@article{KramerLaeuter2019, author = {Kramer, Tobias and L{\"a}uter, Matthias}, title = {Outgassing induced acceleration of comet 67P/Churyumov-Gerasimenko}, volume = {630}, journal = {Astronomy \& Astrophysics}, arxiv = {http://arxiv.org/abs/1902.02701}, doi = {10.1051/0004-6361/201935229}, pages = {A4}, year = {2019}, abstract = {Cometary activity affects the orbital motion and rotation state due to sublimation induced forces. The availability of precise rotation-axis orientation and position data from the Rosetta mission allows one to accurately determine the outgassing of comet Churyumov-Gerasimenko/67P (67P). We derive the observed non-gravitational acceleration of 67P directly from the Rosetta spacecraft trajectory. From the non-gravitational acceleration we recover the diurnal outgassing variations and study a possible delay of the sublimation response with respect to the peak solar illumination. This allows us to compare the non-gravitational acceleration of 67P with expectations based on empirical models and common assumptions about the sublimation process. We use an iterative orbit refinement and Fourier decomposition of the diurnal activity to derive the outgassing induced non-gravitational acceleration. The uncertainties of the data reduction are established by a sensitivity analysis of an ensemble of best-fit orbits for comet 67P. We find that the Marsden non-gravitational acceleration parameters reproduce part of the non-gravitational acceleration but need to be augmented by an analysis of the nucleus geometry and surface illumination to draw conclusions about the sublimation process on the surface. The non-gravitational acceleration follows closely the subsolar latitude (seasonal illumination), with a small lag angle with respect to local noon around perihelion. The observed minor changes of the rotation axis do not favor forced precession models for the non-gravitational acceleration. In contrast to the sublimation induced torques, the non-gravitational acceleration does not put strong constraints on localized active areas on the nucleus. We find a close agreement of the orbit deduced non-gravitational acceleration and the water production independently derived from Rosetta in-situ measurement.}, language = {en} } @article{DelleSiteKrekelerWhittakeretal.2019, author = {Delle Site, Luigi and Krekeler, Christian and Whittaker, John and Agarwal, Animesh and Klein, Rupert and H{\"o}fling, Felix}, title = {Molecular Dynamics of Open Systems: Construction of a Mean-Field Particle Reservoir}, volume = {2}, journal = {Advanced Theory and Simulations}, arxiv = {http://arxiv.org/abs/1902.07067}, doi = {10.1002/adts.201900014}, pages = {1900014}, year = {2019}, abstract = {The simulation of open molecular systems requires explicit or implicit reservoirs of energy and particles. Whereas full atomistic resolution is desired in the region of interest, there is some freedom in the implementation of the reservoirs. Here, a combined, explicit reservoir is constructed by interfacing the atomistic region with regions of point-like, non-interacting particles (tracers) embedded in a thermodynamic mean field. The tracer molecules acquire atomistic resolution upon entering the atomistic region and equilibrate with this environment, while atomistic molecules become tracers governed by an effective mean-field potential after crossing the atomistic boundary. The approach is extensively tested on thermodynamic, structural, and dynamic properties of liquid water. Conceptual and numerical advantages of the procedure as well as new perspectives are highlighted and discussed.}, language = {en} } @article{RodriguezKramer2019, author = {Rodr{\´i}guez, Mirta and Kramer, Tobias}, title = {Machine Learning of Two-Dimensional Spectroscopic Data}, volume = {520}, journal = {Chemical Physics}, arxiv = {http://arxiv.org/abs/1810.01124}, doi = {10.1016/j.chemphys.2019.01.002}, pages = {52 -- 60}, year = {2019}, abstract = {Two-dimensional electronic spectroscopy has become one of the main experimental tools for analyzing the dynamics of excitonic energy transfer in large molecular complexes. Simplified theoretical models are usually employed to extract model parameters from the experimental spectral data. Here we show that computationally expensive but exact theoretical methods encoded into a neural network can be used to extract model parameters and infer structural information such as dipole orientation from two dimensional electronic spectra (2DES) or reversely, to produce 2DES from model parameters. We propose to use machine learning as a tool to predict unknown parameters in the models underlying recorded spectra and as a way to encode computationally expensive numerical methods into efficient prediction tools. We showcase the use of a trained neural network to efficiently compute disordered averaged spectra and demonstrate that disorder averaging has non-trivial effects for polarization controlled 2DES.