Download paper (Available to CMG members only, Click here to learn more.)

Since the practice of capacity planning first spun off from the performance measurement and tuning area around 1980, it has generally evolved into fairly commonly used techniques. However, a number of things have changed over the last 20 years . There is less staff to do the very labor intensive work, CPU upgrades are less complex and upgrade costs are much lower. But, we are still doing the job the same way. There is another way that is faster, takes less effort and costs less. And, even though there are some tradeoffs and it requires some management buy in, it may work for you.

Download paper (Available to CMG members only, Click here to learn more.)

An architectural description of an existing application is often needed, especially for performance work. The architecture is the highest level description of an implemented system in response to its requirements; it is needed to understand interactions for performance analysis. Academic descriptions can help, but existing system documentation, combined with extensive interviews, are needed to discover the most important characteristics.

Download paper (Available to CMG members only, Click here to learn more.)

This paper examines the evolution of one reporting portal from a static to a dynamic mode of operation. The benefits of a user-customizable, data driven, webpage authoring capability will be demonstrated using ASP to create dynamic HTML and a z/OS HFS to store large quantities of reporting objects created from an MXG PDB. The distinction between dynamic HTML and dynamic report generation will be explored along with a look at some SAS and non-SAS tools that can be used to dynamically create reporting objects from a SAS data source in response to user requests.

Download paper (Available to CMG members only, Click here to learn more.)

The history of computing is a history of virtualization. Each increase in the number of abstraction layers separating the end user from the hardware makes life easier for the user, but harder for the planner who must choose a configuration to guarantee performance.

We review the architectures of several contemporary virtual systems, and report on experiments that show how naive interpretations of traditional metrics like “utilization” can lead planners astray. Then we propose a generic set of performance metrics and simple prediction guidelines that can help planners manage those systems.

Download paper (Available to CMG members only, Click here to learn more.)

Traditional alerting on performance metrics works well when servers are dedicated to one application -- the likely source of the alert is known. However when multiple workloads are hosted on a consolidated server, probable cause analysis for alerts becomes more challenging. An approach is presented for stratifying alerts by both metric and workload so that the distribution of alerts across workloads for each performance metric can be seen. An Alerts Map gives a birds-eye view of all alerting activity. The challenges and successes in implementing the multi-workload analysis are also discussed.

Download paper (Available to CMG members only, Click here to learn more.)

Sense and respond systems sense, and then respond, to opportunities and threats. A sense and respond platform can be configured to develop specific sense and respond applications. Sense and respond applications arise in homeland security, healthcare, finance, supply chains, energy, environmental protection, security and - most importantly for this talk - the management of IT infrastructure. The CMG community can make a profound impact on the space of sense and respond applications because this community has expertise in the relevant technologies and mathematics: CMG papers deal with measurement of asynchronous events, statistics, probabilistic models, information fusion, and real time. Sense and respond applications are increasingly important to society, and CMG has a major role to play. This talk will survey the field of sense and respond applications; identify applications spaces that are critical for society; survey the technologies and software architectures used in these applications; describe fundamental problems in experimentation, systems design and theory; explore the rate of growth of this space; show how the CMG community's experience is directly relevant; and discuss experience with developing applications from platforms.

Download paper (Available to CMG members only, Click here to learn more.)

Bandwidth and latency are familiar topics for IT. Both relate to system performance, but in a different fashion; both have improved significantly over the years, but at a very different pace. Their performance impact is also changing as hardware and software technology progresses. We may have to update design strategies in hardware, software, and protocols to cope. This paper examines their impact to response time from a performance analysis perspective and sheds some new light on how to manage the bandwidth imbalance at different devices and imbalance between bandwidth and latency.

Download paper (Available to CMG members only, Click here to learn more.)

QNS is an online system for the self study of the most important queuing models. QNS is publicly available on the Internet. It offers a complete educational support including: (1) theoretical presentation of material, (2) queuing theory models, (3) graphical simulator with queuing network animation, (4) laboratory experiments based on numerical solver, (5) quiz subsystem with automatic grading, (6) control system with GUI, and (7) remote access support. The paper presents the design and implementation of QNS using JSP, HTML, Java, and Tomcat.

Download paper (Available to CMG members only, Click here to learn more.)

This paper introduces load testing, an important component in optimizing software quality. Proper load testing can help to mitigate the costs of poor quality. The paper will explain the importance of load testing, when in the software development process and how optimal performance can be achieved by conducting proper load testing and capacity planning. In addition to discussing the various strategies for implementing load testing within an organization, we will also explore the very real benefits load testing returns to the organization.

