Filters








33 Hits in 3.7 sec

Near-Optimal Latency Versus Cost Tradeoffs in Geo-Distributed Storage

Muhammed Uluyol, Anthony Huang, Ayush Goel, Mosharaf Chowdhury, Harsha V. Madhyastha
2020 Symposium on Networked Systems Design and Implementation  
We show that the key to addressing this sub-optimality is to 1) allow for erasure coding, not just replication, of data across data centers, and 2) mitigate the resultant increase in read and write latencies  ...  However, when sacrificing data consistency is not an option, we show that service providers have to today incur significantly higher cost to meet desired latency goals than the lowest cost theoretically  ...  3: Example execution of RS-Paxos on an erasure-coded Figure 4 : 4 Figure 4: (a) Reusing read quorums in Phase 1 of writes enables reduction in read latency without impacting (Phase 1 + Phase 2) latency  ... 
dblp:conf/nsdi/UluyolHGCM20 fatcat:erjb326xg5azdiferrtw2seimu

RMWPaxos: Fault-Tolerant In-Place Consensus Sequences [article]

Jan Skrzypczak, Florian Schintke, Thorsten Schütt
2020 arXiv   pre-print
The protocol builds a fault-tolerant atomic register that supports arbitrary read-modify-write operations.  ...  Reading a consolidated consensus is done without state modification and is thereby free of concurrency control and demand for serialisation.  ...  As no locking or lease mechanism is used, an ongoing fast write sequence can be interrupted at any time by other proposers.  ... 
arXiv:2001.03362v1 fatcat:7wcbduubzbhzjp7vgafqgsb7t4

RMWPaxos: Fault-Tolerant In-Place Consensus Sequences

Jan Skrzypczak, Florian Schintke, Thorsten Schuett
2020 IEEE Transactions on Parallel and Distributed Systems  
The protocol builds a fault-tolerant atomic register that supports arbitrary read-modify-write operations.  ...  Reading a consolidated consensus is done without state modifications and is thereby free of concurrency control and demand for serialisation.  ...  As no locking or lease mechanism is used, an ongoing fast write sequence can be interrupted at any time by other proposers.  ... 
doi:10.1109/tpds.2020.2981891 fatcat:5rkwiyimd5ejbh6mm54gxnzoua

Kite

Vasilis Gavrielatos, Antonios Katsarakis, Vijay Nagarajan, Boris Grot, Arpit Joshi
2020 Proceedings of the 25th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming  
Kite enforces RC barriers through a novel fast/slow path mechanism that leverages the absence of failures in the typical case to maximize performance while relying on the slow path for progress.  ...  asynchronous environment, albeit sacrificing local reads.  ...  Therefore, to enforce RC, Kite must ensure that S1's write to X reaches S2 before the write (release) to f lag. Fast path: RC & ES without asynchrony.  ... 
doi:10.1145/3332466.3374516 dblp:conf/ppopp/GavrielatosKNGJ20 fatcat:micquvio4vhubj7f5myyljotgu

Harmonia: Near-Linear Scalability for Replicated Storage with In-Network Conflict Detection [article]

Hang Zhu, Zhihao Bai, Jialin Li, Ellis Michael, Dan Ports, Ion Stoica, Xin Jin
2019 arXiv   pre-print
To achieve this goal, Harmonia detects read-write conflicts in the network, which enables any replica to serve reads for objects with no pending writes.  ...  We present Harmonia, a replicated storage architecture that exploits the capability of new-generation programmable switches to obviate this tradeoff by providing near-linear scalability without sacrificing  ...  Harmonia fills in the missing piece for reads: it demonstrates how to make reads scalable without sacrificing either write performance or consistency.  ... 
arXiv:1904.08964v1 fatcat:nsek3dbpcfhong4oqgzx2rwz24

FLAIR: Accelerating Reads with Consistency-Aware Network Routing

Hatem Takruri, Ibrahim Kettaneh, Ahmed Alquraan, Samer Al-Kiswany
2020 Symposium on Networked Systems Design and Implementation  
serve the read without sacrificing linearizability.  ...  FLAIR leverages the capabilities of the new generation of programmable switches to serve reads from follower replicas without compromising consistency.  ...  These changes significantly improved Raft's performance.  Quorum-based reads (Fast Paxos). An alternative to the leader-based design is the quorum design [40, 41, 53] .  ... 
dblp:conf/nsdi/TakruriKAA20 fatcat:rlat7zqbcfghfmtga764i422tq

Megastore: Providing Scalable, Highly Available Storage for Interactive Services

Jason Baker, Chris Bond, James C. Corbett, J. J. Furman, Andrey Khorlin, James Larson, Jean-Michel Leon, Yawei Li, Alexander Lloyd, Vadim Yushprakh
2011 Conference on Innovative Data Systems Research  
This partitioning allows us to synchronously replicate each write across a wide area network with reasonable latency and support seamless failover between datacenters.  ...  Fast Reads We set an early requirement that current reads should usually execute on any replica without inter-replica RPCs.  ...  For any log positions without a known-committed value available, invoke Paxos to propose a no-op write.  ... 
dblp:conf/cidr/BakerBCFKLLLLY11 fatcat:q75xpp6cefc2tonkrovexhnroi