}, language = {en} } @article{KramerLaeuterHviidetal.2019, author = {Kramer, Tobias and L{\"a}uter, Matthias and Hviid, Stubbe and Jorda, Laurent and Keller, Horst Uwe and K{\"u}hrt, Ekkehard}, title = {Comet 67P/Churyumov-Gerasimenko rotation changes derived from sublimation induced torques}, volume = {630}, journal = {Astronomy \& Astrophysics}, arxiv = {http://arxiv.org/abs/1812.01490}, doi = {10.1051/0004-6361/201834349}, pages = {A3}, year = {2019}, abstract = {Context. The change of the rotation period and the orientation of the rotation axis of comet 67P/Churyumov-Gerasimenko (67P/C-G) is deducible from images taken by the scientific imaging instruments on-board the Rosetta mission with high precision. Non gravitational forces are a natural explanation for these data. Aims. We describe observed changes for the orientation of the rotation axis and the rotation period of 67P/C-G. For these changes we give an explanation based on a sublimation model with a best-fit for the surface active fraction (model P). Torque effects of periodically changing gas emissions on the surface are considered. Methods. We solve the equation of state for the angular momentum in the inertial and the body- fixed frames and provide an analytic theory of the rotation changes in terms of Fourier coefficients, generally applicable to periodically forced rigid body dynamics. Results. The torque induced changes of the rotation state constrain the physical properties of the surface, the sublimation rate and the local active fraction of the surface. Conclusions. We determine a distribution of the local surface active fraction in agreement with the rotation properties, period and orientation, of 67P/C-G. The torque movement confirms that the sublimation increases faster than the insolation towards perihelion. The derived relatively uniform activity pattern is discussed in terms of related surface features.}, language = {en} } @misc{SchneckWeiserWende2019, author = {Schneck, Jakob and Weiser, Martin and Wende, Florian}, title = {Impact of mixed precision and storage layout on additive Schwarz smoothers}, issn = {1438-0064}, url = {http://nbn-resolving.de/urn:nbn:de:0297-zib-71305}, year = {2019}, abstract = {The growing discrepancy between CPU computing power and memory bandwidth drives more and more numerical algorithms into a bandwidth- bound regime. One example is the overlapping Schwarz smoother, a highly effective building block for iterative multigrid solution of elliptic equations with higher order finite elements. Two options of reducing the required memory bandwidth are sparsity exploiting storage layouts and representing matrix entries with reduced precision in floating point or fixed point format. We investigate the impact of several options on storage demand and contraction rate, both analytically in the context of subspace correction methods and numerically at an example of solid mechanics. Both perspectives agree on the favourite scheme: fixed point representation of Cholesky factors in nested dissection storage.}, language = {en} } @misc{Deppisch2019, type = {Master Thesis}, author = {Deppisch, Daniel}, title = {Advancing the Heterogeneous Active Messages Approach}, year = {2019}, language = {en} } @article{LaeuterKramerRubinetal.2019, author = {L{\"a}uter, Matthias and Kramer, Tobias and Rubin, Martin and Altwegg, Kathrin}, title = {Surface localization of gas sources on comet 67P/Churyumov-Gerasimenko based on DFMS/COPS data}, volume = {483}, journal = {Monthly Notices of the Royal Astronomical Society}, publisher = {Monthly Notices of the Royal Astronomical Society}, arxiv = {http://arxiv.org/abs/1804.06696}, doi = {10.1093/mnras/sty3103}, pages = {852 -- 861}, year = {2019}, abstract = {We reconstruct the temporal evolution of the source distribution for the four major gas species H2O, CO2, CO, and O2 on the surface of comet 67P/Churyumov-Gerasimenko during its 2015 apparition. The analysis applies an inverse coma model and fits to data between August 6th 2014 and September 5th 2016 measured with the Double Focusing Mass Spectrometer (DFMS) of the Rosetta Orbiter Spectrometer for Ion and Neutral Analysis (ROSINA) and the COmet Pressure Sensor (COPS). The spatial distribution of gas sources with their temporal variation allows one to construct surface maps for gas emissions and to evaluate integrated productions rates. For all species peak production rates and integrated productions rates per orbit are evaluated separately for the northern and the southern hemisphere. The nine most active emitting areas on the comet's surface are defined and their correlation to emissions for each of the species is discussed.