Citation Context

... 31, 35, 38, 39] suggest that BFT may yield significant benefits even without resorting to n-version programming [4, 15, 32]. Third, improvements to the state of the art in BFT replication techniques =-=[3, 9, 10, 17, 32, 40]-=- make BFT replication increasingly practical by narrowing the gap between BFT replication costs and costs already being paid for non-BFT replication. For example, by default, the Google file system us...

Citation Context

...ers by Lamport, Pease, and Shostak [34, 45]. State machine replication [33, 49] is a classic technique for masking a limited number of such Byzantine faults. Byzantine fault tolerance (BFT) protocols =-=[13,47,56]-=- can mask faults as long as less than one-third of the nodes are faulty [9]. Aiyer et al. [1] introduced the BAR model, which can tolerate a limited number of Byzantine nodes plus an unlimited number ...

Citation Context

... type of sandboxing, ranging from fullscale emulators such as VMWare, to system call sandboxes [24], to narrowly applicable schemes such as StackGuard [13]. • Byzantine fault-tolerance schemes (e.g., =-=[34]-=-) which use voting among a number of service instances to select the correct answer, under the assumption that only a minority of the replicas will exhibit faulty behavior. The contribution of this pa...

Citation Context

...lica sends a message to all remaining replicas, stating their agreement on an ordering that was proposed by a primary replica. An important optimization decouples the agreement from request execution =-=[21]-=- reducing the number of the more expensive storage replicas to 2f + 1 but still retaining the quadratic communication costs. 8 Conclusions This paper presents HQ, a new protocol for Byzantinefault-tol...

Citation Context

...) or multiple administrative domains (potentially malicious operators, potential software attacks). 2.3 Notation For conciseness, throughout the paper we use the authentication notation of Yin et al. =-=[39]-=-, according to which we denote by 〈X〉S,D,k an authentication certificate that any node in a set D can regard as proof that k distinct nodes in S said X. For example, a traditional digital signature on...

Citation Context

... failures. The Aardvark protocol consists of 3 stages: client request transmission, replica agreement, and primary view change. This is the same basic structure of PBFT [8] and its direct descendants =-=[4, 18, 19, 33, 32]-=-, but revisited with the goal of achieving an execution path that satisfies the properties outlined in the previous section: acceptable performance, ease of implementation, and robustness against Byza...

Citation Context

...systems have defended against potentially malicious servers by replicating functionality and storage over multiple servers. Protocols, such as Byzantine fault tolerant (BFT) replicated state machines =-=[9, 21, 48]-=- or quorum systems [1, 26], can then guarantee safety and liveness, provided that some fraction of these servers remain non-faulty. Modern approaches optimize performance by, for example, concurrently...

Citation Context

...ior efforts have reduced trust assumptions in storage systems, they have not minimized trust with respect to safety, availability, or both. For example, quorum and replicated state machine approaches =-=[9, 24, 28, 29, 32, 41, 53, 54, 85, 95, 99, 100]-=- tolerate failures by a fraction of servers. However, they sacrifice safety when faults exceed a threshold and availability when too few servers are reachable. Systems like SUNDR [61] and FAUST [20], ...

Citation Context

...od performance even when faults occur. UpRight most directly works to combine the high performance agreement of PBFT [8] and Zyzzyva [18] with low-cost replication from separating order and execution =-=[34]-=- and with Aardvark’s robustness [11]. Notably, where Aardvark uses client signatures and Prime uses a 2 or 3 round pre-ordering stage that uses order node signatures to validate client requests, UpRig...

Citation Context

...yzantine fault tolerant (BFT) state machine systems has demonstrated that generalized state machine replication can be used to improve availability and reliability [22, 9, 21] as well confidentiality =-=[26]-=-. Furthermore, this work suggests that the approach has important practical properties in that it adds low overhead [9, 22, 26], can recover proactively from faults [10], can make use of existing off-...