Download paper (Available to CMG members only, Click here to learn more.)

Virtualization has become so widespread that industry pundits have called it a megatrend. Many forms of Virtualization are available today, each with its own set of potential gains. Of course, with every new technology comes a new set of problems. This paper explores the benefits of Virtualization and discusses the difficulties in measuring the results in the real world. A case study is included to present a practical approach for measuring and forecasting growth for virtual servers.

Download paper (Available to CMG members only, Click here to learn more.)

In this paper, we discuss software design issues that should be considered whenever jobs execute in unreliable computing environments. Specifically, we show that if proper checkpointing mechanisms are not properly implemented, then under certain conditions completion times of applications executing on the system exhibit properties of heavy-tail or power-tail distributions, which can lead to unpredictable and long completion times.

Download paper (Available to CMG members only, Click here to learn more.)

You have a script to run which represents your production environment. What do you measure, what do you look at to make sure it is representative, and examining performance? What to look for when running the tests, how to compare the results between test iterations and re-evaluating performance after hardware/software changes. When you are done, how do your tests compare in your first assessment in a production environment? The case study presented will be SQL server under Windows

Download paper (Available to CMG members only, Click here to learn more.)

If you have been working for an IT company in recent times, you must have already heard about "Excellence in Operations". Let us clarify first the meaning of operations for an IT company: in this paper it generically refers to all the tools, the procedures and the actions, aimed at Service Management, where the Service is what the IT company is providing to its clients. As it easy to understand now, Excellence in Operations basically means performing operations, and so Service Management, in a better way: at its best, in an excellent way.

Download paper (Available to CMG members only, Click here to learn more.)

This presentation will show the challenge of attempting to consolidate almost 30 monitoring tools into one cogent and predictive tool using off-the-shelf software products. The idea here is to show not a success but a process through which companies are contacted, interviewed and in some cases provided a demo to try and establish a link between somewhat disparate troubleshooting solutions. It will show what is needed is a standardization of the database; the framework of every monitoring tool. Once databases are standardized, the overall monitoring of these tools, will be realized.

Download paper (Available to CMG members only, Click here to learn more.)

I spent 18 months with a client who asked for a performance test team. By the end of the engagement, we were able to put in place a true Software Performance Engineering (SPE) process that began at the requirements stage, and that included tasks involving systems engineering, development, support for development tuning, quality assurance, production monitoring, support for evaluation of proposed production configuration changes, and capacity planning.

How was this achieved? This talk discusses the steps we took to make SPE happen for an organization that did not know it wanted SPE.

Download paper (Available to CMG members only, Click here to learn more.)

This paper presents our experience with tuning the JBoss Application Server to run the SPECjAppServer2002 benchmark. We used our knowledge of high-performance, highly scalable systems to push the JBoss Application Server to its limits. This paper describes how those performance gains were accomplished, highlights various significant improvements, and provides information about the performance gains that were attained. The reader should be able to use the information presented in this paper to tune his or her own applications deployed to the JBoss Application Server.

Download paper (Available to CMG members only, Click here to learn more.)

The HotSpot® Virtual Machine for the Java 2 Platform provides a variety of garbage collection algorithms geared towards different application behavior or requirements. This paper discusses the pros and cons of each algorithm, and shows how to gather and analyze statistics provided by each algorithm. With a greater understanding of the garbage collection options available, and how to analyze their performance, the reader will be better equipped to choose the proper algorithm for use with his or her Java applications.

Download paper (Available to CMG members only, Click here to learn more.)

How can statistical analysis bring new insight to process management? Six Sigma, which means six standard deviations from the arithmetic mean, is a measurement-based methodology for eliminating defects, or errors from any process. This paper is an introduction to the statistics commonly used in Six Sigma's DMAIC phases. A case study highlights the use of statistical tests and the appropriate use of confidence intervals, P value and control charts to validate the data and determine its statistical significance. Finally, a road map to select the appropriate tests is discussed and outlined.

Download paper (Available to CMG members only, Click here to learn more.)

The IT world has changed. With job losses exceeding 400,000 in the last four years, survival in this New Age requires much more than just putting in a good day’s work. The tools and strategies you need include: career assessment, ‘managing up,’ personal public relations, networking and much more. Learn how to outwit, outplay and outlast the competition – to thrive, instead of just survive.