}, language = {en} } @inproceedings{WeinholdLackorzynskiBierbaumetal.2019, author = {Weinhold, Carsten and Lackorzynski, Adam and Bierbaum, Jan and K{\"u}ttler, Martin and Planeta, Maksym and Weisbach, Hannes and Hille, Matthias and H{\"a}rtig, Hermann and Margolin, Alexander and Sharf, Dror and Levy, Ely and Gak, Pavel and Barak, Amnon and Gholami, Masoud and Schintke, Florian and Sch{\"u}tt, Thorsten and Reinefeld, Alexander and Lieber, Matthias and Nagel, Wolfgang}, title = {FFMK: A Fast and Fault-Tolerant Microkernel-Based System for Exascale Computing}, booktitle = {Software for Exascale Computing - SPPEXA 2016-2019}, publisher = {Springer}, doi = {10.1007/978-3-030-47956-5_16}, pages = {483 -- 516}, year = {2019}, language = {en} } @article{AbbottStraubeAartsetal.2019, author = {Abbott, Joshua L. and Straube, Arthur and Aarts, Dirk G. A. L. and Dullens, Roel P. A.}, title = {Transport of a colloidal particle driven across a temporally oscillating optical potential energy landscape}, volume = {21}, journal = {New J. Phys.}, doi = {10.1088/1367-2630/ab3765}, pages = {083027}, year = {2019}, abstract = {A colloidal particle is driven across a temporally oscillating one-dimensional optical potential energy landscape and its particle motion is analysed. Different modes of dynamic mode locking are observed and are confirmed with the use of phase portraits. The effect of the oscillation frequency on the mode locked step width is addressed and the results are discussed in light of a high-frequency theory and compared to simulations. Furthermore, the influence of the coupling between the particle and the optical landscape on mode locking is probed by increasing the maximum depth of the optical landscape. Stronger coupling is seen to increase the width of mode locked steps. Finally, transport across the temporally oscillating landscape is studied by measuring the effective diffusion coefficient of a mobile particle, which is seen to be highly sensitive to the driving velocity and mode locking.}, language = {en} } @article{StraubePagesTiernoetal.2019, author = {Straube, Arthur and Pag{\`e}s, Josep M. and Tierno, Pietro and Ign{\´e}s-Mullol, Jordi and Sagu{\´e}s, Francesc}, title = {Collective dynamics and conformal ordering in electrophoretically driven nematic colloids}, volume = {1}, journal = {Phys. Rev. Research}, doi = {10.1103/PhysRevResearch.1.022008}, pages = {022008}, year = {2019}, abstract = {We present a theoretical framework to understand the collective dynamics of an ensemble of electrophoretically driven colloidal particles that are forced to assemble around a single topological defect in a nematic liquid crystal by an alternating current electric field. Our generic model combines phoretic propulsion with electrostatic interactions and liquid-crystal-mediated hydrodynamics, which are effectively cast into a long-range interparticle repulsion, while nematic elasticity plays a subdominant role. Simulations based on this model fully capture the collective organization process observed in the experiments and other striking effects as the emergence of conformal ordering and a nearly frequency-independent repulsive interaction above 10Hz. Our results demonstrate the importance of hydrodynamic interactions on the assembly of driven microscale matter in anisotropic media.}, language = {en} } @article{SalemSchintkeSchuettetal.2019, author = {Salem, Farouk and Schintke, Florian and Sch{\"u}tt, Thorsten and Reinefeld, Alexander}, title = {Improving the throughput of a scalable FLESnet using the Data-Flow Scheduler}, journal = {CBM Progress Report 2018}, isbn = {978-3-9815227-6-1}, doi = {10.15120/GSI-2019-01018}, pages = {149 -- 150}, year = {2019}, language = {en} } @inproceedings{ClauseckerReinefeld2019, author = {Clausecker, Robert and Reinefeld, Alexander}, title = {Zero-Aware Pattern Databases with 1-Bit Compression for Sliding Tile Puzzles}, booktitle = {Proceedings of the Twelfth International Symposium on Combinatorial Search (SoCS 2019)}, pages = {35 -- 43}, year = {2019}, language = {en} } @inproceedings{Hartung2019, author = {Hartung, Marc}, title = {ZIB_Glucose: Luby Blocked Restarts and Dynamic