A Heuristic Approach to Protocol Tuning for High Performance Data Transfers

Obtaining optimal data transfer performance is of utmost importance to today's data-intensive distributed applications and wide-area data replication services. Doing so necessitates effectively utilizing available network bandwidth and resources, yet in practice transfers seldom reach the levels of utilization they potentially could. Tuning protocol parameters such as pipelining, parallelism, and concurrency can significantly increase utilization and performance, however determining the best settings for these parameters is a difficult problem, as network conditions can vary greatly between sites and over time. Nevertheless, it is an important problem, since poor tuning can cause either under- or over-utilization of network resources and thus degrade transfer performance. In this paper, we present three algorithms for application-level tuning of different protocol parameters for maximizing transfer throughput in wide-area networks. Our algorithms dynamically tune the number of parallel data streams per file (for large file optimization), the level of control channel pipelining (for small file optimization), and the number of concurrent file transfers to increase I/O throughput (a technique useful for all types of files). The proposed heuristic algorithms improve the transfer throughput up to 10x compared to the baseline and 7x compared to the state of the art solutions.

[1]  Eitan Altman,et al.  Parallel TCP Sockets: Simple Model, Throughput and Validation , 2006, Proceedings IEEE INFOCOM 2006. 25TH IEEE International Conference on Computer Communications.

[2]  Ian T. Foster,et al.  A data transfer framework for large-scale science experiments , 2010, HPDC '10.

[3]  Tevfik Kosar,et al.  Energy-performance trade-offs in data transfer tuning at the end-systems , 2014, Sustain. Comput. Informatics Syst..

[4]  Tevfik Kosar,et al.  A highly-accurate and low-overhead prediction model for transfer throughput optimization , 2012, 2012 SC Companion: High Performance Computing, Networking Storage and Analysis.

[5]  Brian D. Noble,et al.  Adaptive data block scheduling for parallel TCP streams , 2005, HPDC-14. Proceedings. 14th IEEE International Symposium on High Performance Distributed Computing, 2005..

[6]  Tevfik Kosar,et al.  Dynamic Protocol Tuning Algorithms for High Performance Data Transfers , 2013, Euro-Par.

[7]  Mark Handley,et al.  Data center networking with multipath TCP , 2010, Hotnets-IX.

[8]  Miron Livny,et al.  Stork: making data placement a first class citizen in the grid , 2004, 24th International Conference on Distributed Computing Systems, 2004. Proceedings..

[9]  Tevfik Kosar,et al.  Network-aware end-to-end data throughput optimization , 2011, NDM '11.

[10]  Bing Zhang,et al.  StorkCloud: data transfer scheduling and optimization as a service , 2013, Science Cloud '13.

[11]  Ned Freed,et al.  SMTP Service Extension for Command Pipelining , 1997, RFC.

[12]  Mehmet Balman,et al.  Stork data scheduler: mitigating the data bottleneck in e-Science , 2011, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences.

[13]  Tevfik Kosar,et al.  Energy-aware data transfer algorithms , 2015, SC15: International Conference for High Performance Computing, Networking, Storage and Analysis.

[14]  Tevfik Kosar,et al.  Balancing TCP buffer vs parallel streams in application level throughput optimization , 2009, DADC '09.

[15]  Brian D. Noble,et al.  The end-to-end performance effects of parallel TCP sockets on a lossy wide-area network , 2002, Proceedings 16th International Parallel and Distributed Processing Symposium.

[16]  Tevfik Kosar,et al.  Hysteresis-based optimization of data transfer throughput , 2015, NDM '15.

[17]  Mehmet Balman,et al.  A new paradigm: Data-aware scheduling in grid computing , 2009, Future Gener. Comput. Syst..

[18]  Tevfik Kosar,et al.  End-to-End Data-Flow Parallelism for Throughput Optimization in High-Speed Networks , 2012, Journal of Grid Computing.

[19]  Ian Foster,et al.  GridFTP Pipelining , 2007 .

[20]  Tevfik Kosar,et al.  HARP: Predictive Transfer Optimization Based on Historical Analysis and Real-Time Probing , 2016, SC16: International Conference for High Performance Computing, Networking, Storage and Analysis.

[21]  Tevfik Kosar,et al.  A Data Throughput Prediction and Optimization Service for Widely Distributed Many-Task Computing , 2011, IEEE Transactions on Parallel and Distributed Systems.

[22]  Tevfik Kosar,et al.  How GridFTP Pipelining, Parallelism and Concurrency Work: A Guide for Optimizing Large Dataset Transfers , 2012, 2012 SC Companion: High Performance Computing, Networking Storage and Analysis.

[23]  Tevfik Kosar,et al.  Application-Level Optimization of Big Data Transfers through Pipelining, Parallelism and Concurrency , 2016, IEEE Transactions on Cloud Computing.

[24]  Mehmet Balman,et al.  Dynamically tuning level of parallelism in wide area data transfers , 2008, DADC '08.

[25]  Joel H. Saltz,et al.  Using overlays for efficient data transfer over shared wide-area networks , 2008, 2008 SC - International Conference for High Performance Computing, Networking, Storage and Analysis.

[26]  Miron Livny,et al.  Data placement in widely distributed systems , 2005 .

[27]  Ian T. Foster,et al.  Software as a service for data scientists , 2012, Commun. ACM.

[28]  Tevfik Kosar.  Data Intensive Distributed Computing: Challenges and Solutions for Large-scale Information Management , 2012 .

[29]  Tevfik Kosar,et al.  Prediction of Optimal Parallelism Level in Wide Area Data Transfers , 2011, IEEE Transactions on Parallel and Distributed Systems.

[30]  Peter A. Dinda,et al.  Modeling and taming parallel TCP on the wide area network , 2005, 19th IEEE International Parallel and Distributed Processing Symposium.