Download paper (Available to CMG members only, Click here to learn more.)

Sharing middleware application environments has gained significance in the recent past. Heterogeneous mix of transactions that run in the shared environments increases the complexity of both performance management and capacity planning. Some transactions consume more CPU power and memory than others. Different transactions peak at different times of the day. If and when all transactions peak at the same time, there may be a ''perfect storm''. In this paper, I discuss the issues involved in capacity planning for a shared webMethods environment and present a methodology for capacity planning.

Download paper (Available to CMG members only, Click here to learn more.)

A methodology is presented for developing a capacity-planning model for a highly distributed production environment in which Enterprise Application Integration (EAI) is a significant architectural element. The methodology is demonstrated by building and validating a model using data collected from a production system. The level of abstraction is described, along with a variation on the traditional model validation technique. The results demonstrate the usefulness of the methodology and the resulting models for predicting hardware scalability in highly distributed production environments.

Download paper (Available to CMG members only, Click here to learn more.)

Historically, capacity planning has delivered measurable value to the organization in the form of reduced costs, deferred acquisitions, and service level delivery, deliver a return on investment (ROI) for the organization. However, fundamental changes in the underpinnings of capacity planning ROI threaten the traditional value of capacity planning, and raise the question of whether it is still worth it to engage in capacity planning efforts. This paper examines these changes and their impacts and then considers actions capacity planners might take to adapt to the new reality.

Download paper (Available to CMG members only, Click here to learn more.)

Queuing models are solved with process probability distributions for system performance analysis and capacity planning. However, the models depend on service demand as one of the input variables which can be obtained only through measurements of an existing system. This kind of dependency makes it difficult to project the performance and capacity requirement of an enterprise software application prior to deployment in production. In this paper, we demonstrate that this constraint can be removed. Our models help support the notion of “predict and build” for developing enterprise software applications.

Download paper (Available to CMG members only, Click here to learn more.)

Fuzzy logic, also known as approximate reasoning, is an expansion of set theory originated in the 1960s by Lotfi Zadeh. It has been applied successfully in many areas, including control systems. This paper defines key concepts of fuzzy logic, shows examples of fuzzy logic in life and in real-world applications, shows how fuzzy logic can be applied to computer performance work to simplify and speed analysis and reporting, and shows a simple example of program code which implements fuzzy logic. Examples are based on MS-Windows®, but the principles should be applicable to other operating systems.

Download paper (Available to CMG members only, Click here to learn more.)

One of the first steps in the capacity planning process is the summarization of capacity/performance data. This activity aggregates detailed event-level data into a manageable number of workloads and intervals. Interval data can be summarized into average, peak or percentile figures. This paper discusses summarization techniques and intervals, and the gaps between results from different choices of each.

Download paper (Available to CMG members only, Click here to learn more.)

Software Performance Engineering Failure Modes and Effects Analysis (SFMEA) is a method to assess risk and determine the top business processes (bps) to focus on during software architecture and design, performance simulation and optimization, and performance management. Specifically SFMEA identifies how bps violate software performance anti-patterns and combines this risk with the bps’ frequency, current control plan and the voice of the customer. The product of these dimensions is the risk priority number (RPN). The bps with the highest RPN values defines the SPE focus

Download paper (Available to CMG members only, Click here to learn more.)

Summer 2004, our company brought together all of our Capacity Planners under one roof. From this group, our team was asked to lead a high priority effort that was chartered with finding unused resources and reclaiming them. These resources could then be earmarked for redeployment. This paper focuses on the first domain we choose to begin with consisting of 36 multi-processsor shared servers. All projects had been rolled out to new hardware within the past couple of years. This paper will have plenty of good ideas that YOU will be able to use at your shop to do a lot with a little.

Download paper (Available to CMG members only, Click here to learn more.)

A must task in load testing is workload generation: how to apply a load to your system. It is important to understand all possible options; a single approach may not work in all situations. The main choices are to generate workload manually, to use a load testing tool or to create a program to generate a load. Many tools allow you to use different ways of recording/playback and programming. This paper discusses pros and cons of each approach mainly based on experience with distributed business applications.

Download paper (Available to CMG members only, Click here to learn more.)