Vivification}, booktitle = {Proceedings of SAT Race 2019 : Solver and Benchmark Descriptions}, pages = {43 -- 44}, year = {2019}, language = {en} } @inproceedings{GholamiSchintke2019, author = {Gholami, Masoud and Schintke, Florian}, title = {Multilevel Checkpoint/Restart for Large Computational Jobs on Distributed Computing Resources}, booktitle = {2019 IEEE 38th Symposium on Reliable Distributed Systems (SRDS)}, doi = {10.1109/SRDS47363.2019.00025}, pages = {143 -- 152}, year = {2019}, language = {en} } @inproceedings{HartungSchintkeSchuett2019, author = {Hartung, Marc and Schintke, Florian and Sch{\"u}tt, Thorsten}, title = {Pinpoint Data Races via Testing and Classification}, booktitle = {2019 IEEE International Symposium on Software Reliability Engineering Workshops (ISSREW); 3rd International Workshop on Software Faults (IWSF 2019)}, doi = {10.1109/ISSREW.2019.00100}, pages = {386 -- 393}, year = {2019}, language = {en} } @article{HellerFleischmannKramer2019, author = {Heller, Eric J. and Fleischmann, Ragnar and Kramer, Tobias}, title = {Branched Flow}, arxiv = {http://arxiv.org/abs/1910.07086}, year = {2019}, abstract = {In many physical situations involving diverse length scales, waves or rays representing them travel through media characterized by spatially smooth, random, modest refractive index variations. "Primary" diffraction (by individual sub-wavelength features) is absent. Eventually the weak refraction leads to imperfect focal "cusps". Much later, a statistical regime characterized by momentum diffusion is manifested. An important intermediate regime is often overlooked, one that is diffusive only in an ensemble sense. Each realization of the ensemble possesses dramatic ray limit structure that guides the waves (in the same sense that ray optics is used to design lens systems). This structure is a universal phenomenon called branched flow. Many important phenomena develop in this intermediate regime. Here we give examples and some of the physics of this emerging field.}, language = {en} } @misc{ChristgauSchnor2019, author = {Christgau, Steffen and Schnor, Bettina}, title = {MPI Passive Target Synchronization on a Non-Cache-Coherent Shared-Memory Processor}, issn = {1438-0064}, doi = {10.12752/7477}, url = {http://nbn-resolving.de/urn:nbn:de:0297-zib-74774}, year = {2019}, abstract = {MPI passive target synchronization offers exclusive and shared locks. These are the building blocks for the implementation of applications with Readers \& Writers semantic, like for example distributed hash tables. This paper discusses the implementation of MPI passive target synchronization on a non-cache-coherent multicore, the Intel Single-Chip Cloud Computer. The considered algorithms differ in their communication style (message based versus shared memory), their data structures (centralized versus distributed) and their semantics (with/without Writer preference). It is shown that shared memory approaches scale very well and deliver good performance, even in absence of cache coherence.}, language = {en} } @misc{Mattes2019, type = {Master Thesis}, author = {Mattes, Dirk}, title = {Ein System zur deterministischen Wiedergabe von verteilten Algorithmen auf Anwendungsebene}, year = {2019}, language = {de} } @article{DibakFroehnerNoeetal.2019, author = {Dibak, Manuel and Fr{\"o}hner, Christoph and No{\´e}, Frank and H{\"o}fling, Felix}, title = {Diffusion-influenced reaction rates in the presence of pair interactions}, volume = {151}, journal = {The Journal of Chemical Physics}, arxiv = {http://arxiv.org/abs/1908.07764}, doi = {10.1063/1.5124728}, pages = {164105}, year = {2019}, abstract = {The kinetics of bimolecular reactions in solution depends, among other factors, on intermolecular forces such as steric repulsion or electrostatic interaction. Microscopically, a pair of molecules first has to meet by diffusion before the reaction can take place. In this work, we establish an extension of Doi's volume reaction model to molecules interacting via pair potentials, which is a key ingredient for interacting-particle-based reaction-diffusion (iPRD) simulations. As a central result, we relate model parameters and macroscopic reaction rate constants in this situation. We solve the corresponding reaction-diffusion equation in the steady state and derive semi- analytical expressions for the reaction rate constant and the local concentration profiles. Our results apply to the full spectrum from well-mixed to diffusion-limited kinetics. For limiting cases, we give explicit formulas, and we provide a computationally inexpensive numerical scheme for the general case, including the intermediate, diffusion-influenced regime. The obtained rate constants decompose uniquely into encounter and formation rates, and we discuss the effect of the potential on both subprocesses, exemplified for a soft harmonic repulsion and a Lennard-Jones potential. The analysis is complemented by extensive stochastic iPRD simulations, and we find excellent agreement with the theoretical predictions.