More Like this

1. Experience-Driven Research on Programmable Networks

   Hyojoon Kim, Xiaoqi Chen ( January 2021 , Computer communication review)

   null (Ed.)

   Many promising networking research ideas in programmable networks never see the light of day. Yet, deploying research prototypes in production networks can help validate research ideas, improve them with faster feedback, uncover new research questions, and also ease the subsequent transition to practice. In this paper, we show how researchers can run and validate their research ideas in their own backyards—on their production campus networks—and we have seen that such a demonstrator can expedite the deployment of a research idea in practice to solve real network operation problems. We present P4Campus, a proof-of-concept that encompasses tools, an infrastructure design, strategies, and best practices—–both technical and non-technical–—that can help researchers run experiments against their programmable network idea in their own network. We use network tapping devices, packet brokers, and commodity programmable switches to enable running experiments to evaluate research ideas on a production campus network. We present several compelling data-plane applications as use cases that run on our campus and solve production network problems. By sharing our experiences and open-sourcing our P4 apps [28], we hope to encourage similar efforts on other campuses. 

   more » « less
2. Issues in the Reproducibility of Deep Learning Results

   https://doi.org/10.1109/SPMB47826.2019.9037840  

   Jean-Paul, S. ; Elseify, T. ; Obeid, I. ; Picone, J. ( December 2019 , IEEE Signal Processing in Medicine and Biology Symposium (SPMB))

   Obeid, I. ; Selesnik, I. ; Picone, J. (Ed.)

   The Neuronix high-performance computing cluster allows us to conduct extensive machine learning experiments on big data [1]. This heterogeneous cluster uses innovative scheduling technology, Slurm [2], that manages a network of CPUs and graphics processing units (GPUs). The GPU farm consists of a variety of processors ranging from low-end consumer grade devices such as the Nvidia GTX 970 to higher-end devices such as the GeForce RTX 2080. These GPUs are essential to our research since they allow extremely compute-intensive deep learning tasks to be executed on massive data resources such as the TUH EEG Corpus [2]. We use TensorFlow [3] as the core machine learning library for our deep learning systems, and routinely employ multiple GPUs to accelerate the training process. Reproducible results are essential to machine learning research. Reproducibility in this context means the ability to replicate an existing experiment – performance metrics such as error rates should be identical and floating-point calculations should match closely. Three examples of ways we typically expect an experiment to be replicable are: (1) The same job run on the same processor should produce the same results each time it is run. (2) A job run on a CPU and GPU should produce identical results. (3) A job should produce comparable results if the data is presented in a different order. System optimization requires an ability to directly compare error rates for algorithms evaluated under comparable operating conditions. However, it is a difficult task to exactly reproduce the results for large, complex deep learning systems that often require more than a trillion calculations per experiment [5]. This is a fairly well-known issue and one we will explore in this poster. Researchers must be able to replicate results on a specific data set to establish the integrity of an implementation. They can then use that implementation as a baseline for comparison purposes. A lack of reproducibility makes it very difficult to debug algorithms and validate changes to the system. Equally important, since many results in deep learning research are dependent on the order in which the system is exposed to the data, the specific processors used, and even the order in which those processors are accessed, it becomes a challenging problem to compare two algorithms since each system must be individually optimized for a specific data set or processor. This is extremely time-consuming for algorithm research in which a single run often taxes a computing environment to its limits. Well-known techniques such as cross-validation [5,6] can be used to mitigate these effects, but this is also computationally expensive. These issues are further compounded by the fact that most deep learning algorithms are susceptible to the way computational noise propagates through the system. GPUs are particularly notorious for this because, in a clustered environment, it becomes more difficult to control which processors are used at various points in time. Another equally frustrating issue is that upgrades to the deep learning package, such as the transition from TensorFlow v1.9 to v1.13, can also result in large fluctuations in error rates when re-running the same experiment. Since TensorFlow is constantly updating functions to support GPU use, maintaining an historical archive of experimental results that can be used to calibrate algorithm research is quite a challenge. This makes it very difficult to optimize the system or select the best configurations. The overall impact of all of these issues described above is significant as error rates can fluctuate by as much as 25% due to these types of computational issues. Cross-validation is one technique used to mitigate this, but that is expensive since you need to do multiple runs over the data, which further taxes a computing infrastructure already running at max capacity. GPUs are preferred when training a large network since these systems train at least two orders of magnitude faster than CPUs [7]. Large-scale experiments are simply not feasible without using GPUs. However, there is a tradeoff to gain this performance. Since all our GPUs use the NVIDIA CUDA® Deep Neural Network library (cuDNN) [8], a GPU-accelerated library of primitives for deep neural networks, it adds an element of randomness into the experiment. When a GPU is used to train a network in TensorFlow, it automatically searches for a cuDNN implementation. NVIDIA’s cuDNN implementation provides algorithms that increase the performance and help the model train quicker, but they are non-deterministic algorithms [9,10]. Since our networks have many complex layers, there is no easy way to avoid this randomness. Instead of comparing each epoch, we compare the average performance of the experiment because it gives us a hint of how our model is performing per experiment, and if the changes we make are efficient. In this poster, we will discuss a variety of issues related to reproducibility and introduce ways we mitigate these effects. For example, TensorFlow uses a random number generator (RNG) which is not seeded by default. TensorFlow determines the initialization point and how certain functions execute using the RNG. The solution for this is seeding all the necessary components before training the model. This forces TensorFlow to use the same initialization point and sets how certain layers work (e.g., dropout layers). However, seeding all the RNGs will not guarantee a controlled experiment. Other variables can affect the outcome of the experiment such as training using GPUs, allowing multi-threading on CPUs, using certain layers, etc. To mitigate our problems with reproducibility, we first make sure that the data is processed in the same order during training. Therefore, we save the data from the last experiment and to make sure the newer experiment follows the same order. If we allow the data to be shuffled, it can affect the performance due to how the model was exposed to the data. We also specify the float data type to be 32-bit since Python defaults to 64-bit. We try to avoid using 64-bit precision because the numbers produced by a GPU can vary significantly depending on the GPU architecture [11-13]. Controlling precision somewhat reduces differences due to computational noise even though technically it increases the amount of computational noise. We are currently developing more advanced techniques for preserving the efficiency of our training process while also maintaining the ability to reproduce models. In our poster presentation we will demonstrate these issues using some novel visualization tools, present several examples of the extent to which these issues influence research results on electroencephalography (EEG) and digital pathology experiments and introduce new ways to manage such computational issues. 

