Sunday, May 21, 2017

How Cross-stack Configuration Errors can Ruin a 360 Degree Panorama Website

by Mohammed SayaghEcole Polytechnique de Montréal.

Associate Editor: Sarah Nadi, University of Alberta (@sarahnadi)

Wordpress user “msayagh” could not upload his latest 360 degree panorama of Buenos Aires’ Plaza de Mayo to his Wordpress site (Figure 1). That's strange, since he was using the well-known “NextGen” Wordpress plugin. Being a software engineer at heart, and PhD student at work, he delved into the plugin’s source code, without finding any clue. Little did he know that the NextGen problem was not due to a source code problem, but due to an incorrect configuration option.

Basically, software configuration is the mechanism used to adapt a software system to different situations. For example, one has to only change an option like “file_uploads” to enable/disable file uploads in a particular system. While this gives a lot of flexibility to people to customize their system, it also means that assigning a wrong value to one option can lead the software to an incorrect behaviour. In the research literature, researchers found that such errors are frustrating as they are severe and require a lot of efforts to be fixed.
Figure 1: Configuration error encountered by msayagh’s 360 degree panorama.

Unfortunately, msayagh’s configuration error was not an ordinary configuration error, because only much later was the poor student able to find that the misconfigured option was not an option of “NextGen”, but of the PHP interpreter. That’s right! Although the error message popped up in NextGen’s GUI layer, the actual configuration option causing the trouble had to be fixed much deeper inside the Wordpress LAMP stack (Fig. 2). In other words, a particular system’s configuration is determined by the integration of configurations of each layer in a stack, i.e., each layer’s configuration is linked to each other (more on that later).

Figure 2: LAMP Stack Architecture

Was msayagh just using the wrong Wordpress plugin on the wrong day, or are cross-stack configuration errors a fundamental problem? By manually studying 1,082 configuration errors from StackOverflow and StackExchange, he found that cross-stack configuration errors are intrinsically more severe than single-layer errors and often occur during devops operations [1]. On the other hand, single-layer errors particularly occur during the setup of an environment and the maintenance of a website, which is indeed less severe.

Heartbroken from the 360 degree panorama incident, msayagh started working on an approach to help other Wordpress users like himself to debug cross-stack configuration errors using modern source code analysis techniques. He observed how existing analysis techniques only targeted single-layer errors, yet were quite good at that. For example, let’s consider the example of Figure 3 with three stack layers, each of which use a given programming language (PHP and C) and have their own configuration options.

Figure 2: LAMP Stack Architecture.

Was msayagh just using the wrong Wordpress plugin on the wrong day, or are cross-stack configuration errors a fundamental problem? By manually studying 1,082 configuration errors from StackOverflow and StackExchange, he found that cross-stack configuration errors are intrinsically more severe than single-layer errors and often occur during devops operations [1]. On the other hand, single-layer errors particularly occur during the setup of an environment and the maintenance of a website, which is indeed less severe.

Heartbroken from the 360 degree panorama incident, msayagh started working on an approach to help other Wordpress users like himself to debug cross-stack configuration errors using modern source code analysis techniques. He observed how existing analysis techniques only targeted single-layer errors, yet were quite good at that. For example, let’s consider the example of Figure 3 with three stack layers, each of which use a given programming language (PHP and C) and have their own configuration options.

Figure 3:Simplified source code of 3 LAMP stack layers (left) with their corresponding dependency graphs (right).

To analyze this example, a single layer approach for a given layer would rely on a graph model of the layer’s data [2] and control-flow dependencies [3], as shown on the right side of Figure 3. If the misconfigured option (i.e., the option that msayagh would like to find using his approach) in the example is “Option4” and the error message is printed by the print statement circled in red, a single-layer approach would very efficiently perform a breadth-first search starting from the circled node trying to find nodes that use a given option node, yielding only “Option2” (i.e., a false alarm). If there would only be a way to “link” the middle layer’s dependency graph to the graphs of upper and lower layers, the code accessing Option4 (i.e., the culprit option) would have been found!

