Random Notes
  • Introduction
  • Reading list
  • Theory
    • Index
      • Impossibility of Distributed Consensus with One Faulty Process
      • Time, Clocks, and the Ordering of Events in a Distributed System
      • Using Reasoning About Knowledge to analyze Distributed Systems
      • CAP Twelve Years Later: How the “Rules” Have Changed
      • A Note on Distributed Computing
  • Operating System
    • Index
  • Storage
    • Index
      • Tachyon: Reliable, Memory Speed Storage for Cluster Computing Frameworks
      • Exploiting Commutativity For Practical Fast Replication
      • Don’t Settle for Eventual: Scalable Causal Consistency for Wide-Area Storage with COPS
      • Building Consistent Transactions with Inconsistent Replication
      • Managing Update Conflicts in Bayou, a Weakly Connected Replicated Storage System
      • Spanner: Google's Globally-Distributed Database
      • Bigtable: A Distributed Storage System for Structured Data
      • The Google File System
      • Dynamo: Amazon’s Highly Available Key-value Store
      • Chord: A Scalable Peer-to-peer Lookup Service for Internet Applications
      • Replicated Data Consistency Explained Through Baseball
      • Session Guarantees for Weakly Consistent Replicated Data
      • Flat Datacenter Storage
      • Small Cache, Big Effect: Provable Load Balancing forRandomly Partitioned Cluster Services
      • DistCache: provable load balancing for large-scale storage systems with distributed caching
      • Short Summaries
  • Coordination
    • Index
      • Logical Physical Clocks and Consistent Snapshots in Globally Distributed Databases
      • Paxos made simple
      • ZooKeeper: Wait-free coordination for Internet-scale systems
      • Just Say NO to Paxos Overhead: Replacing Consensus with Network Ordering
      • Keeping CALM: When Distributed Consistency is Easy
      • In Search of an Understandable Consensus Algorithm
      • A comprehensive study of Convergent and Commutative Replicated Data Types
  • Fault Tolerance
    • Index
      • The Mystery Machine: End-to-end Performance Analysis of Large-scale Internet Services
      • Gray Failure: The Achilles’ Heel of Cloud-Scale Systems
      • Capturing and Enhancing In Situ System Observability for Failure Detection
      • Check before You Change: Preventing Correlated Failures in Service Updates
      • Efficient Scalable Thread-Safety-Violation Detection
      • REPT: Reverse Debugging of Failures in Deployed Software
      • Redundancy Does Not Imply Fault Tolerance
      • Fixed It For You:Protocol Repair Using Lineage Graphs
      • The Good, the Bad, and the Differences: Better Network Diagnostics with Differential Provenance
      • Lineage-driven Fault Injection
      • Short Summaries
  • Cloud Computing
    • Index
      • Improving MapReduce Performance in Heterogeneous Environments
      • CLARINET: WAN-Aware Optimization for Analytics Queries
      • MapReduce: Simplified Data Processing on Large Clusters
      • Dryad: Distributed Data-Parallel Programs from Sequential Building Blocks
      • Resource Management
      • Apache Hadoop YARN: Yet Another Resource Negotiator
      • Mesos: A Platform for Fine-Grained Resource Sharing in the Data Center
      • Dominant Resource Fairness: Fair Allocation of Multiple Resource Types
      • Large-scale cluster management at Google with Borg
      • MapReduce Online
      • Delay Scheduling: A Simple Technique for Achieving Locality and Fairness in Cluster Scheduling
      • Reining in the Outliers in Map-Reduce Clusters using Mantri
      • Effective Straggler Mitigation: Attack of the Clones
      • Resilient Distributed Datasets: A Fault-Tolerant Abstraction for In-Memory Cluster Computing
      • Discretized Streams: Fault-Tolerant Streaming Computation at Scale
      • Sparrow: Distributed, Low Latency Scheduling
      • Making Sense of Performance in Data Analytics Framework
      • Monotasks: Architecting for Performance Clarity in Data Analytics Frameworks
      • Drizzle: Fast and Adaptable Stream Processing at Scale
      • Naiad: A Timely Dataflow System
      • The Dataflow Model:A Practical Approach to Balancing Correctness, Latency, and Cost in Massive-Scale
      • Interruptible Tasks:Treating Memory Pressure AsInterrupts for Highly Scalable Data-Parallel Program
      • PACMan: Coordinated Memory Caching for Parallel Jobs
      • Multi-Resource Packing for Cluster Schedulers
      • Other interesting papers
  • Systems for ML
    • Index
      • A Berkeley View of Systems Challenges for AI
      • Tiresias: A GPU Cluster Managerfor Distributed Deep Learning
      • Gandiva: Introspective Cluster Scheduling for Deep Learning
      • Workshop papers
      • Hidden Technical Debt in Machine Learning Systems
      • Inference Systems
      • Parameter Servers and AllReduce
      • Federated Learning at Scale - Part I
      • Federated Learning at Scale - Part II
      • Learning From Non-IID data
      • Ray: A Distributed Framework for Emerging AI Applications
      • PipeDream: Generalized Pipeline Parallelism for DNN Training
      • DeepXplore: Automated Whitebox Testingof Deep Learning Systems
      • Distributed Machine Learning Misc.
  • ML for Systems
    • Index
      • Short Summaries
  • Machine Learning
    • Index
      • Deep Learning with Differential Privacy
      • Accelerating Deep Learning via Importance Sampling
      • A Few Useful Things to Know About Machine Learning
  • Video Analytics
    • Index
      • Scaling Video Analytics on Constrained Edge Nodes
      • Focus: Querying Large Video Datasets with Low Latency and Low Cost
      • NoScope: Optimizing Neural Network Queriesover Video at Scale
      • Live Video Analytics at Scale with Approximation and Delay-Tolerance
      • Chameleon: Scalable Adaptation of Video Analytics
      • End-to-end Learning of Action Detection from Frame Glimpses in Videos
      • Short Summaries
  • Networking
    • Index
      • Salsify: Low-Latency Network Video through Tighter Integration between a Video Codec and a Transport
      • Learning in situ: a randomized experiment in video streaming
      • Short Summaries
  • Serverless
    • Index
      • Serverless Computing: One Step Forward, Two Steps Back
      • Encoding, Fast and Slow: Low-Latency Video Processing Using Thousands of Tiny Threads
      • SAND: Towards High-Performance Serverless Computing
      • Pocket: Elastic Ephemeral Storage for Serverless Analytics
      • Fault-tolerant and Transactional Stateful Serverless Workflows
  • Resource Disaggregation
    • Index
  • Edge Computing
    • Index
  • Security/Privacy
    • Index
      • Differential Privacy
      • Honeycrisp: Large-Scale Differentially Private Aggregation Without a Trusted Core
      • Short Summaries
  • Misc.
    • Index
      • Rate Limiting
      • Load Balancing
      • Consistency Models in Distributed System
      • Managing Complexity
      • System Design
      • Deep Dive into the Spark Scheduler
      • The Actor Model
      • Python Global Interpreter Lock
      • About Research and PhD
Powered by GitBook
On this page
  • TL;DR:
  • Summary:
  • Comments:
  • Related Links:

Was this helpful?

  1. Storage
  2. Index

Building Consistent Transactions with Inconsistent Replication

https://syslab.cs.washington.edu/papers/tapir-tr14.pdf

PreviousDon’t Settle for Eventual: Scalable Causal Consistency for Wide-Area Storage with COPSNextManaging Update Conflicts in Bayou, a Weakly Connected Replicated Storage System

Last updated 5 years ago

Was this helpful?

TL;DR:

TAPIR is layered approach to supporting distribution transactions, showing that a Transactional Application Protocol can be built on top of an Inconsistent Replication protocol

Summary:

This paper describes a layered approach for distributed transactions. It introduces TAPIR(“the Transaction Application Protocol for Inconsistent Replication”), which is built on top of an inconsistent replication that provides no consistency.

Distributed transactions with strong consistency are very useful when we are building a key-value store. However, in the paper, the authors observe that combining strong consistent replication protocol(e.g., Paxos) and distributed transaction protocol(e.g., 2-Phase commit) wastes work. They both provide linearizable ordering, which leads to high latency and low throughput. Specifically, if a system that implements 2-Phase commit and Paxos(I think Spanner is an example of such system?), the transactions will first send to the leader in each partition and the leaders are responsible for ordering the transactions. Such design requires at least 2 RTTs to commit a transaction, and the leader will cause leader bottleneck. Thus, the goal of TAPIR is to eliminate the linearizable ordering from replication layer.

The paper co-designed two protocols. The first one(IR) works in replication layer. IR provides fault-tolerance and agreement(a majority of replicas agrees on a result) without operation ordering. In other words, agreement means the replicas will return the same result to distributed transaction protocol. IR uses Quorum-like technique to detect conflicts. IR is efficient because of no leader and coordination. Then, the paper introduces TAPIR, which is designed explicitly to work with IR's unordered operations. TAPIR uses OCC to detect conflicts and loosely synchronized clocks to order transactions. Interestingly, TAPIR uses the client as the coordinator in 2PL. Multi-versioning is implemented in TAPIR to deal with inconsistent replicas.

Comments:

I think the overall idea is great. TAPIR provides cheaper transaction with same guarantees. The experimental result shows that TAPIR_KB has lower latency and better throughput than conventional systems(e.g., Spanner, MongoDB). However, what is not good is the complexity of the protocols. IR and TAPIR both involve sophisticated ideas and techniques. Also, if we use Paxos as replication protocol, any transaction protocol works. But, if we use IR, we need a particular transaction protocol(like TAPIR). Any conventional OCC or 2PL won't work.

Another interesting thing I found in this paper is that it exposes a tradeoff space between performance and programmability. the top-level protocol is very-hard-to-understand. but the savings are substantial. how many times does the top-level need to be re-implemented? this hearkens back to our conversation about Raft.

Related Links:

Irene Zhang blog post:

Irene Zhang's talk in SOSP' 15:

Review by the Morning Paper:

Irene Y. Zhang: Lessons learned from TAPIR(s)
LogoBuilding Consistent Transactions with Inconsistent Replication | the morning paper