}, language = {en} } @inproceedings{Wende2019, author = {Wende, Florian}, title = {C++ Data Layout Abstractions through Proxy Types}, booktitle = {2019 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW), 14th International Workshop on Automatic Performance Tunings (iWAPT)}, doi = {10.1109/IPDPSW.2019.00126}, pages = {758 -- 767}, year = {2019}, abstract = {Programs that process linearly indexed fields with structured element types in a data-parallel way usually suffer from the fact that compilers fail to generate efficient code if the selected data layout appears inappropriate for the chosen target architecture. If their internal heuristics cannot proof a performance gain from a data-parallel execution, compilers may fall back to scalar code generation. Data access through proxy types together with a customized container is one means to assist the compiler in generating efficient machine code in these cases without changing the user code. We present an automated proxy-type generator (using Clang's LibTooling) and a configurable C++ container that supports different data layouts in a transparent way.}, language = {en} } @inproceedings{KnaustMayerSteinke2019, author = {Knaust, Marius and Mayer, Florian and Steinke, Thomas}, title = {OpenMP to FPGA Offloading Prototype Using OpenCL SDK}, booktitle = {2019 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW)}, doi = {10.1109/IPDPSW.2019.00072}, pages = {387 -- 390}, year = {2019}, abstract = {Field-programmable gate arrays (FPGAs) are of great interest for future high-performance computing and data analytics systems, since they are capable of efficient, highly-parallel data processing. Even though high-level synthesis became more popular in the last years, the effort of porting existing scientific software onto FPGAs is still considerable. We propose to use OpenMP target offloading as a solution, which we implement in a first prototype, making use of the preexisting OpenCL SDK of the FPGA vendor. Early results demonstrate the feasibility of this approach and also reveal that further optimizations will be necessary such that code can be written in an FPGA-agnostic way.}, language = {en} } @inproceedings{MayerKnaustPhilippsen2019, author = {Mayer, Florian and Knaust, Marius and Philippsen, Michael}, title = {OpenMP on FPGAs—A Survey}, booktitle = {OpenMP: Conquering the Full Hardware Spectrum}, doi = {10.1007/978-3-030-28596-8_7}, pages = {94 -- 108}, year = {2019}, abstract = {Due to the ubiquity of OpenMP and the rise of FPGA-based accelerators in the HPC world, several research groups have attempted to bring the two together by building OpenMP-to-FPGA compilers. This paper is a survey of the current state of the art (with a focus on the OpenMP target pragma). It first introduces and explains a design space for the compilers. Design space dimensions include how FPGA infrastructure is generated, how work is distributed, and where/how target outlining is done. A table concisely condenses the available information on the surveyed projects which are also summarized and compared. The paper concludes with possible future research directions.}, language = {en} } @inproceedings{NoackFochtSteinke2019, author = {Noack, Matthias and Focht, Erich and Steinke, Thomas}, title = {Heterogeneous Active Messages for Offloading on the NEC SX-Aurora TSUBASA}, booktitle = {2019 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW), Heterogeneity in Computing Workshop (HCW 2019)}, year = {2019}, abstract = {The NEC SX-Aurora TSUBASA is a new generation of vector processing architectures that combines a standard Intel Xeon host with the newly developed NEC Vector Engine co-processor cards. One way to use these co-processors is offloading suitable parts of the program from the host to the Vector Engines. Currently, the only vendor-provided offloading solutions are the low-level Vector Engine Offloading (VEO) library, and a builtin reverse-offloading mechanism named VHcall. In this work, we extend the portable Heterogeneous Active Messages (HAM) based HAM-Offload framework with support for the NEC SX-Aurora TSUBASA. Therefore, we design, implement, and evaluate two messaging protocols aimed at minimising offloading cost. This sheds some light on how to achieve fast communication between host CPU and the Vector Engines of the NEC SX-Aurora TSUBASA. Compared with VEO, the DMA-based protocol reduces offloading overhead by a factor of 13×. The resulting framework enables users to write portable offload applications with low overhead, that do neither require a language extension like OpenMP, nor a special language like OpenCL. Existing HAM-Offload applications are now ready to run on the NEC SX-Aurora TSUBASA.