Resources

By Neil J. Gunther on June 1, 2010

Commercial Clusters and Scalability

In this paper we present an introductory analysis of throughput scalability for update intensive workloads (such as measured by the TPC-C or TPC-W benchmarks) and how that scaling is limited by serialization effects in the software-hardware combination that comprises any platform.

UNIX Load Average Part 1: How It Works

In this online article Dr. Gunther digs down into the UNIX kernel to find out how load averages (the “LA Triplets”) are calculated and how appropriate they are as capacity planning metrics.

UNIX Load Average Part 2: Not Your Average Average

This is the second in a two part-series where we explore the use of UNIX load averages in performance analysis and capacity planning.

UNIX Load Average: Reweighed

This is an unexpected Part 3 to the discussion about the UNIX load average metric answering the question of where the weight factor comes from.

How to Get Unbelievable Load Test Results

This article is about delusions that arise from incorrect interpretation of load test measurements.

Capacity Calculations: Handle with Care

This paper discusses avoiding calculation results that are more precise than is justified by precision of corresponding measurement input data.

Let the Bridge Fall—As long as it falls on time

Dissect the psychology of IT managers that prefer to let a project fail as long as it fails on time in this provocative article by Dr. Gunther.

Of Buses and Bunching: Strangeness in the Queue

This article explains how correlated or bunched requests can impact capacity planning results.

How to Measure an Elephant

This article explains how correlated or bunched requests can impact capacity planning results.

Evaluating Scalability Parameters: A Fitting End

This is the final online article concerning the concept of application scalability. Here, you will learn how to determine value of the parameters that control scalability.

The “LA Triplets” Quiz

This is a little quiz to test your understanding of the triplet of numbers that appear in the UNIX® load average (LA) performance metric.