Conflict detection tradeoffs for replicated data

Michael J. Carey, Miron Livny
1991 ACM Transactions on Database Systems  
To achieve this goal, Harmonia detects read-write conflicts in the network, which enables any replica to serve reads for objects with no pending writes.  ...  We present Harmonia, a replicated storage architecture that exploits the capability of new-generation programmable switches to obviate this tradeoff by providing near-linear scalability without sacrificing  ...  Harmonia fills in the missing piece for reads: it demonstrates how to make reads scalable without sacrificing either write performance or consistency.  ... 
doi:10.1145/115302.115289 fatcat:rt4ewjk4bvcafbfi4ggwg7loai

Exploiting Commutativity For Practical Fast Replication [article]

Seo Jin Park, John Ousterhout
2017 arXiv   pre-print
In RAMCloud, CURP improved write latency by ~2x (13.8 us -> 7.3 us) and write throughput by 4x.  ...  CURP can be used with RAMCloud without sacrificing its fast crash recovery [15] ).  ...  To avoid replicating read requests, consensus protocols with strong leaders usually use a lease for the leader; a leader with a valid lease can execute read operations without replication.  ... 
arXiv:1710.09921v1 fatcat:ox5t6b2jmnfi3cy4mvczjwydt4

Making Reads in BFT State Machine Replication Fast, Linearizable, and Live [article]

Christian Berger, Hans P. Reiser, Alysson Bessani
2021 arXiv   pre-print
We show this weakness by presenting an attack in which a malicious leader blocks correct clients and present two solutions for patching the protocol, making read-only operations fast and correct.  ...  of reads from five to two communication steps.  ...  Choosing a Paxos quorum for a lease helps with managing leases and reducing overhead as lease revocation and Paxos messages can be combined.  ... 
arXiv:2107.11144v1 fatcat:bzqzh7usb5gfpgsrxddgpoqfge

PolarFS

Wei Cao, Zhenjun Liu, Peng Wang, Sen Chen, Caifeng Zhu, Song Zheng, Yuhui Wang, Guoqing Ma
2018 Proceedings of the VLDB Endowment  
In this way, the end-toend latency of PolarFS has been reduced drastically and our experiments show that the write latency of PolarFS is quite close to that of local file system on SSD.  ...  ParallelRaft relaxes Raft's strictly sequential order of write without sacrificing the consistency of storage semantics, which improves the performance of parallel writes of PolarFS.  ...  : insert = 2:1:1) and read/write-mixed (read : write = 7:2) OLTP workloads.  ... 
doi:10.14778/3229863.3229872 fatcat:gc4yrd4y5nebvh4rvf3ju4k44m

Minimizing Commit Latency of Transactions in Geo-Replicated Data Stores

Faisal Nawab, Vaibhav Arora, Divyakant Agrawal, Amr El Abbadi
2015 Proceedings of the 2015 ACM SIGMOD International Conference on Management of Data - SIGMOD '15  
MDCC [27] uses Fast Paxos [32] to enable committing transactions in one round of communication to a single, but larger than a majority, quorum.  ...  The version timestamp is the timestamp of the most recent write operation that wrote the data object. Blind writes are allowed, meaning that an object can exist in the write-set without being read.  ... 
doi:10.1145/2723372.2723729 dblp:conf/sigmod/NawabAAA15 fatcat:tn4m4ox2kvdeldlbztetudkjm4

Specification Of Security Enablers For Data Management

Marko Vukolic
2016 Zenodo  
This technique consists in forcing readers to write-back the data they read to ensure this data will be available in a quorum when the read completes [165, 227, 54] .  ...  We want to maximize parallelism between operation execution and partition transfer without sacrificing the consistency of the service (i.e., linearizability [175] ). 3.  ... 
doi:10.5281/zenodo.263831 fatcat:ksk4ze2tuncfbexjzsorjnoabe

The FuzzyLog: A Partially Ordered Shared Log

Joshua Lockerman, Jose M. Faleiro, Juno Kim, Soham Sankaran, Daniel J. Abadi, James Aspnes, Siddhartha Sen, Mahesh Balakrishnan
2018 USENIX Symposium on Operating Systems Design and Implementation  
By exposing a partial order, the FuzzyLog enables three key capabilities for applications: linear scaling for throughput and capacity (without sacrificing atomicity), weaker consistency guarantees, and  ...  On a 6-node Dapple deployment, our FuzzyLogbased ZooKeeper supports 3M/sec single-key writes, and 150K/sec atomic cross-shard renames.  ...  In TXMap (417 LOC), each server executes read-write transactions speculatively [8, 10] , tracking read-sets and buffering write-sets.  ... 
dblp:conf/osdi/LockermanFKSAA018 fatcat:iqovgum7ojhg3lnal253tc3t4i

Invalidation-Based Protocols for Replicated Datastores [article]

Antonios Katsarakis
2021 arXiv   pre-print
Paxos quorum leases: Fast reads without sacrificing writes. In Proceedings of the Symposium on Cloud Computing, SOCC ’14, pages 1–13, USA, 2014. ACM.  ...  ., execute reads and writes) as long as their lease has not expired.  ... 
arXiv:2112.02405v1 fatcat:toi3wflouvbzngpdfzypw3l4li
« Previous Showing results 1 — 15 out of 33 results