At this point, msayagh suddenly realized that the missing links in the graphs are in fact the “Physical Links” between layers of a stack, such as the calling conventions from a plugin to Wordpress, or from a PHP program to the language primitives implemented in C. Since such links are documented (or in the worst case can be found in the stack layers’ code), they are straightforward to find and add to the dependency graph (see Figure 4). For example, the function “bar()” in the top layer is connected to its implementation “php_bar()” the lower layer by following a common naming convention in PHP. Similarly, we have the connections between the call to the function “mysql_bar()” and its implementation in the lower layer. By performing a simple breadth-first traversal of the unified stack dependency graph, the approach now yields (in that order) Option 4, Option1/2 and Option 3, with the first recommendation being the right answer. Since this new approach basically reuses the results of existing single-layer approaches, it is easy to generate the dependency graphs of additional layers or swap a single-layer approach for an equivalent one.
Figure 4: Model with “Physical links”

The approach basically reports as output the suspected configuration options to have an incorrect value. While the first reported option is the most likely to be misconfigured, the last one is the least likely.

Being an empirical researcher, msayagh evaluated the effectiveness of his approach on 36 real Cross-stack configuration errors, and found that the approach was able to identify the misconfigured option for 32 cases, with 23 misconfigured options reported as top recommendation. Each time, the approach finished its analysis in a median time of 230 seconds (asking a question on StackOverflow would have taken a median of 2.33 hours, without a strong guarantee of correctness).

In other words, in less than 4 minutes, our approach would enable any Wordpress user to upload their favourite 360 degree panoramas, or configure any other plugin in their Wordpress system! For more details, please see our ICSE 2017 paper [1] or if you happen to be in Bueones Aires next week, come see our presentation!

[1] Mohammed Sayagh, Noureddine Kerzazi, Bram Adams, “On Cross-stack Configuration Errors”, in Proc. of the 39th ICSE, 2017. Link:
[2] Data flow:
[3] Control flow:

Sunday, April 2, 2017

Scaling up the software development process

by Mark Basham, Diamond Light Source (@basham_mark)
Associate Editor: Christoph Treude (@ctreude)

What should I talk about at the first RSE conference in the UK?

New conferences don't start every day, and so when a community which you have been a part of for a while decides to run one and asks for abstracts, it's important to take part.  The community I’m talking about is the UK’s RSE or Research Software Engineer community, and it has been gathering strength over the last few years.  Originally started by the SSI (Software Sustainability Institute, the initial aim of the group was to create a job title and career path for the oft underappreciated group of researchers who support their research group by writing and maintaining, often critical, software.  Given this audience the question is really what should I talk about?  I contribute to several open source software packages, which could always do with more publicity (,, but that felt like a shameless plug which would only interest a very minor group, and I wanted something which would appeal to a larger audience.

Project Management?

One more general area in which I have been working is project managing a large cross group software project at Diamond Light Source (  When I started working the project about 2 years ago, I had little experience of project managing such a complex project, although Diamond runs projects routinely.  I cannot understate that the amount I have learnt from my peers at Diamond whilst making the transition from developer to manager during this project is huge, and wasn't long before I was directed towards a selection of texts in the field.  The wealth of information available on the topic is absolutely astounding, and almost immediately influenced my management decisions, but coming from my background as a scientist, this wealth of knowledge had passed me by almost entirely.  It seemed like exactly the kind of thing on which to base a presentation.

The Books?

There are many good books about project management for software, but the ones which I discovered and found particularly relevant were the Mythical Man Month, Peopleware, and User Story Mapping.

The Mythical Man Month

This book is from the 70s but still has many relevant points for the modern developer.  The key take home for me was that if you want to write software well, as in tested, integrated and ready for others to use, it takes about 10 times longer than to just hack it up.


This is a great book, and I have enjoyed other books by the Atlantic Systems Guild, such as “Adrenaline Junkies to Template Zombies” and “Slack”.  The thing that resonated with me here was how damaging context switching was, and it is something I see regularly in our predictably overstretched team.

User Story Mapping

The last book I wanted to talk about was User Story Mapping, simply because getting requirements right is always difficult, and the book suggested a clear and concise method for doing this, as well as a nice discussion on the merits and pitfalls of user stories.

So what are you going to talk about?

OK, so three books do not make a presentation, but they do lay the foundations for a talk based around work we conducted at Diamond to try to get the aforementioned project back on track.  There was one deliverable which for many valid reasons was falling behind the rest of the project, and put the whole initiative in jeopardy.  Diamond took action. First we identified a team, and freed up their time to focus on the deliverable to stop them from having to multitask.  We moved them out of their usual open plan environment and into a new office for the duration of the intervention, increasing their workspace by around 70% and reducing off topic distractions.  Finally we trialled a new method for identifying requirements and planning the work based on the Story Mapping approach.  The cost of these interventions was high, but the result was significant, bringing the deliverable back on track with a generally high level of design, code quality and tests in place. This journey seemed like a perfect topic for a talk to a growing community which may meet similar challenges.