}, language = {en} } @inproceedings{Noack2019, author = {Noack, Matthias}, title = {Heterogeneous Active Messages (HAM) — Implementing Lightweight Remote Procedure Calls in C++}, booktitle = {Proceedings of the 5th International Workshop on OpenCL, The Distributed \& Heterogeneous Programming in C/C++ (DHPCC++ 2019) Conference}, doi = {10.1145/3318170.3318195}, year = {2019}, abstract = {We present HAM (Heterogeneous Active Messages), a C++-only active messaging solution for heterogeneous distributed systems.Combined with a communication protocol, HAM can be used as a generic Remote Procedure Call (RPC) mechanism. It has been used in HAM-Offload to implement a low-overhead offloading framework for inter- and intra-node offloading between different architectures including accelerators like the Intel Xeon Phi x100 series and the NEC SX-Aurora TSUBASA Vector Engine. HAM uses template meta-programming to implicitly generate active message types and their corresponding handler functions. Heterogeneity is enabled by providing an efficient address translation mechanism between the individual handler code addresses of processes running different binaries on different architectures, as well a hooks to inject serialisation and deserialisation code on a per-type basis. Implementing such a solution in modern C++ sheds some light on the shortcomings and grey areas of the C++ standard when it comes to distributed and heterogeneous environments.}, language = {en} } @inproceedings{SkrzypczakSchintkeSchuett2019, author = {Skrzypczak, Jan and Schintke, Florian and Sch{\"u}tt, Thorsten}, title = {Linearizable State Machine Replication of State-Based CRDTs without Logs}, booktitle = {Proceedings of the 2019 ACM Symposium on Principles of Distributed Computing, PODC 2019}, doi = {10.1145/3293611.3331568}, pages = {455 -- 457}, year = {2019}, abstract = {General solutions of state machine replication have to ensure that all replicas apply the same commands in the same order, even in the presence of failures. Such strict ordering incurs high synchronization costs due to the use of distributed consensus or a leader. This paper presents a protocol for linearizable state machine replication of conflict-free replicated data types (CRDTs) that neither requires consensus nor a leader. By leveraging the properties of state-based CRDTs—in particular the monotonic growth of a join semilattice—synchronization overhead is greatly reduced. In addition, updates just need a single round trip and modify the state 'in-place' without the need for a log. Furthermore, the message size overhead for coordination consists of a single counter per message. While reads in the presence of concurrent updates are not wait-free without a coordinator, we show that more than 97 \% of reads can be handled in one or two round trips under highly concurrent accesses. Our protocol achieves high throughput without auxiliary processes such as command log management or leader election. It is well suited for all practical scenarios that need linearizable access on CRDT data on a fine-granular scale.}, language = {en} } @inproceedings{HartungSchintke2019, author = {Hartung, Marc and Schintke, Florian}, title = {Learned Clause Minimization in Parallel SAT Solvers}, booktitle = {Pragmatics of SAT 2019}, arxiv = {http://arxiv.org/abs/arXiv:1908.01624v1}, pages = {1 -- 11}, year = {2019}, abstract = {Learned clauses minimization (LCM) let to performance improvements of modern SAT solvers especially in solving hard SAT instances. Despite the success of LCM approaches in sequential solvers, they are not widely incorporated in parallel SAT solvers. In this paper we explore the potential of LCM for parallel SAT solvers by defining multiple LCM approaches based on clause vivification, comparing their runtime in different SAT solvers and discussing reasons for performance gains and losses. Results show that LCM only boosts performance of parallel SAT solvers on a fraction of SAT instances. More commonly applying LCM decreases performance. Only certain LCM approaches are able to improve the overall performance of parallel SAT solvers.