Modern performance monitoring tools allow the collecting of a vast amount of metrics for many dozens or even hundreds of nodes. In most cases, there is neither the necessity nor real opportunity to take into account all of the redundant measurements. In this paper we demonstrate how the method of principal components helps in the analysis of such multi-dimensional systems, and in particular, how it facilitates managing the collected data in a compressed form, without losing important information about the system.

Download paper (Available to CMG members only, Click here to learn more.)

In today's multi-tiered mass server infrastructure, it is highly desirable to be able to provide capacity reports in a timely, proactive and cost effective manner. This paper presents a methodology for producing a level one automated capacity planning report for a large heterogeneous, distributed systems server population. The paper describes the forecasting techniques employed to generate the report and the capacity planning process that it supports.

Download paper (Available to CMG members only, Click here to learn more.)

As Service Level Management increasingly encompasses multi-tier J2EE applications, it is no longer sufficient to do one-time capacity planning or static performance analysis and predictions. Utility Computing requires dynamic performance management capabilities that include modeling, analysis and predictions based on dynamic data. This paper analyzes various approaches to this problem and describes a methodology that can help solve this problem for real-world, multi-tier enterprise applications.

Download paper (Available to CMG members only, Click here to learn more.)

The need to calculate the reliability and expected up-time of new systems has become critical. This paper describes some simple methods that can be used to determine the reliability of a current system and the underlying components. It also shows how these techniques can be used to determine if an improvement is warranted and if so where. The paper starts by reviewing the concepts behind the calculations, using single system failures. It then goes onto calculating the probability of failure across numerous components and then shows how these techniques can be used to analyze and improve the reliability of a complex system

Download paper (Available to CMG members only, Click here to learn more.)

While scalability is important to today's software applications, few organizations understand how to quantitatively evaluate their software's scalability. This paper describes the Quantitative Scalability Evaluation Method, QSEM. QSEM uses straightforward measurements to quantify the scalability of a software application. The results provide an understanding of the application's scalability that makes it possible to extrapolate behavior to larger configurations with confidence. The seven steps of the QSEM method are described and illustrated with a case study.

Download paper (Available to CMG members only, Click here to learn more.)

Kaiser Permanente has been using application network traffic analysis as part of our application performance problem determination. Unencrypted TCP/IP network activity often identifies the portion(s) of the application that result in poor overall performance.

Training in this technique has the ability to identify poor performing application modules/objects, CPU constrained components, implementation configuration issues, and specific poor performing application SQL.

This paper presents the basic techniques of this type of problem determination, along with several real life examples.

Download paper (Available to CMG members only, Click here to learn more.)

This paper will explore benchmarking using tools and products available in the commercial marketplace. The material will include overviews of commonly used open systems benchmarks, what benchmarks are designed to measure, how vendors use and misuse benchmarks, how customer organizations can use benchmark results and the value of customers establishing an internal benchmarking process. This paper presents published results from several current industry-standard benchmarks and analyzes them.

Download paper (Available to CMG members only, Click here to learn more.)

Download paper (Available to CMG members only, Click here to learn more.)

The software industry is abuzz with excitement over promises that Service Oriented Architecture (SOA) will provide unprecedented business value by containing IT complexity, aligning IT with business, and facilitating IT agility – while also spurring a momentous paradigm shift towards virtualizing and uniformly connecting disparate software programs at the business logic layer. The key to achieving these goals is the rise of a horizontal service infrastructure layer/business application framework. This nascent but pivotal concept is illustrated through familiar metaphors and examples.

Download paper (Available to CMG members only, Click here to learn more.)

Grid computing is an emerging concept that will change the way we use and think about technology. This paper will explore the questions such as how the role of capacity planning will change in respect to the rise of grid technology and why the field is becoming ever more critical, despite rapid hardware price-performance improvements. The paper will also discuss the benefits and challenges that capacity planning analysts will face when working with grids and explore the new capacity planning objectives and innovative performance improvement strategies promised by the grid technology.

Download paper (Available to CMG members only, Click here to learn more.)

What should not be surprising, especially to anyone who has attempted to implement J2EE application management, is that performance data collection is difficult. The goal of this paper is to describe how to use ARM to instrument J2EE applications so that they can provide performance data, and implement self-discovery and self-monitoring capabilities, which when integrated with existing system management facilities effect an adaptive J2EE Application Management Framework (AMF).

Download paper (Available to CMG members only, Click here to learn more.)

