TripS: Automated Multi-tiered Data Placement in a Geo-distributed Cloud Environment

Motivation

Many Internet service providers such as Netflix, Uber, and AirBnB use cloud providers' multiple data centers (DCs) to provide reduced user-perceived latency and better data availability without having their own DCs on multiple locations. Typically, those Internet applications want to get composite benefits from multiple storage services in each DC e.g., hot data in memory and cold data in object storage. Thus, applications developers need to find optimized DC locations and (combined) storage services to realize a tradeoff suitable for their requirements. However, this task of choosing both DC locations and combining different storage tiers to realize composite benefits are tedious and not an easy task due to

• Too many DCs on multiple locations and many storage options from each DC (the number of them only increases)
• Dynamics from cloud environment and applications

TripS Architecture

TripS (Switch Storage System) is a light-weight data palcement decision system considering both DC locations and storage tiers. TripS can run on top of any system (or application) that needs to find the optimized data placement in a multi-cloud environment. TripS also can help applications to handle dynamics from cloud infrastructure and applications (especially, short-term dynamics e.g., burst access, transient DC (network) failure, or overloaded node).

To evaluate the optimized data plamcement in a multi-cloud environment, TripS requires inputs from an underlying system.

• Application desired goals e.g., SLA, consistency model, and degree of fault tolerance
• Cost Information e.g., Storage and network cost.
• Latency Information e.g., Storage and network (between DCs) latency
• Workload Information e.g., number of requests (get and put) and average data size

People

• Principal Investigators
• Jon Weissman
• Abhishek Chandra

• Students
• Kwangsung Oh

Publications

• TripS: Automated Multi-tiered Data Placement in a Geo-distributed Cloud Environment [pdf] [slides]
  Kwangsung Oh, Abhishek Chandra and Jon Weissman.
  In Proceedings of the 10th ACM International Systems and Storage Conference (SYSTOR 2017), Haifa, Israel, May 2017.
  

Sponsor