SOSCON 2019

SCP(Samsung Cloud Platform)는 Kubernetes를 Telco 환경에 사용하기 위해 CNI 및 EPA (Enhanced Platform Awareness) 기능을 최적화한 solution입니다. 이 세션에서는 Telco Application을 Cloud Native 하게 만들기 위해 Infrastructure 에서 준비해야 할 내용들을 공유하고자 합니다.

Presenter Jonghan Park(Samsung Electronics)

삼성전자 네트워크 사업부 수석연구원
2014년부터 Telco 환경에 적합한 On-Premise Cloud 를 개발해 왔으며, 최근에는 Cloud Native 구조를 Telco 에 접목시키는 일을 하고 있습니다.

The Fourth Industrial Revolution is also the Data Revolution. The lifetime data of many things, for example of tuna caught in the Pacific Ocean and served on our table or the lifetime data of a TGV train from France to Switzerland, is recorded and shared in the international standard language GS1. The Oliot open-source used to implement the GS1 international standard is the data-based IoT platform that can be utilized in more than 25 global industries. This presentation introduces the data platforms constructed by many countries, international organizations, and global enterprises using the GS1 international standard.

PresenterDaeyoung Kim(KAIST)

2005년부터, GS1 국제표준기구의 사물인터넷 국제공동연구 컨소시엄인 Auto-ID Labs(KAIST, MIT, Cambridge, ETH Zurich, Fudan, Keio)의 멤버로 GS1 국제표준화 작업과 함께 Oliot오픈소스 프로젝트를 리딩하고 있습니다.Oliot은 국제식품기구와 세계자연기금의 수산물/해양생물의 이력추적 프로젝트부터 스마트시티에 이르기까지 다수의 글로벌 프로젝트에서 활용되고 있습니다.

In this session, I would share the technical issues that I faced while operating Apache Openwhisk in the production environment for lasat 3 years and lesson-learned from the experience while contributing to the Apache OpenWhisk project.

* A Fun happening while contributing to Apache OpenWhisk
* Docker networking issue on Docker for Mac
* CouchDB performance issue
* Peroformance issue of Docker daemon
* Performance improvement of Apache OpenWhisk

PresenterDongkyoung Kim(NAVER)

Dominic is on a mission to enable and enlarge the serverless platform at Naver. As a senior software engineer, he designed the overall architecture to build a highly scalable and fault tolerable system based on Apache OpenWhisk. With nearly a decade of experience in cloud computing at Samsung and Naver, he gained expertise in the container, distributed system, and especially on building and operating a large scale system. He is also actively participating in Apache OpenWhisks project, and he recently became a committer and a member of PMC in recognition of his contribution to the performance area.

The infrastructure is also developing into a cloud-native service form following the technological advancement in network virtualization, MEC, and media. Furthermore, the demand for new services has grown in the 4G era. As such, SKT has been developing cloud-native infrastructure technology that combines OpenStack and Kubernetes since 2016 and has actively contributed codes in the global open-source software projects such as OpenStack-Helm and Airship. TACO (SKT All Container Orchestrator) is the "declarative infrastructure delivery platform” developed by SKT and based on such technologies. It distributes and manages the containerized infrastructure software like cloud-native applications using the life cycle management function of Kubernetes. SKT opened TACO in open-source software ([https://github.com/openinfradev](https://github.com/openinfradev)). This presentation begins with a technical description and a simple demonstration of TACO and discusses 1) various technical issues and how to resolve them through the commercial application of TACT to the SKT networks, IT environments, and external projects, 2) a description of the CI pipeline architecture and technology that we built while creating the open-source software based solutions, and 3) a description of new technology patterns introduced to the cloud-native ecosystem such as CRD, operator, and cluster-API and the implementation plan for managing the infrastructure delivery & lifecycle in the new form based on them.

PresenterJaesuk Ahn(SK Telecom)

As the Cloud Native Development Team Manager at SK Telecom, Jaesuk Ahn is leading the development of cloud platforms based on open-source software technology such as OpenStack and Kubernetes. He has been participating in the OpenStack project since the very beginning (2010) and has founded an OpenStack community in Korea and helped to grow it into Korea’s leading open-source software community. Since 2017, he has launched the OpenStack-Helm and Airship projects as a joint development with AT&T and developed and commercialized TACO (SKT All Container Orchestrator). He received an MS and PhD from the University of Texas at Austin and was in charge of cloud platform development based on KT open-source software in Korea. He strives to promote the open-source software ecosystem in Korea and abroad as an active User Committee member of the OpenStack Global Community and as the coordinator of the OpenStack Community and Kubernetest Community in Korea.