Virtualization was invented more than thirty years ago to allow expensive mainframes to be shared among different applications. As hardware prices went down, the need for virtualization faded away. More recently, virtualization at all levels (system, storage, and network) became important again to improve system security, reliability, and to reduce costs. This paper explains the basics of system virtualization and addresses performance issues related to modeling virtualized systems using analytic performance models. A case study on server consolidation is used to illustrate the points.

Download paper (Available to CMG members only, Click here to learn more.)

As labor costs continue to rise, IT environments are installing automation to convert labor based tasks. There are heterogeneous niche products that have been proven very effective. The Data Center Markup Language (DCML) standard has been proposed by a consortium of companies to allow one to codify the current state of an IT environment, and provide a metalanguage for defining the policies for how that environment should be run. This paper discusses DCML, and takes a look at how performance information can be defined in a common format so that disparate products may use the information.

Download paper (Available to CMG members only, Click here to learn more.)

The performance implications of consolidating Windows systems under VMware ESX Server are analyzed using a simulation model. First VMware ESX Server overhead is measured based on data from Windows systems running as virtual machines under VMware ESX Server. Then a simulation model is used to estimate the contention for the physical CPUs by the virtual CPUs being dispatched from the Windows virtual machines. The simulation results are compared to some benchmark runs to validate the assumptions of the simulation model.

Download paper (Available to CMG members only, Click here to learn more.)

Service Orientation is emerging as the fourth wave of the computing paradigm shift because it promises to enable broad-scale interoperability and unprecedented business agility in a service value-net (ecosystem). Containing IT complexity and aligning IT with business through a set of sound and robust design principles are pivotal to the transformational power of Service Oriented Architecture (SOA). This paper looks for insights into containing IT complexity by studying the time-tested tenets and dynamics of complex fractal-like forms that abound in Nature.

Download paper (Available to CMG members only, Click here to learn more.)

Performance Engineering uses multiple performance assessment tools depending on the state of the software and the amount of performance data available. This paper demonstrates how Web Services can be used to facilitate the use of modeling tools in a plug-and-play manner thus enabling the use of the tool best suited to the analysis. The paper describes the design and implementation of a prototype Web Service for a performance modeling tool. Additionally, it shows experimental results that prove the viability of such a Web Service.

Download paper (Available to CMG members only, Click here to learn more.)

Server virtualization is one of the hottest topics for the capacity planner in 2005. How does a hardware virtualization environment like VMware ESX impact the performance measurements we have been accustomed to getting from Windows and Linux? If you can’t measure it, you can’t manage it, so how are we to measure and understand this new environment? This paper presents a lab-based study designed specifically to aid in understanding performance measurements available directly from VMware ESX. The study sheds light on the impact of virtualization on Windows and Linux performance information.

Download paper (Available to CMG members only, Click here to learn more.)

Download paper (Available to CMG members only, Click here to learn more.)

Business Service Management (BSM), rooted in ITIL, helps organizations manage IT from a business perspective. By identifying and mapping business-critical processes to the underlying IT infrastructure and services, BSM connects key business services to the IT services that manage them, such as routers, servers, and applications. While BSM does the mapping, SOA serves as the framework that connects the infrastructure. Hence, the BSM methodology and SOA are synergistic and the integration of BSM into SOA is pivotal to IT agility.

Download paper (Available to CMG members only, Click here to learn more.)

Download paper (Available to CMG members only, Click here to learn more.)

DB2 DDF processing, where the database is centralized and the application is distributed, offers some interesting challenges when diagnosing ''End-To-End'' performance problems. This paper uses a number of different sources (from SMF to UNIX log files to web reports), to create a single view of a distributed workload as part of a performance problem diagnosis. Warning: Success is NOT guaranteed!

Download paper (Available to CMG members only, Click here to learn more.)

Starting with z/OS V1R6 on z890 and z990 servers, Java applications can run on a new type of processor called the eServer zSeries Application Assist Processor (zAAP). The zAAP is a relatively inexpensive solution for installations running a large amount of Java work. This paper presents an overview of zAAP processors, describes how zAAP processors interact with z/OS, discusses some performance considerations when implementing zAAP processors, and describes the data available in RMF that can be used to analyze zAAP performance. This paper also describes the zAAP enhancements that are available with IBM's new z9 109 server.

Download paper (Available to CMG members only, Click here to learn more.)

