Citation Context

...Because Depot does not require nodes to trust each other, different data centers in Figure 1 could be operated by different SSPs. Doing so might reduce the risk of correlated failures across replicas =-=[6, 38]-=-. For simplicity, we describe and evaluate only single-SSP configurations. the assignment of keys to volumes to the layers above Depot. The servers for each volume may be geographically distributed, a...

Citation Context

... Introduction Cloud storage provides an on-demand remote backup solution. However, using a single cloud storage vendor raises concerns such as having a single point of failure [3] and vendor lock-ins =-=[1]-=-. As suggested in [1, 3], a plausible solution is to stripe data across different cloud vendors. While striping data with conventional erasure codes performs well when some clouds experience shortterm...

Citation Context

...ew cloud provider, then it can revoke policy P . Thus, all files on cloud X will be deleted. We argue that switching cloud providers has its potential application, such as in avoiding vendor lock-ins =-=[1]-=-. 3 FADE OVERVIEW We now overview the design of FADE, a system that provides guarantees of access control and assured deletion for outsourced data in cloud storage. We present the necessary components...

Citation Context

...ouds Computing Security This section will discuss the migration of cloud computing from single to multi-clouds to ensure the security of the user’s data. 4.1 Multi-Clouds: Preliminary The term “multi-clouds” is similar to the terms “interclouds” or “cloud-of-clouds” that were introduced by Vukolic [54]. These terms suggest that cloud computing should not end with a single cloud. Using their illustration, a cloudy sky incorporates different colors and shapes of clouds which leads to different implementations and administrative domains. Recent research has focused on the multi-cloud environment [3],[8],[10],[11] which control several clouds and avoids dependency on any one individual cloud. Cachin et al. [11] identify two layers in the multicloud environment: the bottom layer is the inner-cloud, while the second layer is the inter-cloud. In the intercloud, the Byzantine fault tolerance finds its place. We will first summarize the previous Byzantine protocols over the last three decades. 4.2 Introduction of Byzantine Protocols In cloud computing, any faults in software or hardware are known as Byzantine faults that usually relate to inappropriate behavior and intrusion tolerance. In addi...

Citation Context

...g., [7], [9], [13]) evaluate random linear codes for peer-to-peer storage from a practical perspective, but they are mainly based on simulations without actually deploying a real storage system. RACS =-=[1]-=- and DEPSKY [4] are cloud storage proxies that interconnect multiple cloud storage providers, and they are built on traditional erasure codes. To our knowledge, this paper is the first work that evalu...

Citation Context

...storage providers (all offering a KVS interface), using the guarantees offered by robust distributed storage algorithms [21, 8]. Data replication across different clouds is a topic of active research =-=[2, 12, 36, 10]-=-. 1.2 Problem Our data replication scheme relies on multiple providers of raw storage, called base objects here, and emulates a single, more reliable shared storage abstraction, which we model as a re...

Citation Context

...yers say about Byzantine agreement being too paranoid, too hard, or too abstract, BA is the right problem formulation for building secure networks. To see this, consider the many systems papers (e.g. =-=[1, 50, 19, 26, 21, 24, 22, 55]-=-) from the recent years that say that a key component of what they want to do requires BA. The problem is the solution. Or rather the lack of a solution. We still do not have a general algorithm to so...

Citation Context

...the common case. They also show that storing two copies is necessary, in order to achieve wait-free client operations. Data replication in the intercloud has recently received a lot of attention [4], =-=[5]-=-, [6], [7]. One class of such systems assume specialized interfaces on the storage replicas, which are capable of limited processing and command execution (like active disks [8] or the storage servers...

Citation Context

...orage can suffer latency spikes even when there is no spike in that application’s workload. Combining cloud providers. Others have combined the use of multiple cloud providers to improve availability =-=[11, 28]-=-, to offer more secure storage [17], and to reduce cost [39, 41]. CosTLO uses cloud storage services offered by multiple providers because 1) the combination offers more data center pairs that are clo...