Operators, Service Mesh, and Serverless are technologies that are gaining traction in the industry quickly and are set to define components of next-generation, containerized application development and deployment. With Operators, customers can build business logic into applications for enhanced deployment and management. Service mesh enables communications infrastructure between services deployed inside individual containers. Serverless enables new compute consumption models with minimal overheads. In this session, we’ll describe how these technologies help you build and deploy faster and more consistently.

PresenterGyuseok Lee(Red Hat)

Gyuseok Lee is a Specialist Solution Architect at Red Hat Korea, responsible for the OpenShift container platform and middleware. He is thinking and suggesting how to present realistic and sympathetic solutions, and he shares his knowledge and practices on the latest trends DevOps, MSA and Cloud-Native development. Prior to joining Red Hat, he worked on MSA-based server development and PM. He conveys the problems that need to be solved in that direction and what to consider. In addition, based on his previous work as a WEB / WAS field engineer, he identifies obstacles, designs and proposes solutions when transitioning from the existing multitier service environment to the new paradigm.

What are the methods for packet processing on Linux? And how fast are each packet processing methods? In this presentation, we will learn how to handle packets on Linux (User space, socket filter, netfilter, tc), and compare performance with analysis of where each packet processing is done in the network stack (hook point). Also, we will discuss packet processing using XDP, an in-kernel fast-path recently added to the Linux kernel. eXpress Data Path (XDP) is a high-performance programmable network data-path within the Linux kernel. The XDP is located at the lowest level of access through SW in the network stack, the point at which driver receives the packet. By using the eBPF infrastructure at this hook point, the network stack can be expanded without modifying the kernel.

PresenterDaniel T. Lee(Kosslab)

Daniel T. Lee currently works as Software Engineer at Kosslab and contributing to Linux kernel BPF project. He has interest in cloud, Linux networking, and tracing technologies, and likes to analyze the kernel's internal using BPF technology.

3D / VR engines for game and virtual reality development in commercial, educational and research markets are converged to the Unity3D and Unreal, the majors of the current authoring market. Some content creators with development capabilities are using their own development engines but the numbers are shrinking. In addition to these popular commercial SW, Godot, Torque, Xenko, and Ogre are used. The Danuri engine currently being developed is not intended to be another engine in this group of open SW engines, but to be the easy and friendly tool like Unity3D, but with open sources. allowing content developers to modify the tool and engine thus enabling the generation of another tools based on our engine and authoring tool. We are working to provide usability that is close to the Unity3D, and the sources arerelatively easy to understand and could be used as base code for the university student, and commercial services. Codefarm, coding education software for young children, consisting of Scratch like programming interface with 3D graphical user interface was developed on the Danuri engine, and currently distributed by thje ATC, teacher’s association of coding education and Educational Broadcasting System. Since 2017, it has been released at www.danuri-vr.org under a license similar to the MIT type license, aka Danuri license.

PresenterSeongwon Ryu(ETRI)

Ryu Seong Won, Principal Researcher at Korea Electronics and Telecommunications Research Institute, has conducted various projects related to virtual reality engines, content authoring tools, and real-time 3D graphics technology. He has been researching the virtual reality environment where the virtual world interacts with the participants through the various virtual reality devices under the specific contextual ways, and the field that enhances the effect of the experience by using them. He is interested in developing the VR system, incorporation the VR engines wirh authoring environments and the user interfaces, and hopes to distributes the system in open software world.

CUBRID is the system software based on participation, openness, and sharing. It was switched to the open-source software for the first time in Korea in 2008. We strive to spread users and promote projects for them. Working with GitHub, CUBRID encourages the participation and development of domestic and foreign developers. We are creating the virtual cyclic structure for the software ecosystem by supporting the developers’ communities and collaborating with third-party solutions. Since the switch to open-source software, the CUBRID product has been downloaded more than 300,000 times and has been applied to more than 600 services of the clouds in the public and defense sectors. Although it was until recently only used for constructing the software ecosystem in Korea, it has enhanced its brand awareness abroad since it was selected as one of “16 Powerful Open-Source DBMS” by a global IT periodical. We plan to begin the community activities to expand it to a global ecosystem.

PresenterHyunsuk Jang(CUBRID)

공공 및 민간부문에서 오픈소스SW인 CUBRID DBMS의 확산과 발전을 위하여 활동하고 있습니다.

