• Journal of Platform Technology
• /
• v.11 no.4
• /
• pp.71-79
• /
• 2023

In this paper, we analyze the characteristics of machine learning workloads and, based on them, propose a distributed in-memory caching technique to improve the performance of machine learning workloads. The core of machine learning workload is model training, and model training is a computationally intensive task. Performing machine learning workloads in a Kubernetes-based cloud environment in which the computing framework and storage are separated can effectively allocate resources, but delays can occur because IO must be performed through network communication. In this paper, we propose a distributed in-memory caching technique to improve the performance of machine learning workloads performed in such an environment. In particular, we propose a new method of precaching data required for machine learning workloads into the distributed in-memory cache by considering Kubflow pipelines, a Kubernetes-based machine learning pipeline management tool.

• Journal of KIISE:Software and Applications
• /
• v.26 no.8
• /
• pp.974-987
• /
• 1999

이동 컴퓨팅 환경에서 자주 접근하는 데이터에 대한 캐싱은 무선 채널의 좁은 대역폭에서 경쟁을 줄일 수 있는 유용한 기술이다. 그러나, 트랜잭션 캐시 일관성을 지원하는 전통적인 클라이언트/서버 전략은 클라이언트와 서버간에 많은 양의 통신을 필요로 하기 때문에 이동 클라이언트/서버 컴퓨팅 환경에서는 적절하지 않다. 본 논문에서는 브로드캐스트-기반 캐시 무효화 정책을 사용하면서 트랜잭션 캐시 일관성을 지원하는 OCC-UTS (Optimistic Concurrency Control with Update TimeStamp) 프로토콜을 제안한다. 접근한 데이터에 대한 일관성 검사 및 완료 프로토콜은 캐시 무효화 과정의 내부 과정으로 완전 분산 형태로 효율적으로 구현되며, 일관성 체크의 대부분이 이동 클라이언트에서 수행된다. 또한, 분석 모델에 기반한 성능 비교를 통해, 본 논문에서 제안하는 OCC-UTS 프로토콜이 다른 경쟁 프로토콜보다 높은 트랜잭션 처리율을 얻으며, 데이터 항목을 자주 접근하면 할수록 지역 캐시를 사용하는 OCC-UTS 프로토콜이 더 효율적임을 보인다. 이동 클라이언트의 접속 단절에 대해서는 무효화 브로드캐스트 윈도우를 크게 하여 접속 단절에 적절히 대처할 수 있다.Abstract In a mobile computing environment, caching of frequently accessed data has been shown to be a useful technique for reducing contention on the narrow bandwidth of the wireless channels. However, the traditional client/server strategies for supporting transactional cache consistency that require extensive communications between a client and a server are not appropriate in a mobile client/server computing environment. In this paper, we propose a new protocol, called OCC-UTS (Optimisitic Concurrency Control with Update TimeStamp), to support transactional cache consistency in a mobile client/server computing environment by utilizing the broadcast-based solutions for the problem of invalidating caches. The consistency check on accessed data and the commitment protocol are implemented in a truly distributed fashion as an integral part of cache invalidation process, with most burden of consistency check being downloaded to mobile clients. Also, our experiments based on an analytical model substantiate the basic idea and study the performance characteristics. Experimental results show that OCC-UTS protocol without local cache outperforms other competitor protocol, and the more frequent a mobile client accesses data items the more efficient OCC-UTS protocol with local cache is. With respect to disconnection, the tolerance to disconnection is improved if the invalidation broadcast window size is extended.

• Proceedings of the Korean Information Science Society Conference
• /
• 2012.06d
• /
• pp.249-250
• /
• 2012

스마트 캐쉬 시스템은 콘텐츠 전달망에서 콘텐츠 트래픽 양을 감소시키기 위해 필수적인 기술로써 가입자 네트워크 구간에 콘텐츠를 자동 저장하여 중복 전송을 최소화 하는 시스템이다. 부하 분산 시스템을 통해 특정 스마트 캐쉬 서버로 집중되는 요청을 분산하고 콘텐츠 요청자의 주소에 따른 최적의 스마트 캐쉬를 지정하는 기술이다. 이를 통해 스마트 캐쉬를 적용한 글로벌 딜리버리 클라우드 플랫폼을 구축하여 콘텐츠 전송 플랫폼 기술의 국내 기업 경쟁력을 확보할 수 있고, 차세대 콘텐츠 유통 시장에서의 기술력 확보와 국내 업체의 세계 시장 지배력 강화할 수 있다.

• Journal of Convergence for Information Technology
• /
• v.10 no.1
• /
• pp.22-27
• /
• 2020

With the deployment of 5G infrastructure, content delivery network (CDN) will be a key role to provide explosive growing services for the independent media and YouTube which contain high-speed mobile contents. Without a local cache, the mobile backhaul and fronthaul should endure huge burden of bandwidth request for users as the increase number of direct accesses from contents providers. To deal with this issue, this paper fist presents both fronthaul solutions for CDN that use dark fibers and a passive optical network (PON). On top of that, we propose the aggregated content request specialized for PON caching and evaluate and compare its performance to legacy schemes through the simulation. The proposed PON caching scheme can reduce average access time of up to 0.5 seconds, 1/n received request packets, and save 60% of backhaul bandwidth compared to the no caching scheme. This work can be a useful reference for service providers and will be extended to further improve the hit ratio of cache in the future.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.16 no.6
• /
• pp.1121-1132
• /
• 2021

Nowadays, it is more difficult to find people who do not use mobile devices such as smartphones and tablets. Contents that can be accessed at the touch of a finger is overflowing. However, the existing network has a structure in which it is difficult to efficiently respond to the problems caused by overflowing contents. In particular, the bottleneck problem that occurs when multiple users intensively request content from the server at the same time is a representative problem. To solve this problem, the CCN has emerged as an alternative to future networks. CCN uses the network bandwidth efficiently through the In-Network Cache function of the intermediate node to improve the traffic required for user to request to reach the server, to reduce response time, and to distribute traffic concentration within the network. I propose a cache policy that can improve efficiency in such a CCN environment.

