More Like this

1. Network Capabilities for the HL-LHC Era

   https://doi.org/10.1051/epjconf/202024507051  

   Babik, Marian ; McKee, Shawn ( January 2020 , EPJ Web of Conferences)

   Doglioni, C. ; Kim, D. ; Stewart, G.A. ; Silvestris, L. ; Jackson, P. ; Kamleh, W. (Ed.)

   High Energy Physics (HEP) experiments rely on the networks as one of the critical parts of their infrastructure both within the participating laboratories and sites as well as globally to interconnect the sites, data centres and experiments instrumentation. Network virtualisation and programmable networks are two key enablers that facilitate agile, fast and more economical network infrastructures as well as service development, deployment and provisioning. Adoption of these technologies by HEP sites and experiments will allow them to design more scalable and robust networks while decreasing the overall cost and improving the effectiveness of the resource utilization. The primary challenge we currently face is ensuring that WLCG and its constituent collaborations will have the networking capabilities required to most effectively exploit LHC data for the lifetime of the LHC. In this paper we provide a high level summary of the HEPiX NFV Working Group report that explored some of the novel network capabilities that could potentially be deployment in time for HL-LHC. 

   more » « less
2. Secure and Reliable Network Updates

   https://doi.org/10.1145/3556542  

   Lembke, James ; Ravi, Srivatsan ; Roman, Pierre-Louis ; Eugster, Patrick ( February 2023 , ACM Transactions on Privacy and Security)

   Software-defined wide area networking (SD-WAN) enables dynamic network policy control over a large distributed network via network updates . To be practical, network updates must be consistent (i.e., free of transient errors caused by updates to multiple switches), secure (i.e., only be executed when sent from valid controllers), and reliable (i.e., function despite the presence of faulty or malicious members in the control plane), while imposing only minimal overhead on controllers and switches. We present SERENE: a protocol for se cure and re liable ne twork updates for SD-WAN environments. In short: Consistency is provided through the combination of an update scheduler and a distributed transactional protocol. Security is preserved by authenticating network events and updates, the latter with an adaptive threshold cryptographic scheme. Reliability is provided by replicating the control plane and making it resilient to a dynamic adversary by using a distributed ledger as a controller failure detector. We ensure practicality by providing a mechanism for scalability through the definition of independent network domains and exploiting the parallelism of network updates both within and across domains. We formally define SERENE’s protocol and prove its safety with regards to event-linearizability. Extensive experiments show that SERENE imposes minimal switch burden and scales to large networks running multiple network applications all requiring concurrent network updates, imposing at worst a 16% overhead on short-lived flow completion and negligible overhead on anticipated normal workloads. 

   more » « less
3. 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
4. Efficient service handoff across edge servers via docker container migration

   https://doi.org/10.1145/3132211.3134460  

   Ma, Lele ; Yi, Shanhe ; Li, Qun ( October 2017 , ACM/IEEE Symposium on Edge Computing)

   Supporting smooth movement of mobile clients is important when offloading services on an edge computing platform. Interruption free client mobility demands seamless migration of the offloading service to nearby edge servers. However, fast migration of offloading services across edge servers in a WAN environment poses significant challenges to the handoff service design. In this paper, we present a novel service handoff system which seamlessly migrates offloading services to the nearest edge server, while the mobile client is moving. Service handoff is achieved via container migration. We identify an important performance problem during Docker container migration. Based on our systematic study of container layer management and image stacking, we propose a migration method which leverages the layered storage system to reduce file system synchronization overhead, without dependence on the distributed file system. We implement a prototype system and conduct experiments using real world product applications. Evaluation results reveal that compared to state-of-the-art service handoff systems designed for edge computing platforms, our system reduces the total duration of service handoff time by 80% (56%) with network bandwidth 5Mbps (20Mbps). 

   more » « less
5. Optimal Cache Allocation under Network-Wide Capacity Constraint

   https://doi.org/10.1109/ICCNC.2019.8685587  

   Mai, Van Sy ; Ioannidis, Stratis ; Pesavento, Davide ; Benmohamed, Lotfi ( April 2019 , 2019 International Conference on Computing, Networking and Communications (ICNC))

   Network cache allocation and management are important aspects of an Information-Centric Network (ICN) design, such as one based on Named Data Networking (NDN). We address the problem of optimal cache size allocation and content placement in an ICN in order to maximize the caching gain resulting from routing cost savings. While prior art assumes a given cache size at each network node and focuses on content placement, we study the problem when a global, network-wide cache storage budget is given and we solve for the optimal per-node cache allocation. This problem arises in cloud-based network settings where each network node is virtualized and housed within a cloud data center node with associated dynamic storage resources acquired from the cloud node as needed. As the offline centralized version of the optimal cache allocation problem is NP-hard, we develop a distributed adaptive algorithm that provides an approximate solution within a constant factor from the optimal. Performance evaluation of the algorithm is carried out through extensive simulations over multiple network topologies, demonstrating that our proposal significantly outperforms existing cache allocation algorithms. 

   more » « less