Final Title?

“Scaling up the software development process, a case study highlighting the complexities of large team software development” (  Personally I enjoyed preparing it, and the associated paper is available if you are interested in the results of the experiment.

Monday, March 20, 2017

IEEE January/February Issue, Blog, and SE Radio Summary

Associate Editor: Brittany Johnson (@brittjaydlf)

The January/February issue of IEEE Software continues to deliver new and exciting things happening in the software world. This issue covers a few different of topics, such as global software engineering and software requirements, but focuses on software engineering for the internet of things. A large subset of the articles in this issue pertain to this topic.

This issue features an article on requirements engineering titled "Guidelines for Managing Requirements Rationale" by Anil Kumar Thurimella, Mathias Schubanz, Andreas Pleuss, and Goetz Botterweck. In this article, the authors discuss both the challenges involved in requirements rationales along with detailed guidelines for creating them. Based on the article, the guidelines presented are best suited for organizations that require both transparency and well thought out decisions.

Most of the papers in this issue discuss challenges and concerns that need to be addressed when software engineering for the Internet of Things (IoT). Aside from the introduction to this issue's focus, titled "Software Engineering for the Internet of Things" by Xabier Larrucea, Annie Combelles, John Favaro, and Kunal Taneja, this issue features the following articles on the IoT:

Some articles provide general suggestions for IoT software engineering. For example, in "Scalable-Application Design for the IoT", Vankatesh and colleagues propose a modular approach to designing and implementing context-aware IoT applications. Given the dynamic nature of IoT applications and large amounts of raw data produced, scalability becomes an important issue to address early on in the design of IoT applications. The authors propose and analyze an approach that uses context engines, or small and simple functional units, to improve scalability in context-aware IoT applications.

In "Key Abstractions for IoT-Oriented Software Engineering", Zambonelli presents common feature across IoT software and applications to determine the important abstractions for engineering IoT software. The common features he discusses come from existing IoT research on designing and developing IoT systems. Based on the common features, such as "things" and software infrastructure "glue", the author identified the following key abstractions to consider during analysis, design, and development of IoT systems: Stakeholders and Users (Analysis), Requirements (Analysis), Groups and coalitions (Design), Avatars (Design and Development), and Smart Things (Development).

With any growth and change come challenges that must be overcome. In the article "A Roadmap to the Programmable World: Software Challenges in the IoT Era", Taivalsaari and Mikkonen provide a forward-looking roadmap to a programmable world and discuss some of the challenges software developers may face along the way.  The authors highlight the differences between IoT and typical software development, such as the system of devices that comprise an IoT application, and implications of these differences along with possible challenges. Some of the challenges highlighted in this article include considerations for multi-device programming, security of IoT systems, and dealing with the distributed, dynamic, and potentially migratory nature of IoT software.

IEEE Software Blog

The blogs posts from the past couple months continue to reflect on and discuss the state of software development and provide insights into new technologies and approaches relevant to the work we do. January's blog posts focus on recent innovations in new and old technologies. Most interesting is a new tool introduced called the Unified ASAT Visualizer (UAV) which is used for comparing multiple static analysis tools on a given project. This aims to developers decide whether the costs for using more than one tool (i.e. FindBugs + Checkstyle) are worth it. February touches on the theme from the IEEE Software focus topic, Internet of Things, but also discusses other topics such as rethinking the role of developers in OSS and sustainable software design.