Concerns about the risks that the Internet of Things has posed on personal privacy have been raised for a few years. Data protection legislations, such as GDPR, advocates "privacy by design" and encourages us to re-prioritize digital privacy when deploying technologies and building services. One promise of the open web platform is enhanced privacy in comparison to other technology stacks. This talk will walk you through the end to end IoT system we have built onsite using web technologies, from new device on-boarding to remote access. At each stage, privacy considerations will be addressed and our solutions will be explained. To address the issue of resource restriction in constrained IoT devices and create REST Web interfaces, we have ported Web of Things APIs from Node.js to IoT.js, a JavaScript platform based on the ultra-light JerryScript engine by Samsung. The open gateway framework is node.js based and targets at decentralized ‘Internet of Things’ with privacy and security in mind. To have digital replica of a “Thing” using Web, we have introduced “Web of Twin” concept based on Digital Twin. This concept is explained through a smart home demo using Web of Things, Immersive Web and PWA technologies. At network infrastructure aspect, we will discuss and demo our implementations of Federated Learning, a Collaborative Machine Learning Framework. With Federated Learning in place, privacy is preserved by keeping all the user data on local device and only sending updates of “model” to server.

PresenterZiran Sun(Samsung Research, UK)

Ziran Sun is an Open Source Engineer at Samsung Research UK. She has been involved in a few open source projects, including webinos, blink/chromium and IoTivity. Currently, Ziran is looking at 5G network related area.

The eye problem is getting worse. According to the American Academy of Ophthalmology, half of the entire population is expecting to have eyeglasses in the near future. Following to this, the Korean National Health Insurance Service also mentioned that the number of pediatric patients with amblyopia is steadily increasing and this is occured by refractive errors which is the cause of higher rate of eyeglasses. However, compare to the increasing incidence of pediatric amblyopia, eye test for children is still a challenge around the world. Amblyopia can be prevented by 95% if the disease is detected before 4, but if the treatment timing is missed as the child is older than the age of 8, the cure rate rapidly decreases to 23% which leads to a lifetime suffer that the child's vision cannot be corrected by glasses. Although the test for amblyopia can be done by a simple early detection, the fundamental problem is that children younger than the age of four mostly do not have the ability to distinguish letters and are not very cooperative. There are pediatric eye test devices, but it is very hard for private eye clinics or developing countires where has the medical imbalance to afford these devices due to high cost. To this end, we have developed an innovative solution that enables eye test with smartphone which is closest to our living life in these days, and we do not only want to provide pediatric eye test in our country, but also want to create an opportunity to protect children's eye health in developing countires. This innovative solution is called KIZOPTER and has been researched, developed, and produced by using open source and deep learning. I would like to introduce our story that happened during the process of producing and twits and turns as as junior developers. And as well as the stories of my pioneering the DevOps environment which was used by the R&D department.

PresenterZgero Yang(Pixel Display Inc. Research & Development Dept)

Currently, I have been working on R&D at Pixel Display for 2 years. I completed the Global School of Media Studies at Soongsil University and worked as a designer at an entertainment company and as a video producer for about three years. Although open source will be limited to programming, I always study to foster wide coverage with the idea that the spirit of sharing and win-win development of open source will be gradually advanced to all fields.

We are at this moment using cloud services provided by cloud providers like Amazon, Google, and MS. How are the cloud services that we use configured? How are they operated? NHN TOAST services public and private clouds with OpenStack. This presentation describes how we implement public clouds with open-source, including OpenStack, and shares various experiences of operating them.

PresenterSeongsoo Cho(NHN Cloud System Team)

Seongsoo Cho is operating the TOAST IaaS service and infrastructure at NHN. He is also working to promote Korea's OpenStack ecosystem as an active member of the OpenStack Korean User Group.

The reinforcement learning algorithm itself is important in the study of reinforcement learning, but it is also very important to construct an environment for testing the reinforcement learning algorithm. For example, if you want to apply the reinforcement learning algorithm to the drone simulation environment, you need a drone simulator. However, developing this kine of simulator is very difficult. In particular, simulators in 3D environments or simulators with physical models are more difficult to implement. However, in September 2017, with the release of Unity ML-agents, it became much simpler to implement a game and simulation environment to test reinforcement learning algorithms. Unity makes it easy to implement 3D environments with physical models. Also, ML-agents simplify the communication between the environment and reinforcement learning algorithms. This presentation introduces Unity ML-agents and provides examples of reinforcement learning environments implemented with Unity ML-agents.