}, language = {en} } @inproceedings{SalemSchuettSchintkeetal.2019, author = {Salem, Farouk and Sch{\"u}tt, Thorsten and Schintke, Florian and Reinefeld, Alexander}, title = {Scheduling Data Streams for Low Latency and High Throughput on a Cray XC40 Using Libfabric}, booktitle = {CUG Conference Proceedings}, year = {2019}, abstract = {Achieving efficient many-to-many communication on a given network topology is a challenging task when many data streams from different sources have to be scattered concurrently to many destinations with low variance in arrival times. In such scenarios, it is critical to saturate but not to congest the bisectional bandwidth of the network topology in order to achieve a good aggregate throughput. When there are many concurrent point-to-point connections, the communication pattern needs to be dynamically scheduled in a fine-grained manner to avoid network congestion (links, switches), overload in the node's incoming links, and receive buffer overflow. Motivated by the use case of the Compressed Baryonic Matter experiment (CBM), we study the performance and variance of such communication patterns on a Cray XC40 with different routing schemes and scheduling approaches. We present a distributed Data Flow Scheduler (DFS) that reduces the variance of arrival times from all sources at least 30 times and increases the achieved aggregate bandwidth by up to 50\%.}, language = {en} } @inproceedings{AlonsoBinnigPandisetal.2019, author = {Alonso, Gustavo and Binnig, Carsten and Pandis, Ippokratis and Salem, Kenneth and Skrzypczak, Jan and Stutsman, Ryan and Thostrup, Lasse and Wang, Tianzheng and Wang, Zeke and Ziegler, Tobias}, title = {DPI: The Data Processing Interface for Modern Networks}, booktitle = {9th Biennial Conference on Innovative Data Systems Research}, year = {2019}, abstract = {As data processing evolves towards large scale, distributed platforms, the network will necessarily play a substantial role in achieving efficiency and performance. Increasingly, switches, network cards, and protocols are becoming more flexible while programmability at all levels (aka, software defined networks) opens up many possibilities to tailor the network to data processing applications and to push processing down to the network elements. In this paper, we propose DPI, an interface providing a set of simple yet powerful abstractions flexible enough to exploit features of modern networks (e.g., RDMA or in-network processing) suitable for data processing. Mirroring the concept behind the Message Passing Interface (MPI) used extensively in high-performance computing, DPI is an interface definition rather than an implementation so as to be able to bridge different networking technologies and to evolve with them. In the paper we motivate and discuss key primitives of the interface and present a number of use cases that show the potential of DPI for data-intensive applications, such as analytic engines and distributed database systems.}, language = {en} } @article{SkrzypczakSchintkeSchuett2019, author = {Skrzypczak, Jan and Schintke, Florian and Sch{\"u}tt, Thorsten}, title = {Linearizable State Machine Replication of State-Based CRDTs without Logs}, journal = {arXiv}, arxiv = {http://arxiv.org/abs/1905.08733v1}, year = {2019}, abstract = {General solutions of state machine replication have to ensure that all replicas apply the same commands in the same order, even in the presence of failures. Such strict ordering incurs high synchronization costs caused by distributed consensus or by the use of a leader. This paper presents a protocol for linearizable state machine replication of conflict-free replicated data types (CRDTs) that neither requires consensus nor a leader. By leveraging the properties of state-based CRDTs - in particular the monotonic growth of a join semilattice - synchronization overhead is greatly reduced. In addition, updates just need a single round trip and modify the state `in-place' without the need for a log. Furthermore, the message size overhead for coordination consists of a single counter per message. While reads in the presence of concurrent updates are not wait-free without a coordinator, we show that more than 97\% of reads can be handled in one or two round trips under highly concurrent accesses. Our protocol achieves high throughput without auxiliary processes like command log management or leader election. It is well suited for all practical scenarios that need linearizable access on CRDT data on a fine-granular scale.}, language = {en} }