SE Radio

Featured on SE Radio this issues is a range of topics. Most of the episodes in this issue focus on software integrity, from testing and dealing with bugs and failures to dealing with the challenges in debugging distributed systems. Invited guests include Florian Gilcher, Gerald Weinberg, James Whittaker, Donny Nadolny, Alexander Tarlinder, John Allspaw, and James Cowling. Also, SE Radio welcomed a new member to the SE Radio Team, Edaena Salinas

Sunday, March 12, 2017

The Spartanizer

Authors: Yossi Gil and Matteo Orrù, Technion, Israel

In this post, we would like to tell you about the Spartanizer, an automatic software-engineering tool developed at the Technion, that helps you code in the “Spartan Programming Style” that means that a programmer phrases his code’s statements in the same way a person of Laconia phrases his statements of speech:

saying the most in a few, clear words.
For example, the Spartanizer takes some verbose piece of code such as
public static String getCurrentLanguage()
  String language;
  if (getBooleanProperty("lang.usedefaultlocale"))
language = Locale.getDefault().getLanguage();
language = getProperty("lang.current", "en");
  return language;
Figure 1: An example taken from jEdit
(found in the Jedit project), and converts it  into the concise spartan form;
public static String getCurrentLanguage() {
 return getBooleanProperty("lang.usedefaultlocale")
           ? Locale.getDefault().getLanguage()
           : getProperty("lang.current", "en");
Figure 2: The spartanized code of Figure 1
Transformations made by the Spartanizer are too many (currently over 150) to enumerate here; their main kinds are compile time evaluation when possible, elimination of dead code, simplification of boolean expressions, removing syntactic baggage such as superfluous call to super(), short-circuiting and early returns, identifying cases in which a for should have been used instead of a while, using standard short names whenever this makes sense.  A frequently recurring theme is refactorings inspired by the distributive rules of arithmetic, ab+ac⇒a(b+c) and ba+ca⇒(b+c)a i.e., identifying a common factor of two
syntactical elements and rewriting it in equivalent way with this common factor occurring only once. The Spartanizer is thus able to unite two adjacent conditionals, factor out commonalities of the two branches of a conditional. Yet another kind of refactorings carried out by the spartanizer are applying programming idioms whenever appropriate, e.g., following fluent API programming style.
The code generated by the Spartanizer is concise, and often looks more functional than imperative, and may need some getting used to conventions. Yet, our experience in teaching spartanization, and applying it in student and non-student projects, including several projects with a couple of hundreds of classes each, shows that after adjustment period is short, and programmers tend to prefer minimal ink to verbosity, as long as the minimal like is regular, i.e., follows a small set of recurring patterns. An empirical study [1] shows that the spartanization of code makes it measurably more predictable.
Spartanized code has unique lean appearance: few statements, minimal number of intermediates, adherence to idioms, use of conditional expressions (and in recent versions Java streams) rather than explicit, general purpose, control such as if, else, while, and for. In fact, spartanized code often assumes the concise and elegant looks of functional programs, even though the original version of the code was purely imperative. For example, this function  for comparing two lists of nodes  (drawn from the Spartanizer’s implementation) was created  automatically from its imperative version.

<N extends ASTNode> boolean same(final List<N> ns1, final List<N> ns2) {
  return ns1 == ns2 ||
    ns1.size() == ns2.size() &&
       .allMatch -> same(ns1.get(λ), ns2.get(λ)));
Figure 3: An example of spartanized code taken from the Spartanizer project

Conversion can be done either interactively, or in batch form. The Spartanizer features an Eclipse plugin, which offers “spartanization” tips to the programmer. The programmer will reach the concise form by following the six tips offered by the Spartanizer to the jEdit function in Figure  Figure. Unlike many code analysis and code generation tools, the conversion from the verbose to the concise is continuous: one little change at a time. An experimental feature in the Spartanizer, available from version 2.8 is capable of doing the opposite conversion, from the concise to the verbose, making it possible for the programmer to use the mouse wheel for both zooming in and zooming out the code.
Alternatively, spartanization can be done in batch application. Unlike many deep analysis tools automatic spartanization is fast, with more than five thousand tips per minute on a contemporary computer.  Like other refactoring tools, the Spartanizer is optimistic, making e.g. some assumptions on the use of overloading, or that the order of evaluation of arguments to a function is immaterial to correctness.
Internally, the Spartanizer follows this extensible modular structure: spartanization tips are generated by small software modules, called tippers. Each of these generate its own kind of spartanization tips. Tippers are applied automatically or interactively by tip applicators, which apply automatic spartanization to user selected portions of the code (function, class, file or project).
The Spartanizer is available at the Eclipse marketplace, and open source on GitHub.

