BlockParty: Cooperative offsite backup among friends
نویسندگان
چکیده
We are keeping an increasing amount of valuable data in digital format only. Offsite backup systems are ideal for ensuring that such irreplaceable personal information will be preserved in the face of disk failures, unexpected disastrous events and operator mistakes. It is time to bring offsite backup solutions, traditionally a luxury to only deep pocketed enterprises, to the masses of home users today. There are two difficulties in building a cooperative backup system on a traditional peer-to-peer network. First, there is not enough bandwidth to repair lost data due to a high node departure rate [2]. The presence of frequent node departures is a consequence of users failing to display long term cooperativeness without proper incentives: after downloading a file, users have no good reason to continue participating in the network. Second, unlike file sharing, we cannot induce cooperative behavior in backup systems using tit-for-tat [3, 4]. Tit-for-tat requires that both peers involved an exchange have some resource that catches the other’s fancy. However, nodes that seek to restore data from others following a disk crash have no useful data for other peers and therefore are inevitably at the mercy of others’ altruism. Most existing peer-to-peer storage systems run under a centralized administration [1] so that any detected misbehavior or unwillingness to cooperate can potentially result in a peer’s expulsion from the system. Such a trusted central authority can be a serious nuance in situations when there is no natural central party trusted by all users. In BlockParty, a node’s owner explicitly specify a small number of overlay neighbors to perform cooperative backups with. Nodes that agree out-of-band to cooperative with each other are likely to be owned by users who have real-world friendships among them. In essence, each BlockParty node acts as its own centralized membership authority. BlockParty’s particular system structure brings us advantages as well as challenges. From a resource donor’s point of view, the ability to explicitly specify which other nodes she contributes her resources to gives a node’s owner better incentive to participate in the system over a long time and increase her node’s availability to others. From a resource recipient’s point of view, entrusting data only to those nodes’ whose owners are trusted “friends” allows each node to act as its own trusted authority to deter others from misbehaving and denying services with such credible threats that is the ire of friends in real life. For such credible threats to be effective, BlockParty cannot rely on transitive trust [5, 7], i.e. nodes never store data on those that they do not have a direct real world relationship with. We do not think this is overly restrictive because unlike peer-to-peer file sharing and live streaming applications, there is little value in making use a huge set of unreliable nodes to store one’s backup data. Restricting where data is stored limits BlockParty’s ability to utilize idle disk spaces in the system effectively. It is common that a node could not back up her data fully because none of its neighbors has any idle disk space remaining. BlockParty attacks this resource imbalance problem by having overloaded nodes store coded data blocks that belong to different neighbors, effectively increasing an overloaded node’s ability to store more information without literally requiring it to contribute more disk spaces. Although a node trusts its neighbors not to purposefully deny service at the time of restore, it still verifies the presence of its backed up data at neighbors periodically to be able to detect data loss and start the repair process promptly. The verification protocol must only involve transferring a small amount of data between nodes. Furthermore, it must accommodate coded information belonging to multiple nodes. BlockParty uses homomorphic hash functions [6] to allow a node to compute a secure hash value of any individual block from the hashes of the aggregate coded block and the hashes of remaining individual blocks.
منابع مشابه
AllNet: using Social Connections to Inform Traffic Prioritization and Resource Allocation
AllNet is a new networking protocol designed to provide communication utilizing all available means, including Internet and cellular communications, but when these are not available, also ad-hoc networking and delay-tolerant networking. These latter mechanisms are best for low-bandwidth communications. Effective support of low-bandwidth networking needs message prioritization, which can benefit...
متن کاملA SWOT Analysis of the Various Backup Scenarios Used in Electronic Medical Record Systems
OBJECTIVES Electronic medical records (EMRs) are increasingly being used by health care services. Currently, if an EMR shutdown occurs, even for a moment, patient safety and care can be seriously impacted. Our goal was to determine the methodology needed to develop an effective and reliable EMR backup system. METHODS Our "independent backup system by medical organizations" paradigm implies th...
متن کاملSecure Processing Offload in Recombining Media Segments for Mobile Access
We develop an architecture for a federation of home gateways. We distinguish inner and outer circle friends and describe a mechanism for social-aware backup of content and sharing with friends. To facilitate remote access, content can be segmented and redundantly stored at gateways of friends. We develop innovative solutions that allow secure and redundant storage of segments with inner and out...
متن کاملA Survey of Cooperative Backup Mechanisms
Storage capacity, like computing power, follows its Moore’s law and grows dramatically. Consequently, the need for data backup services increases. The domain of data backup was recently hit by the peer-to-peer wave: several projects propose to use a cooperative approach to the backup problem. In this paper we survey these projects, the various problems they face and the techniques they propose ...
متن کاملA Simulation-based Evaluation of a Hybrid Storage System Combining P2P, F2F, and Cloud storage with a Distributed Reputation System
As the amount of valuable data that the average person owns increases, there is a growing need for personal low cost backup services. A variety of methods have been developed to fulfill this need, from cloud based backup services to cooperative methods where users share spare resources to store each other’s data, either using a peer-to-peer (P2P) network to store data among a large number of di...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2007