PresenterKyushik Min(PhD Candidate of Hanyang University)

Currently, I am PhD candidate at automotive engineering department of Hanyang University. I am conducting research on autonomous vehicle and driver assistance system. Also, I am studying deep learning, reinforcement learning and Unity. I am working as one of the Unity Masters which is a Unity expert group certified by Unity Korea. I won the Creative Application Award at the Unity ML-agents Challenge which is hosted by Unity. Also, I developed simulators with Unity ML-agents and published journal and conference paper using the simulator. In addition, I have conducted a number of seminars about Unity ML-agents.

Analyze and visualize several datasets provided by Seoul Public Data Portal with Elastic Stack. Get stats of people movement, by timestam, stations, locations and so on and see what kind of insights we can earn from this study.

PresenterJongmin Kim(Elastic)

Jongmin Kim is a Tech Evangelist at Elastic. He runs an Elasticsearch technology community, and communicates with developers through various publications, seminars, and presentations.

This presentation shares the experiences of using ShardingSphere, which was recently raised to the Apache project, and describes how we can use it in various ways. I am not a developer but an open-source data engineer who must code programs when needed. My presentation covers how we have solved and continue to solve problems in unique ways.

PresenterChan Sung(kakaobank)

Engineer for perhaps Korea’s first open-source database for the bank that aims to provide a robust service that never breaks down.
KT Hitel > TMON > Kakao > Kakao Bank (Current)

In a future society driven by the Fourth Industrial Revolution (4IR), the automation would create a whole new set of jobs with computing technology combined with every area in a society. In this presentation, I would like to envision an ever more diverse society with new lifestyles and to touch upon what we would need in that society. In particular, the presentation focuses on the value of SW education in an automated society by computing technology.

※ Time : 14:30 ~ 15:00

PresenterDai Young Kwon(Research Professor, Korea University)

Professor Kwon supported projects of the Junior SW Heroes.

In deep learning production, engineering art that is 5-10 times the model code may be necessary, and as such various open-source ecosystem exists to support them. This presentation reviews the stable, scalable, and advancing AI service design methodologies and the infrastructure elements for them, such as AI DevOps, AI BackOffice, CI/CD fault-tolerant model distribution, and auto scale out/scale down techniques and introduces the optimization tools for the cloud and open-source.

PrsenterHoondong Kim(SK Telecom)

AI Product DevOps Team Leader, AI Center of SK Telecom
2016 - 2017. Received MVP (Most Valuable Professional) in BigData from Microsoft
2018 - 2019. Received MVP (Most Valuable Professional) in AI from Microsoft
Operating Committee member of the Korea Spark User Group
Hoondong Kim is currently contributing to the communities related to the production of BigData scale deep learning infrastructure as well as distributed training/distributed serving.

As the requirement for the distributed cloud infrastructure environment like the Edge Cloud has increased, declarative infrastructure management is becoming more important. We can distribute and manage the bare-metal hosts in Kubernetes Native using Metal³.
This presentation discusses the Kubernetes Native technologies such as cluster-api and bare metal operators used in Metal³ projects like the OpenStack ironic projects used for provisioning bare metal hosts.
This presentation covers cluster-api, bare metal operator, and ironic.

Presenter한승진(SK Telecom)

He has developed OpenStack programs and implemented and operated private cloud platforms. He is currently developing the TACO service, which is a Kubernetes-based solution at SK Telecom. Utilizing his experience of operating open-source-based CEPH, he is also an operating member of the Ceph Korea User Group, which is a Facebook page.

The software is designed to improve a delivery review system through machine learning and natural language processing and to run RUR-PLE on the system.

※ Time : 15:10 ~ 16:10

PresenterDong-yoon Jung(Chadwick International School)

A winner of the Junior SW Heroes.

The platform-based software is designed to provide feedback on therapeutic postures at home./ (If the technlogy is shown to the audience) The software is designed to analyze and correct postures for rehabilitation using Tensorflow 2D Pose Estimation.

※ Time : 15:10 ~ 16:10

PresenterMin Kim(Maesong Middle School)

A winner of the Junior SW Heroes.

PresenterMin Seo Kim(Phillips Exeter Academy)

A winner of the Junior SW Heroes.

PresenterHo-jun Youn(homeschooling)

A winner of the Junior SW Heroes.

The software is designed to provide safety information by combining public data on natural disasters and the do's and don'ts to remember.

※ Time : 15:10 ~ 16:10

PresenterHa-eun Park(Changshin Elementary School)

A winner of the Junior SW Heroes.

PresenterSu-ha Lee(PohangJecheol(Posco) Elementary School)

A winner of the Junior SW Heroes.

