Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TierBase: A Workload-Driven Cost-Optimized Key-Value Store

Zhitao Shen, Shiyu Yang, Weibo Chen, Kunming Wang, Yue Li, Jiabao Jin, Wei Jia, Junwei Chen, Yuan Su, Xiaoxia Duan, Wei Chen, Lei Wang, Jie Song, Ruoyi Ruan, Xuemin Lin

TL;DR

This work tackles the challenge of cost-efficient, low-latency online data serving by introducing a Space-Performance Cost Model and TierBase, a disaggregated, tiered storage key-value store. The approach balances performance and storage costs through cache/storage disaggregation, pre-trained compression, persistent memory, and elastic threading, guided by a workload-aware optimization framework. Through synthetic benchmarks and real Ant Group case studies, TierBase demonstrates substantial cost savings (up to 62% in primary scenarios) and favorable space–performance trade-offs compared with Redis, Cassandra, and HBase. The paper provides a practical framework for adapting storage configurations to diverse workloads in production environments, with broad implications for large-scale online data serving systems.

Abstract

In the current era of data-intensive applications, the demand for high-performance, cost-effective storage solutions is paramount. This paper introduces a Space-Performance Cost Model for key-value store, designed to guide cost-effective storage configuration decisions. The model quantifies the trade-offs between performance and storage costs, providing a framework for optimizing resource allocation in large-scale data serving environments. Guided by this cost model, we present TierBase, a distributed key-value store developed by Ant Group that optimizes total cost by strategically synchronizing data between cache and storage tiers, maximizing resource utilization and effectively handling skewed workloads. To enhance cost-efficiency, TierBase incorporates several optimization techniques, including pre-trained data compression, elastic threading mechanisms, and the utilization of persistent memory. We detail TierBase's architecture, key components, and the implementation of cost optimization strategies. Extensive evaluations using both synthetic benchmarks and real-world workloads demonstrate TierBase's superior cost-effectiveness compared to existing solutions. Furthermore, case studies from Ant Group's production environments showcase TierBase's ability to achieve up to 62% cost reduction in primary scenarios, highlighting its practical impact in large-scale online data serving.

TierBase: A Workload-Driven Cost-Optimized Key-Value Store

TL;DR

This work tackles the challenge of cost-efficient, low-latency online data serving by introducing a Space-Performance Cost Model and TierBase, a disaggregated, tiered storage key-value store. The approach balances performance and storage costs through cache/storage disaggregation, pre-trained compression, persistent memory, and elastic threading, guided by a workload-aware optimization framework. Through synthetic benchmarks and real Ant Group case studies, TierBase demonstrates substantial cost savings (up to 62% in primary scenarios) and favorable space–performance trade-offs compared with Redis, Cassandra, and HBase. The paper provides a practical framework for adapting storage configurations to diverse workloads in production environments, with broad implications for large-scale online data serving systems.

Abstract

In the current era of data-intensive applications, the demand for high-performance, cost-effective storage solutions is paramount. This paper introduces a Space-Performance Cost Model for key-value store, designed to guide cost-effective storage configuration decisions. The model quantifies the trade-offs between performance and storage costs, providing a framework for optimizing resource allocation in large-scale data serving environments. Guided by this cost model, we present TierBase, a distributed key-value store developed by Ant Group that optimizes total cost by strategically synchronizing data between cache and storage tiers, maximizing resource utilization and effectively handling skewed workloads. To enhance cost-efficiency, TierBase incorporates several optimization techniques, including pre-trained data compression, elastic threading mechanisms, and the utilization of persistent memory. We detail TierBase's architecture, key components, and the implementation of cost optimization strategies. Extensive evaluations using both synthetic benchmarks and real-world workloads demonstrate TierBase's superior cost-effectiveness compared to existing solutions. Furthermore, case studies from Ant Group's production environments showcase TierBase's ability to achieve up to 62% cost reduction in primary scenarios, highlighting its practical impact in large-scale online data serving.
Paper Structure (48 sections, 2 theorems, 10 equations, 13 figures, 3 tables)

This paper contains 48 sections, 2 theorems, 10 equations, 13 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Space-performance Cost Modeling
  3. Cost Analysis Framework
  4. Cost Efficiency Metrics
  5. Space-Performance Trade-off and Optimal Cost Theorem
  6. Cost Model with Tiered Storage
  7. Cost Optimization Strategies
  8. Single-Tier Storage Optimization
  9. Tiered Storage Optimization
  10. Workload-Driven Optimization Approaches
  11. TierBase System Design
  12. Cost Optimization Strategies
  13. Tiered Storage
  14. Write-through Caching
  15. Write-back Caching
  16. ...and 33 more sections

Key Result

theorem 1

For a given workload $w$ with requirements $QPS$ and $DataSize$, and a set of storage configurations $S$, the optimal cost $C^*$ is achieved by selecting the configuration $s^* \in S$ that minimizes the overall cost while balancing performance and space costs: $C^* = \min_{s \in S} \max(PC_{s}, SC_{

Figures (13)

  • Figure 1: Cost comparison in TierBase
  • Figure 2: Space-Performance Cost Model
  • Figure 3: TierBase architecture overview
  • Figure 4: Caching policy for write operation
  • Figure 5: The framework of pre-trained based compression
  • ...and 8 more figures

Theorems & Definitions (5)

  • Definition 1: Cost of workload $w$
  • Definition 2: Cost Metrics
  • Definition 3: Space-Performance Trade-off of Storage Configurations
  • theorem 1: Optimal Cost $C^*$
  • theorem 2: Optimal Cache Tier Cost