DB2 requests originate from a variety of diverse sources including batch, CICS, DDF and SAP. Related DB2 CPU usage can be recorded in RMF, DB2, SMF 30, and other subsystem records. Proper selection and interpretation of these values will vary with transaction source, DB2 environment, product levels, and analysis objectives. Analysts must be careful to include all desired values and avoid multiple counting of the same logical utilization. This paper will discuss the sources and analysis of DB2 CPU metrics and corresponding response times. Examples will include CICS, DDF, and SAP.

Download paper (Available to CMG members only, Click here to learn more.)

FICON technology has been in place for five years now, and in the past two years significant advances have been made. One of those advances is cascaded FICON. This paper will discuss FICON cascading in depth: what it is, what its benefits are, what the important z/OS considerations are when implementing, how to implement, and finally what are the performance considerations and how to get better performance via optimizing buffer to buffer credits. The author will then go through a real life example of modeling and testing that was done to optimize a cascaded FICON environment.

Download paper (Available to CMG members only, Click here to learn more.)

A set of trained neural networks are used as mapping functions to translate the published LSPR benchmarks between the zOS 1.4 and OS/390 v2.10 sets. The neural networks provide an estimate of the latest zSeries performance under the old benchmarks. Capacity planners using previously validated LSPR workload distributions can provide an estimate directly comparable to the older machines being replaced.

Download paper (Available to CMG members only, Click here to learn more.)

DB2’s Distributed Data Facility’s (DDF) veil has been lifted from the eyes of the eBusiness applications personnel. With the ability to view and measure the performance of DDF SQL, which runs outside of the normal DB2 address space, IT personnel finally have insight into their systems. This paper will discuss: the evolution of using DDF - from a Gestalt perspective; DDF’s components and how they interact with MVS - key elements in understanding this process. Using special tools to enter the former “black box” of DDF’s address space, its SQL is revealed and thus tunable.

Download paper (Available to CMG members only, Click here to learn more.)

Java is ten years old this year and is being used more and more as the language of choice for mainframe applications under WebSphere. But Java is not COBOL. It is important that analysts understand the performance implications of this new world. This paper will use actual front line application examples to illustrate some of the performance issues involved with Java running under WebSphere on the mainframe.

Download paper (Available to CMG members only, Click here to learn more.)

In September 2003, five months after our previous mainframe upgrade, John Lewis added another two processors to our zSeries machine. Within days the peak hour CPU reached 94%; and we hadn’t even reached our busiest time of year! Fifteen months later, we are running with the same capacity, and have provided excellent service in the run up to Christmas 2004, despite an increase of over 25% in business transactions. This paper tells the tale of how, by a mixture of good luck and good judgement, John Lewis have moved from reactive panic to controlled and effective capacity management.

Download paper (Available to CMG members only, Click here to learn more.)

In IT management, CPU utilization is arguably the most basic performance metric. It is also one of the most multi-faceted, with IBM’s RMF and SMF providing over 60 variations of zSeries CPU utilizations, spanning differing degrees of granularity and complexity. So which CPU utilization metrics are right for you? Why, when, and how do you use each of them? We advocate a new, multi-tiered approach built on a broad-based data normalization technique. In addition, we also offer guidance on how to select CPU utilization metrics to address specific problems, such as capacity planning.

Download paper (Available to CMG members only, Click here to learn more.)

Best-in-class companies save tremendously by reducing their software costs via applying the principals of SAM, Software Asset Management. Often companies think that pursuing one course of action vs another will provide the “promised land” of best-in-class cost savings; however, in the author’s research and experience, there are myths associated with various strategies. You will see these myths exposed, learn what to do about them, and how to pursue an enlightened path to lower software costs.

Download paper (Available to CMG members only, Click here to learn more.)

It is quite possible, and sometimes very easy, to make dramatic savings in the load on the system caused by some applications. This presentation explains the reasons behind application performance tuning, the tool kit required, goes into some case studies or war stories, and finally gives a set of performance hints and tips to reduce overall CPU and elapsed time in applications.

Download paper (Available to CMG members only, Click here to learn more.)

Download paper (Available to CMG members only, Click here to learn more.)

The One Minute Manager suggests three principles for effective management: one-minute goals, one-minute praisings and one-minute reprimands. These principles apply equally well to TCP network analysis. Regularly finding and fixing the problems in one minute of TCP stack activity can keep your network running smoothly. If you find and fix the problems in the packets transferred in one minute from your TCP stack every week for a year, you will have found and fixed hundreds of problems.

Download paper (Available to CMG members only, Click here to learn more.)