Open Network Automation Platform (ONAP) as Virtual Network Function Orchestrator is going to play a key role in the security of whole 5G deployment. That's why we decided to start security penetration testing of this open source project at its very early stage. This talk is a summary of first ever ONAP penetration testing and its influence on the ONAP community.

PresenterKrzysztof Opasiak(Samsung Research)

Krzysztof Opasiak is a PhD student at Warsaw University of Technology. He works as Open Source Developer at Samsung R&D Institute Poland. Initially involved in Linux Kernel and libusbgx. Now focused on OpenStack and ONAP. Open Source enthusiast and speaker at several Linux and Open Source Conferences.

Nowadays many services are being built using Micro Service Architecture and moving to Cloud Native, In the field of monitoring, the word observability is increasingly mentioned. We will introduce what we are doing to ensure observability of over 10,000 service instances in Naver. and show how Pinpoint is used to solve problems in a distributed environment. This talk will also cover how Pinpoint, a global open source project, is evolving.

PresenterSungwook Kim(NAVER)

Sung Wook Kim is a software engineer working at Naver, since 2017. Currently, one of the maintainers in open-source project Pinpoint. He has many interests in the open source community and loves to meet new people. Also participated as a presenter in a number of national and international open source conferences. Before this, he worked for seven years at Samsung.

PresenterHyungil Jeong(NAVER)

Pinpoint maintainer since 2015, a software engineer at Naver trying to make life a bit easier for developers around the world. Based on his experience with open source Pinpoint, he has been a presenter at several national and international open source conferences.

The software is designed to classify speakers' emotions through machine learning and natural language processing.

※ Time : 16:30 ~ 17:10

PresenterTae-uk Gang(Korea International School, Jeju Campus)

A winner of the Junior SW Heroes.

The software is designed to provide pubic data on water quality in a practical and handy way.

※ Time : 16:30 ~ 17:10

PresenterMin-kyum Kim(Yeongnam Middle School)

A winner of the Junior SW Heroes.

PresenterJi-hwan Kim(Yeongnam Middle School)

A winner of the Junior SW Heroes.

PresenterHyeon-sung Jo(Yeongnam Middle School)

A winner of the Junior SW Heroes.

In order to develop competitive software, it is necessary not only to actively use various open source SWs, but also to contribute to open source SW development. In particular, we can secure corporate technical capabilities, enhance our corporate image, and gain new opportunities based on community capabilities. Samsung SDS is making various efforts to participate in the open source SW ecosystem. Among them, Open Source SW, which can measure various quality of SW called Code Analyst, has been released since 2018, and it is released as Maven Central Repository for various applications in other open source SW. Maven Central Repository is a service that makes it easy for anyone to use Java libraries of various open source SWs around the world, but it requires a lot of preparation to publish libraries in Central Repository. In this session, you will learn how to prepare your library for distribution through OSS Repository Hosting (OSSRH) and how to release it through continuous integration, giving you a deep understanding of the overall structure and methodology of the Central Repository.

PresenterSeongkon Han(Samsung SDS)

I am working on Samsung SDS to establish code quality metrics and develop measurement tools, and contribute some of the development tools through GitHub. In addition, I am involved in various activities to improve code quality(refactoring / clean code training, conducting code review and inspection).

Debezium(https://github.com/debezium/debezium) is an open source(Apache License 2.0) project that provides a low latency data streaming platform for CDC(change data capture). Debezium is a set of distributed services that capture row-level changes in your databases so that your applications can see and respond to those changes. Debezium is durable and stores the history of data changes in the replication log and can be stopped and restarted by the application. He describes existing CDC technologies, Debezium architecture, Kafka / Kafka Connect / Schema registry / Mysql database / Kubernetes integration, and Debezium advantages and disadvantages.

PresenterSamuel Kim(KakaoCommerce)

He works as a Developer / Devops in Kakao(Kakao Commerce) through Alticast, Naver, Line, and SK Planet. Now a forty-four-year-old average developer, he gets energy from a variety of challenges and records his development and practical experiences on Kim Yong-hwan's blog (http://knight76.tistory.com). Still looking at the new open source, I live with the belief that "new technologies make the development and operating environment better." He have shared development-related content in the Software Engineering Portal under the NIPA, as well have been publishing Java and DevOps related to Deview and many other conferences / seminars. Whenever he is working, always he feels of inadequacy in knowledge. So He started to translate when I was forty-year-old and already translated 15 books in Korean. He is currently translating ‘Docker In Practice 2nd’ and ‘Seeking SRE’ in Korean.