   more » « less
3. Interpersonal Interactions that Foster Inclusion: Building Supports for Diversity in Engineering Teams

   Rodríguez-Simmonds, Héctor E. ; Jackson, Benjamin P. ; Pearson, Nelson S. ; Langus, Tara C. ; Major, Justin C. ; Kirn, Adam ; Godwin, Allison ( January 2018 , ASEE annual conference & exposition)

   Teaming is a core part of engineering education, especially in the first and last years of engineering when project work is a prevalent focus. The literature on the effects of working in diverse teams is mixed. Negative findings include decreased affect, increased frustration, and sustained conflict in teams. Positive findings include increased productivity, production of high quality products, and divergent-thinking and idea generation. Given these mixed findings, it becomes important to not only understand the practical outputs of working in diverse teams, but also how the experience of working in diverse teams influences whether students see themselves as engineers and whether or not they feel they belong in engineering. Our project, Building Supports for Diversity through Engineering Teams, investigates how students’ attitudes towards diversity influence how students experience work in diverse teams through addressing two main research questions: 1) What changes occur in students’ diversity sensitivity, multicultural effectiveness, and engineering practices as a result of working in diverse teams? 2) How do students’ perceptions of diversity, affect, and engineering practices change because of working on diverse teams? Using a multi-method approach, we deployed survey instruments to determine changes in student’s attitudes about teaming, diversity sensitivity, and openness attitudes. We also observed students working in teams and interviewed these students about their perceptions of diversity and experiences in their teams. Preliminary results of the quantitative phase show that variance in students’ attitudes about diversity significantly increase over the semester, further reflecting the mixed results that have been seen previously in the literature. Additionally, Social Network Analysis was used to characterize the social structure practices of a multi-section, large-enrollment first-year engineering course. This reveals the underlying social structure of the environment, its inclusiveness, and how diverse students work with others on engineering. Initial results indicate that students are included in social networks regardless of gender and race. Preliminary results of the qualitative phase, using Interpretive Phenomenological Analysis, have yielded relationships between student’s definitions, valuation, and enactment of diversity in engineering spaces. Individual student’s incoming attitudes of diversity and previous experiences interact with practical needs in first-year engineering classrooms to create different microclimates within each team. These microclimates depict tensions between what instructors emphasize about diversity, stereotypes of engineering as focused on technical instead of social skills, and pragmatic forces of “getting the job done.” This knowledge can help explain some of the complexity behind the conflicting literature on diversity in teams. Ultimately, this research can help us understand how to build inclusive and diverse environments that guide students to learn how to understand their own complex relationship, understanding, and enactment of diversity in engineering. By understanding how students make sense of diversity in engineering spaces, educators and researchers can figure out how to introduce these concepts in relevant ways so that students can inclusively meet the grand challenges in engineering. This curriculum integration, in turn, can improve team interactions and the climate of engineering for underrepresented groups. 

