Minkyong Kim (김민경)

Samsung Electronics
minkyong.kim @ samsung.com


  • Service Automation

    I have been working on analytics to improve service automation for IBM Technical Support Services (TSS). I received "2015 IBM Technical Accomplishment" for this project.
  • IBM Bluemix: Platform as a Service on Cloud (PaaS)

    I have been working on infrastructure management and monitoring for Bluemix. I'm also working on providing various services (e.g., BigInsights Hadoop) on Bluemix. I gave a guest lecture at Carnegie Mellon University (CMU) on "IBM Bluemix Architecture" (April 22, 2015). I hosted a tutorial on "IBM Bluemix" at IEEE International Conference on Cloud Computing (IEEE Cloud) (June 27 - July 2, 2015). I also hosted a hands-on session at IBM Pulse (February 23-26, 2014).
  • Analytics Platform for Cloud

    I'm working on designing and developing IBM Cloud Analytics Application Services (ICCAS) for enterprise customers (2014-2015). I created an offering of BigInsights Hadoop as a Service and also added dynamic cluster resource management capabilities. In addition to BigInsights Hadoop, I also created an offering of Hortonworks. These on-demand services are managed using IBM Platform Cluster Manager Advanced Edition (PCMAE) on SoftLayer. (Paper: Computing Frontiers 2015)
  • Big Data Analyltics

    I'm leading a project on big data analytics for retail industry. By analyzing fashion data, our system can make outfit recommendations and fashion trend forecasting.
  • Open Collaboration Research (OCR)

    [2013-2014]: I'm collaborating with Prof. Brian Noble at the University of Michigan, under IBM's OCR program, focusing on cloud data center management. (Paper: W-PSDS 2014)
  • Cloud management

    [2011-2013]: In 2011, I was involved in benchmarking various approaches for managing virtual machine (VM) images. (See Middleware2012.) In 2012, I led Cloud Benchmarking project, with focus on IaaS. The goal was to benchmark various cloud management systems, including OpenStack. As the project lead, I played the major role in identifying the key metrics, designing the experiments and analyzing the results. I presented the finding to IBM executives, and was invited to OpenStack Grizzly Summit 2012 to present the results to the broader community. The project successfully provided the guidance for IBM and OpenStack communities for possible future releases of cloud management systems. In 2013, I led the benchmarking of PaaS layer. The results were used to develop the performance models of various cloud applications.

    Besides the cloud benchmarking efforts, I was also involved in the study of similarity among VM images. (See Middelware 2011.) Based on the understanding of VM images, I led the design and development of a p2p-based VM image distribution system for cloud. (See INFOCOM 2012.)
  • Software license discovery for cloud

    [2011-2012]: I led the software license discovery project. I designed and developed a new software license technology capable of addressing major limitations with license entitlement tracking in the cloud. (See ICSOC 2012.)
  • QoS-aware messaging

    I worked in the Responsive, Reliable and Real-Time (R3) Messaging project. This involved designing and building a messaging middleware to deliver messages in wide-area networks while meeting QoS requirements. There have been a lot of news coverages on this project. the New York Times, Wall Street Journal, and CNN. (See Middleware 2010, Middleware 2009 and ICSOC 2009 papers for details.)
  • Scalable services for cloud

    I led the team of researchers to develop a scalable queueing service for cloud. Together we developed a scalable queueing service using a DHT overlay. I also worked on experimenting different consistency levels for a scalable cloud-based queueing service. (See ICSOC 2010.) I was also involved in designing and developing a scalable publish/subscribe system, built on top of Cassandra. (See IPDPS 2011.)
  • Stream processing

    I worked on the Distillery project, designing the distributed version of the System S Stream Computing System. I also worked on enhancing high availability of the system. (See Middleware 2009.) I was also involved in experimenting multipath for real-time streaming applications. (See ICNP 2009.)
  • Mobility models and characterization

    Developing mobility models is challenging. To simulate mobile users (especially for testing wireless networks), it is important to have a realistic mobility model. The commonly-used random waypoint model doesn't reflect real user mobility. Based on the traces collected at Dartmouth College, I developed a realistic model focusing on the arrival and departure rates of hotspots. (See INFOCOM 2006.) I was also involved in developing another model, based on the transition probabilities (at buildings and intersections) extracted from the traces. (See MobiSys 2006.) I also worked on discovering mobility patterns by studying real user traces. I worked on a large data set to discover interesting mobility patterns. I had fun applying Fourier transform and playing with MATLAB. (See PUC 2007, WiTMeMo 2005.)
  • Localization

    I'm generally interested in location sensing and discovery technologies. I looked into the accuracy of the AP locations discovered through war driving or war walking. (See Pervasive 2006.)
  • Distributed file systems

    I worked on designing and developing a distributed file system for wide-area networks. The focus was to help mobile users to access remote files fast when they are traveling. (See Fast 2002 and NetStore 1999.)
  • Mobile network estimation

    I worked on estimating bandwidth and latency in wireless networks. It is always interesting to adapt existing techniques (such as the Kalman filter) and find how well they address the new problems. (See MobiCom 2001)