this session we take a look at how memory is used within an VMware vSphere environment. Starting the journey with the 'Memory Management Concepts' breakfast indulging in Hardware and Software MMU, we take a quick stop at the 'Memory Reclamation and Overcommitment' rest area. Lunch is served at the 'How limits are enforced' and 'what Memory metrics to monitor' restaurant, followed by a visit to the 'Guest Virtual Machine Memory reporting' cafe. Finally back home for a light meal on 'the vRAM licensing model and how it may affect you'? Cup of Hot Chocolate anyone?

Download paper

Practical SPE in a Challenging World. Bringing Value to Companies, Customers and Ourselves.
Enrico Ottolini, Swisscom

Competition is increasing, available resources are stretched, and budgets are limited. The general economy is in large part still in severe crisis. In these challenging conditions, what are the opportunities for implementing and rolling-out valuable System Performance Engineering (SPE) initiatives, be professionally successful and prevent the risk to be considered redundant? This session presents the author's direct experiences, defining and practically implementing a flexible SPE approach, in multiple and actual Web Portals and IPTV projects.

Download paper
Download handouts

Case Study: Software Performance Engineering (SPE) Techniques for Certifying Support Portal
Robert Jahn, Collaborative Consulting

This session presents a case study for the successful application Performance Engineering techniques for the certification of a customer support web portal. This case study will provide an overview of the system; project goals; workload modeling, simulation, and monitoring, approach, tools and techniques; challenges and lastly results. In particular, the scalability issues with the Java caching uncovered and addressed, topology and sizing of SOA tier that were analyzed and optimized, as well as many other valuable lessons were learned will be presented in detail.

Download paper

IT Asset Management in Today's Changing Environment
Jeff Ebbrecht, IBM

As new technologies emerge, how well positioned is your IT Asset Management system or process? What are some of the common practices used today to ensure your IT Asset Management system is delivering value to your forecast and budget? What items are needed to be tracked in your Asset Management system for use in your IT Chargeback rates, cost allocation model, or showback model? How do you make necessary changes to your budget and forecast when budget changes are requested or when the business dynamically improves?

Download paper
Download handouts

Performance Prediction Using Minimal Testing and Parallel Characteristics of a Program
Amol Khanapurkar, Tata Consultancy Services
Mohit Nanda, Tata Consultancy Services Ltd.
Suresh Malan, Tata Consultancy Services

Amdahl’s Law predicts the program speed-up on multiple processors when serial and parallel proportion of the program is known. In practice, estimating these, is not easy. Hence one resorts to testing. Based on testing, the speed-up and the Serial / Parallel proportions can be found. Amdahl’s Law is applicable to activities that can be parallelized on multiple instances. Using this property, we show the Law can be applied to a state-space of a combination of 2 resources to get performance predictions. Performance can thus be projected for an explosive number of test cases based on minimal testing.

Download paper