Suhyun Kim

Replica Placement Algorithm for Highly Available Peer-to-Peer Storage Systems

Peer-to-peer (P2P) technology is an emerging approach to overcoming the limitations of the traditional client –server architecture. However, building a highly available P2P system is quite challenging, in particular a P2P storage system. The reason is due to the fundamental nature of P2P systems: peers can join and leave at any time without any notice. Replication is one of the strategies in overcoming the unpredictable behavior of peers. A good replication algorithm should use the minimum number of replicas to provide the desired availability of data. The popular approach in the previous studies is a random placement of replicas, but it ignores the wide difference in the availability of each peer. In this paper, we develop a replica placement algorithm which exploits the availability pattern of each individual peer. By comparing our algorithm with a random placement scheme, we show that our algorithm dramatically improves the data availability with moderate overhead in terms of memory consumption and processing time in both ideal and practical conditions.

Real-time panoramic video streaming system with overlaid interface concept for social media

In this paper, we propose a real-time panoramic video streaming system with overlaid interface concept for social media. The proposed system provides real-time panorama images for smart displays such as smart TVs, smart phones, and tablet PCs. Panorama images are collected at sporting events via panorama cameras. Contents thus collected are sent to servers and then provided to smart devices via live sports video streams. Users select a panorama camera and choose their viewing angle and zooming factor for the selected panorama camera. The proposed system provides immersive and realistic views of live sporting events on users’ displays. Furthermore, an overlaid panoramic interface concept is proposed for immersive live baseball watching combined with tightly integrated social media experience.

Saveme: client-side aggregation of cloud storage

Nowadays, there are many cloud storage service providers and many of them provide significant amounts of free storage. It is desirable for users to be able to manage the storage space available to them as a single unified space. Otherwise, user files will be scattered over many different providers, negating the advantage of using cloud storage. Another important concern of users is that their data is at the disposal of the service providers. Lastly, in contrast to the hype of extremely high availability, accidents tend to happen and even the most available cloud services are not accessible many times per year or even per month. SaveMe is designed to solve all of these issues, providing aggregation, security, and higher availability purely from the client side. It is the first approach to address all three issues without any additional server component. To achieve these goals, SaveMe is building a single virtual volume by combining public and private cloud storage. In SaveMe, every file is fully encrypted, erasure coded, and distributed. For better security and privacy, all data processes are performed on the client side, unlike the case of conventional cloud storage services. The logical volume of SaveMe is extensible due to a flexible data and volume model. The authors also tested the performance of SaveMe in comparison with those of existing services. SaveMe showed above average performance in terms of the speeds of uploading or downloading files. Furthermore, data availability evaluation clearly demonstrates that the availability of SaveMe is an order of magnitude higher than those of other commercial cloud storage services.

SyncStream: Synchronized Media Streaming System in a Peer-to-Peer Environment

With recent advances in video encoding and network technology, media streaming services are getting popular. Media streaming system based on the traditional server/client or CDN architecture has a scalability problem, and peer-to-peer media streaming system using resources of peers in the system can provide better scalability. Previous media streaming services are suited to play stored contents such as dramas or movies, which are not live or real-time. However, if these services are used to provide live streaming such as broadcasting sports events, variation among play positions of peers in the service will be a problem. The users will demand simultaneity between viewers watching the same sports game through a real-time streaming service. To solve this problem, it is necessary to coordinate the behavior of peers participating in the service. In this paper, we propose SyncStream, a peer-to- peer media streaming system that synchronizes play positions of peers in the service. SyncStream provides synchronization in the same channel to help concurrent interaction among users. In our experiment, the difference in play positions of SyncStream nodes is 1 second at the maximum. This small gap among viewers will encourage group interaction and enrich user experience.

Classifying behavior patterns of user nodes

To increase the scalability of cloud computing, utilizing resources of individual users has been widely adopted especially in video streaming services. Accurately predicting behavior of user nodes is critical to achieve a high efficiency in such a peer-assisted system. Though there have been many measurement studies on peer-to-peer systems, most of them have focused on the design and characterization of the systems. Thus the behavior patterns of individual nodes have seldom been studied. In this paper, we present new techniques for classifying behavior of nodes in terms of availability and compare them with naive manual classification. We apply a k-means clustering algorithm with various classification criteria on real trace data of a peer-to-peer system. Our analysis shows that there are three dominant time zones with respect to the availability peak time. Our study will give a useful hint to a system designer in handling churns more efficiently based on the peer classification.

Mutual exclusion method in client-side aggregation of cloud storage

Consumers continue to produce and spend a huge amount of data using consumer electronic devices. To store consumers’ data, a huge volume of storage is needed. A cloud-of-clouds which is a virtual cloud storage service using heterogeneous public cloud storage services can be a solution. To support concurrent accesses for multiple consumers and devices, a mutual exclusion in a cloud-of-clouds is necessary. In this paper, we propose a mutual exclusion method for concurrent accesses in a cloud-of-clouds. The proposed method includes an atomic operation for cloud-of-clouds, a deadlock recovery and a race condition control.

User generated highlight system for baseball games with social media activities

In this paper, we propose a user-generated baseball highlight system. The proposed system provides environments for baseball highlight generation based on real-time text scoreboard with social media and live media. The proposed system creates short baseball highlights according to analysis of live baseball text scoreboard and social media. Users can generate baseball highlights with selections of short highlight clips.

Mutual exclusion strategy in a Cloud-of-clouds

A cloud-of-clouds is a virtual cloud storage service using heterogeneous public cloud storage services. To support concurrent accesses for multiple clients, a mutual exclusion is necessary in a cloud-of-clouds. In this paper, we propose a mutual exclusion strategy for concurrent accesses in a cloud-of-clouds. The proposed strategy includes an atomic operation for cloud-of-clouds, a deadlock recovery method and a race condition control method.

Coping with bad-mouthing in peer-to-peer file sharing networks

In the recent years, the P2P file sharing systems have adopted rating systems in the hope to stop the propagation of bad files. In a rating system, users rate files after downloading and a file with positive feedback is considered a good file. However, a dishonest rater can undermine the rating system by giving positive rating to bad files and negative rating to good files. In this paper, we design two filters based on probabilistic models such that the good files with negative feedback are not completely kept out of the system. The first filter is based on the binomial distribution of the ratings of a file, and the second filter considers the confidence of the downloading peer and the difference of positive and negative ratings of a file to calculate the probability to take a risk to download the file or reject it. Our filters only need the ratings of a file and this makes them suitable for popular torrent sharing websites that rank the files using a binary rating system without any information about raters. In addition, we can implement them entirely on the client side without any modification to the content sharing sites.

Speech-to-text-based life log system for smartphones

In this paper, we propose a life log system which provides a real-time voice recording with a speech-to-text feature over smartphone environments. The proposed system records data of user life using a microphone of smartphone. Recorded data are sent to a server, analyzed, and stored. Recorded data are dictated using a speech-to-text service, and saved as text files. The proposed system is implemented as a prototype system and evaluated. Users of our system are able to search their life log sound files using text.