Network service levels have long been a neglected practice due to the lack of standard measurement & reporting definitions & the fast pace of emerging technologies. However, some TCP/IP utilities (PING & TRACEROUTE) present a practical opportunity to proactively measure network service level goals. Despite arguments as to their accuracy, or even validity, with nominal coding effort and collaboration of other network tools, these TCP/IP utilities are indeed useful. This paper shares ways to use some of these utilities to gather network service level statistics. Sample reports are also provided.

Download paper (Available to CMG members only, Click here to learn more.)

Web switch is a key component of the Web cluster architecture. It is often built on a network processor (NP). This paper presents an analytic performance model for analyzing performance of NP-based Web switches. The model contains parameters of system configurations, dispatching algorithms, and workload characterizations, allowing users to trouble shoot performance bottlenecks and provision the capacity of the Web switch.

Download paper (Available to CMG members only, Click here to learn more.)

Supporting stateful protocols like CIFS and NFSv4 in a workload model (or a benchmark) requires accurate capture of the order of operations observed in various traces as this has a significant impact on the performance of the storage device.

One common tool for analyzing streams of discrete values that depend on an underlying stateful process is a Hidden Markov Model (HMM). This paper clearly illustrates how HMMs can be used effectively to capture the state behavior of CIFS and NFSv4 traffic.

Download paper (Available to CMG members only, Click here to learn more.)

Patent information is a basis for technical and strategic decisions in enterprises. Due to the huge amount of data and the complexity of the information in different languages carrying out comprehensive searches is burdensome for the inexperienced user.

The system allows an easy access to patent information by using natural language processing and a semantic network of terms together with a dictionary specially developed for the patent area. This interface is used for accessing patent data as well as the International Patent Classification by entering search terms in one of the languages.

Download paper (Available to CMG members only, Click here to learn more.)

Download paper (Available to CMG members only, Click here to learn more.)

With version 2.6 of the Linux operating system, disk schedulers have become totally modular. This paper provides a performance comparison of the four native schedulers included with the 2.6 kernel. Also of the authors’ new algorithm (the table-building bus driver) whose implementation is, like the others, available in a single C source file that may be dropped into the kernel. The algorithms are compared on a real system under a workload designed to emulate heavy web server traffic. The authors’ algorithm is seen to deliver substantial performance improvements over the other four schedulers.

Download paper (Available to CMG members only, Click here to learn more.)

This paper will discuss the aspects and best practices of sizing an ESCON to FICON migration to result in a high performance FICON architecture. We will discuss disk, tape and CTC considerations. The paper will then demonstrate these principles applied to real world examples of FICON migrations using data from actual client sizings using modeling tools.

Download paper (Available to CMG members only, Click here to learn more.)

As the cost of disk continues to decrease, disk-based backups have become more attractive. Although tape backups still have their place in data archival and off-site vaulting, they are not a very efficient method of data recovery. Database backups are staged on disk in preparation for a tape backup routine to copy to tape. The staged backup can be stored on inexpensive ATA disk relieving expensive primary disk storage. This white paper describes ATA disk capacities available for backup solutions, and also presents storage strategies for backups throughout the lifecycle of a database.

Download paper (Available to CMG members only, Click here to learn more.)

The infrastructure of SAN is complex, and finding a proper method to analyze and manage the performance of it in an efficient and economical way is very important for SAN performance management. This paper proposes an analytic method for performance management of SAN system, the way to apply the method, and illustrates the method by analyzing a living example. The comparison of analysis and experiment results shows that they are in agreement. So we can consider this method a sound solution for the performance management or control of SAN.

Download paper (Available to CMG members only, Click here to learn more.)

Increasing business and regulatory pressures are forcing more and more enterprises to consider remote replication (replication of data from one storage array to another). Fortunately, there are more options than ever before. This paper will survey the current landscape of remote replication technologies, examining the technologies and analyzing their advantages and disadvantages.

Download paper (Available to CMG members only, Click here to learn more.)

Can scheduled nighttime processing somehow escape the tyranny of the ''batch window''? Can transactions and large-scale queries be run against the same database at the same time, while maintaining acceptable levels of performance? To those managing database or other systems that require access by multiple applications to a common pool of data, such questions tend to be a key focus. Encouraged by recent developments in the SCSI standard, this paper considers the possibility of delegating such performance management to the individual disk drive.

Download paper (Available to CMG members only, Click here to learn more.)