   more » « less
4. Debugging SDN in HPC Environments

   https://doi.org/10.1145/3219104.3229277  

   Hayashida, Mami ; Rivera, Sergio ; Griffioen, James ; Fei, Zongming ; Song, Yongwook ( July 2018 , PEARC '18 Proceedings of the Practice and Experience on Advanced Research Computing)

   HPC networks and campus networks are beginning to leverage various levels of network programmability ranging from programmable network configuration (e.g., NETCONF/YANG, SNMP, OF-CONFIG) to software-based controllers (e.g., OpenFlow Controllers) to dynamic function placement via network function virtualization (NFV). While programmable networks offer new capabilities, they also make the network more difficult to debug. When applications experience unexpected network behavior, there is no established method to investigate the cause in a programmable network and many of the conventional troubleshooting debugging tools (e.g., ping and traceroute) can turn out to be completely useless. This absence of troubleshooting tools that support programmability is a serious challenge for researchers trying to understand the root cause of their networking problems. This paper explores the challenges of debugging an all-campus science DMZ network that leverages SDN-based network paths for high-performance flows. We propose Flow Tracer, a light-weight, data-plane-based debugging tool for SDN-enabled networks that allows end users to dynamically discover how the network is handling their packets. In particular, we focus on solving the problem of identifying an SDN path by using actual packets from the flow being analyzed as opposed to existing expensive approaches where either probe packets are injected into the network or actual packets are duplicated for tracing purposes. Our simulation experiments show that Flow Tracer has negligible impact on the performance of monitored flows. Moreover, our tool can be extended to obtain further information about the actual switch behavior, topology, and other flow information without privileged access to the SDN control plane. 

   more » « less
5. Supporting collaborative classroom networks through technology: An actor network theory approach to understanding social behaviours and design

   https://doi.org/10.1111/bjet.13274  

   Kumar, Vishesh ; Tissenbaum, Mike ( September 2022 , British Journal of Educational Technology)

   This paper presents an implementation of Connected Spaces (CxS)—an ambient help seeking interface designed and developed for a project‐based computing classroom. We use actor network theory (ANT) to provide an underutilized posthumanist lens to understand the creation of collaborative connections in this Computational Action‐based implementation. Posthumanism offers an emerging and critical extension to sociocultural perspectives on understanding learning, by pushing us to decenter the human, and consider the active roles that human and non‐human entities play in learning environments by actively shaping each other. We analyse how students in this class adjusted their help‐seeking and collaborative habits following the introduction of CxS, a tool designed to foster (more inter‐group) collaboration. ANT proposes generalized symmetry—a principle of considering human, non‐human and more than human entities with equivalent and comparable agency, leading to describing phenomena as networks of actors in different evolving relationships with each other. Analysing collaborative interactions as fostered by CxS using an ANT approach supports design‐based research—an iterative design revision process highlighting understandings about design as well as learning—by providing a temporal and informative lens into the relationship between actors and tools within the environment. Our key findings include a framing of technologies in classrooms as bridgingagentic gapsbetween students and becoming actors engaging in different behaviours; learners enacting new agencies through technologies (for instance a more comfortable non‐intrusive help seeker), and the need for voicing and teachers to connect help networks in CxS equipped classrooms.

   What is already known about this topic

   Collaborative learning is a valuable skill and practice; opportunities to mentor others are critical in empowering minoritized learners, especially in STEM and computing disciplines.

   School norms solidify a power and expertise hierarchy between teachers and learners and fail to productively support learners in learning from each other.

   Additionally, lack of awareness about peers' knowledge is a common hindrance in students knowing who to ask for help and how.

   What this paper adds

   An example of a designed interface called Connected Spaces with potential to foster more inter‐student collaboration, especially outside of mandated within‐group collaboration—in the form of cross‐group help seeking and help giving.

   A design based research study using actor network theory highlighting the limitations of Connected Spaces in sparking notable behaviour change among students by itself but being retooled as a teacher support tool in enabling cross‐group collaborations.

   Presenting conceptions of collaboration through technologies as bridging agentic gaps and acting with new agencies in performing help‐seeking related actions.

   Provoking the idea of testing emerging technologies in classrooms along with sharing our analyses and reflections with the classroom as a key idea in computing education—surfacing the gap between designed intentions and the different kinds of extra social work needed in the on‐ground success of different technologies.

   Implications for practice and/or policy

   Designers and researchers should create and test more interfaces alongside teachers across different classrooms and contexts aimed at supporting different kinds of voluntary collaborative interactions.

   Curricula, standards and school practices should further center providing students with opportunities to engage as mentors and build communities of learning across disciplines to empower minoritized students.

   Researchers engaging in design based research should consider using more posthumanist lenses to examine educational technologies and how they affect change in learning environments.

    

   more » « less