[1] Y. Gil and M. Orrú, “Code Spartanization,” in Proc. of SAC’17, the 32nd ACM Symposium on Applied Computing, Marrakesh, Morocco, April 3–7 2017.
[2] Y. Gil and M. Orrú, “The Spartanizer: Massive Automatic Refactoring,” in Proc. of 24th IEEE International Conference on Software Analysis, Evolution, and Reengineering, Klagenfurt Austria, February 20-24, 2017.
[3] The Spartanizer at the Eclipse Marketplace.
[5] The Spartanizer website:

Monday, March 6, 2017

Helping Developers Configure Application-Level Caches for Web Applications Through DevOps

by Peter (Tse-Hsun) Chen, Queen's University (@petertsehsun)
Associate Editor: Sarah Nadi, University of Alberta (@sarahnadi)

Developers nowadays use application-level caching frameworks, such as Ehcache [1], to cache objects in object-oriented languages in memory. After the objects are cached, application servers do not need to retrieve the object data from external machines such as a database server, which can significantly improve the application performance. 
Alert: Application-level caching frameworks are only useful if you know what you want to cache. However, determining the best way to use such a caching framework may not always be easy. A large-scale web application may have hundreds of different types of objects. Now the challenge is how do we know which type of objects should be cached?
Developers need to manually configure the caching frameworks to enable cache for each type of object (e.g., enable cache for Student objects). Therefore, it is difficult to know what the best cache configuration is, especially given that the benefit of caching is directly related to how users are using the application. For example, enabling caching on frequently-modified objects, will not improve the application performance, but will actually slow down the application due to frequent cache renewal. 
How can we help improve the performance of web applications by finding better cache configurations? 
CacheOptimizer - finding optimal cache configurations by leveraging DevOps.DevOps tries to combine the developer and administrator world to further improve software development. Therefore, one main concept of DevOps is to assist software development by understanding user behaviour (such as knowing how the application is used in production). However, what kind of data can we analyze to recover the user behaviour? One potential way is to instrument the application to record the information that we need. However, this solution is not acceptable for applications running in production, since instrumentation will add too much overhead to the application. 
Instead, we propose a framework, called CacheOptimizer, that analyzes readily-available application runtime logs to automatically help developers find the optimal cache configurations when using application-level caching frameworks. 
Understanding user behaviour by combining static code analysis and runtime logs.CacheOptimizer analyzes web access logs to understand how users are accessing the application. Such access logs are automatically generated by web servers like Tomcat, so we do not add any extra overhead to the application. CacheOptimizer then links the logs to data accesses using static code analysis. For example, consider the following log line: 
200 /user/getDetails/peter
The log line contains information about a user request. Here, 200 represents the HTTP status code and represents the IP of the user. Through static code analysis, we can find the method in the code that handles this particular user request. We can also uncover the type of data access that is called in this request-handler method (e.g., accessing user data in the database). Then, by combining the information in the log and the information that we obtained from the code, we know that the user request is reading detailed information about the user Peter from the database. In this simple example, since we are only reading user objects from the database, the optimal cache configuration would be to enable caching on the User class. 
In short, given the logs, CacheOptimizer can recover how the users are actually using the application and can suggest the optimal cache configuration for the application. 
CacheOptimizer brings significant performance improvement. We apply CacheOptimizer to three open source applications. We find that CacheOptimizer can significantly improve the throughput of the application (it can process more requests per second)! 
We compare CacheOptimizer with several different cache configurations. CacheAll, where we simply enable cache on all objects; DefaultCache, where we use the default cache configuration that already exists in the studied application; and NoCache, where we disable all the caches. The figure below shows the cumulative throughput when using different cache configurations in one of the studied applications. We can see that by using CacheOptimizer, we can achieve a much better throughput compared to all other configurations.
Assisting development using production data. Finding the optimal cache configuration is just one way that we can assist software development by leveraging production data. Due to the agile natural of modern software development, a lot of software design and development decisions are directly affected by how users are actually using the application. The idea of DevOps brings a new set of research challenges and interesting problems. As software practitioners and researchers, can we think of new ways to assist software development by embracing DevOps?
You can find more details about CacheOptimizer in our research paper [2].
[1] Ehcache. Last accessed Feb. 6 2017 [2] Tse-Hsun Chen, Weiyi Shang, Ahmed E. Hassan, Mohamed Nasser, and Parminder Flora. 2016. CacheOptimizer: helping developers configure caching frameworks for hibernate-based database-centric web applications. In Proceedings of the 2016 24th ACM SIGSOFT International Symposium on Foundations of Software Engineering (FSE 2016).