K
Kostas Katrinis
Researcher at IBM
Publications - 76
Citations - 1118
Kostas Katrinis is an academic researcher from IBM. The author has contributed to research in topics: Network topology & Optical burst switching. The author has an hindex of 17, co-authored 76 publications receiving 939 citations.
Papers
More filters
Journal ArticleDOI
A taxonomy of task-based parallel programming technologies for high-performance computing
Peter Thoman,Kiril Dichev,Thomas Heller,Roman Iakymchuk,Xavier Aguilar,Khalid Hasanov,Philipp Gschwandtner,Pierre Lemarinier,Stefano Markidis,Herbert Jordan,Thomas Fahringer,Kostas Katrinis,Erwin Laure,Dimitrios S. Nikolopoulos +13 more
TL;DR: This paper provides an initial task-focused taxonomy for HPC technologies, which covers both programming interfaces and runtime mechanisms and demonstrates the usefulness of the taxonomy by classifying state-of-the-art task-based environments in use today.
Journal ArticleDOI
Media- and TCP-friendly congestion control for scalable video streams
TL;DR: The algorithm integrates two new techniques: i) a utility-based model using the rate-distortion function as the application utility measure for optimizing the overall video quality; and ii) a two-timescale approach of rate averages to satisfy both media and TCP-friendliness.
Proceedings ArticleDOI
Rack-scale disaggregated cloud data centers: The dReDBox project vision
Kostas Katrinis,Dimitris Syrivelis,Dionisios Pnevmatikatos,Georgios Zervas,Dimitris Theodoropoulos,Iordanis Koutsopoulos,K. Hasharoni,Daniel Raho,Christian Pinto,Felix Espina,Sergio Lopez-Buedo,Q. Chen,Mario Nemirovsky,D. Roca,H. Klos,T. Berends +15 more
TL;DR: In this paper, the authors propose a highly modular software-defined architecture for the next generation datacentre, where SoC-based microservers, memory modules and accelerators are placed in separated modular server trays interconnected via a high-speed, low-latency opto-electronic system fabric, and be allocated in arbitrary sets, as driven by fit-for-purpose resource/power management software.
Proceedings Article
MiceTrap: Scalable traffic engineering of datacenter mice flows using OpenFlow
TL;DR: MiceTrap is proposed, an OpenFlow-based TE approach targeting datacenter mice flows that employs scalability against the number of mice flows through flow aggregation, together with a software-configurable weighted routing algorithm that offers improved load balancing for mice flows.
Proceedings ArticleDOI
Pythia: Faster Big Data in Motion through Predictive Software-Defined Network Optimization at Runtime
TL;DR: A system that reduces the skew impact of the communication-heavy phase of MapReduce by transparently predicting data communication volume at runtime and mapping the many end-to-end flows among the various processes to the underlying network, using emerging software-defined networking technologies to avoid hotspots in the network is presented.