• The KIPS Transactions:PartB
• /
• v.9B no.5
• /
• pp.555-562
• /
• 2002

Increasing of electronic commerce users and high-capacity multimedia data is the main causes of increasing of load and network traffic in electronic commerce system. Increasing of load and network traffic brings the delay of service about client's request. And it's becoming factor that decreases customer satisfaction and corporation's competitive ability. So it is needed to research that consider efficient management and response time of electronic commerce system. In this paper, we designed electronic commerce system based on local server with web cashing function in order to solve load problem of electronic commerce system. And we proposed web cashing algorithm for local web server and analyzed performance by an experiment.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.37 no.8C
• /
• pp.670-679
• /
• 2012

The recent dissemination of cloud computing makes the amount of data storage to be increased and the cost of storing the data grow rapidly. Accordingly, data and service requests from users also increases the load on the cloud storage. There have been many works that tries to provide low-cost and high-performance schemes on distributed file systems. However, most of them have some weaknesses on performing parallel and random data accesses as well as data accesses of frequent small workloads. Recently, improving the performance of distributed file system based on caching technology is getting much attention. In this paper, we propose a CHPC(Cloud storage High-Performance Caching) framework, providing parallel caching, distributed caching, and proxy caching in distributed file systems. This study compares the proposed framework with existing cloud systems in regard to the reduction of the server's disk I/O, prevention of the server-side bottleneck, deduplication of the page caches in each client, and improvement of overall IOPS. As a results, we show some optimization possibilities on the cloud storage systems based on some evaluations and comparisons with other conventional methods.

• Journal of KIISE
• /
• v.42 no.9
• /
• pp.1109-1116
• /
• 2015

In this paper, we propose a query processing approach that uses the Spark functional programming and distributed memory system to solve the computational overhead of SPARQL. In the semantic web, RDF ontology data is produced at large scale, and the main challenge for the semantic web is to query and manipulate such a large ontology with a high throughput. The most existing studies on SPARQL have focused on deploying the Hadoop MapReduce framework, and although approaches based on Hadoop MapReduce have shown promising results, they achieve a low level of throughput due to the underlying distributed file processes. Therefore, in order to speed up the query processes, we suggest query- processing methods that are based on memory caching in distributed memory system. Our approach is also integrated with a clause unification method for propagating between the clauses that exploits Spark join, map and filter methods along with caching. In our experiments, we have achieved a high level of performance relative to other approaches. In particular, our performance was nearly similar to that of Sempala, which has been considered to be the fastest query processing system.

• Proceedings of the Korean Information Science Society Conference
• /
• 2002.04a
• /
• pp.619-621
• /
• 2002

90년대 중반에 등장한 그리드(GRID)는 지역적으로 분산되어 있으며 소유자가 서로 다른 다양한 컴퓨팅 자원의 효과적인 공유를 목적으로 하며, 시스템의 구축에 필요한 스케쥴링, 자원 관리, 보안, 성능 측정 및 상태 모니터링 등의 문제를 해결하기 위한 다양한 미들웨어 및 개발 도구가 연구되고 있다. 그리드 시스템에서 사용되는 데이터는 FTP 서버, 파일 시스템, 데이터베이스 등 여러 장치에 저장되어 개별적인 인터페이스를 통해 접근된다. 각각의 인터페이스론 통하여 접근하므로 사용 방법이 어렵고 확장성이 떨어지는 단점이 있다. 이러한 문제점을 해결하기 위하여, 본 연구에서는 데이터베이스, FTP, 파일 시스템에 산재한 그리드 데이터를 동시에 지원하고, URL을 통해 단일 인퍼페이스로 접근하게 해 주며, 3계층 구조와 데이터 캐시를 이용하여 성능과 확장성을 제공하는 시스템을 구현한다.

• Proceedings of the Korea Database Society Conference
• /
• 2001.11a
• /
• pp.536-554
• /
• 2001

인터넷 이용자의 증가와 멀티미디어 등 대용량 콘텐츠에 대한 수요의 증가로 인해 데이터 트래픽은 기하 급수적으로 늘어나는 데 비해 네트웍 성능향상에 대한 투자는 이에 미치지 못함으로서 결과적으로 인터넷은 트래픽의 병목현상, 즉 정체현상이 발생하게 되었으며 점차 그 정도가 심해지고 있다. 이 문제를 해결하기 위해 제안하는 방안이 바로 전략적 CDN(Content Delivery Network) 이다. 현실적으로, 본 연구는 물리적인 망 증설만으로는 트래픽 체증 문제를 해결하기는 어려우므로 주요 ISP 네트웍에 전략적 캐시서버를 설치하여 CP의 콘텐츠를 각 ISP 네트웍마다 분산하여 저장하고 이용자의 요청 시 이용자의 인터넷 회선이 물려 있는 ISP 네트웍에서 바로 콘텐츠를 전송하여 데이터 손실과 속도 저하 등의 문제를 해결하는 CDN의 구성과 그 효과를 실증적으로 밝혔다.