Information Lifecycle Management (ILM) is the alignment of storage resources services to business goals and objectives. It is critically important to have an ILM methodology that associates business elements with their resources. The alignment of resources depends on understanding what is important, what the business needs are, and what corporate and regulatory requirements affect the business (e.g., Sarbanes-Oxley, and HIPAA). This paper focuses on those techniques and strategies for classifying information and data.

Download paper (Available to CMG members only, Click here to learn more.)

An explosion of data, along with new requirements for corporate governance, are generating requirements to build petabyte-sized disk subsystems. Given the current costs and complexities in administration for terabyte-sized systems, new innovations in management and monitoring are needed to scale beyond a petabyte. We are currently constructing a disk subsystem that will exceed a petabyte within the next few years. This paper presents, from a practitioner’s viewpoint, the design and implementation of a management framework that is architected to scale beyond a petabyte of storage.

Download paper (Available to CMG members only, Click here to learn more.)

This study investigates the I/O performance characteristics of raw volumes in a Linux 2.6 environment running on 64-bit processors (AMD Opteron and Intel Itanium 2). Both sequential and random I/O are tested on RAID-0, and RAID-1+0 devices. The volume managers used are VxVM, Linux LVM2, and Linux MD. Synthetic workloads are used in the study. Measurement results are reported.

Download paper (Available to CMG members only, Click here to learn more.)

Download paper (Available to CMG members only, Click here to learn more.)

Operating Systems that implement a simple round robin scheduler are inherently inefficient. In this environment tasks that have consumed or consume more CPU than their peers get equal scheduling priority on the CPU. By implementing a priority queue on the scheduler you can easily improve performance of processes that require short bursts of CPU while continuing to service the processes with higher usage demands.

The example used is based on the Minix operating system. However, this scheduling algorithm can be applied for any operating system that implements a round robin scheduler.

Download paper (Available to CMG members only, Click here to learn more.)

To improve the scalability of an existing production application it was rewritten. The rewritten application was based on a new architecture, using J2EE technology (WebLogic product). The new Application was structured with multiple sub-applications using a set of common components. The common component expected to be the most heavily used was identified; this was the Access Control Facility (ACF). Development of ACF was accelerated to allow testing, analysis and improvements to occur well ahead of the development of the other components. This paper describes the scalability testing of ACF.

Download paper (Available to CMG members only, Click here to learn more.)

Our company was paying rather large licensing fees for third-party products for gathering systems data for capacity planning and performance management. Our focus was to save cost and at the same time obtain data for mid-range systems like UNIX, Linux, and Windows servers. We wrote scripts to gather our own data using native commands.

This paper does comparisons of the various native commands and metrics/data in AIX, Solaris, HP-UX, Linux, and Windows and presents an overview of the capture methodology.

Download paper (Available to CMG members only, Click here to learn more.)

Java is a robust language as well as development platform, for both server and client applications. The JVM is the undergirding support structure for both the language and the runtime environment. We look at the limitations which the JVM places upon applications from both a CPU usage and a serialization perspective. Several real examples from previous projects are examined as test cases, and the solution to the limitations is provided.

Download paper (Available to CMG members only, Click here to learn more.)

In recent years, the Volvo AB group realized many acquisitions that impacted strongly the overall IT processes. This triggered the need for global troubleshooting tools and procedures, as well as a global web-based monitoring facility allowing worldwide servers’ diagnostically proactive survey.

Starting our project, we were seeking for a solution aiming at presenting a consolidated vision of our UNIX servers through a single portal for system administrators, managers and, in some way, customers.

The scope was to get resource utilization for a UNIX server via home-made development.

Download paper (Available to CMG members only, Click here to learn more.)

The paper describes one site's experience of using Multivariate Adaptive Statistical Filtering (MASF) to automatically recognize some common computer system defects such as run-away processes on one or multiple CPUs and memory leaks. A home-made SEDS (Statistical Exception Detection System) that captures any global and application level statistical exceptions was modified to recognize, separately report and alarm about those particular defects.

Download paper (Available to CMG members only, Click here to learn more.)

Download paper (Available to CMG members only, Click here to learn more.)

SQL Server makes many performance counters available. However numerous explanations provided via the Windows Performance Monitor simply restate the name of the counter or provide cryptic explanations. This paper discusses the interpretation of several performance counters that have proven useful in performance studies, elaborates on their Windows Performance Monitor explanations, and proposes potential courses of action.

Download paper (Available to CMG members only, Click here to learn more.)