[Paper Review] The Impact of Distance on Performance and Scalability of Distributed Database Systems in Hybrid Clouds

Yaser Mansouri, M. Ali Babar|arXiv (Cornell University)|Jul 30, 2020

Cloud Computing and Resource Management4 citations

TL;DR

This paper evaluates the impact of inter-cloud distance on throughput and scalability of six modern distributed databases (MongoDB, Cassandra, Riak, CouchDB, Redis, MySQL Cluster) in hybrid cloud environments spanning on-premises and public cloud regions (Sydney, Mumbai, Virginia). It finds that increasing distance degrades performance for most databases, with MongoDB and MySQL Cluster achieving the highest throughput, while vertical scaling (larger VMs) significantly outperforms horizontal scaling (more VMs), especially for I/O-intensive and communication-heavy databases like Cassandra and Redis.

ABSTRACT

The increasing need for managing big data has led the emergence of advanced database management systems. There has been increased efforts aimed at evaluating the performance and scalability of NoSQL and Relational databases hosted by either private or public cloud datacenters. However, there has been little work on evaluating the performance and scalability of these databases in hybrid clouds, where the distance between private and public cloud datacenters can be one of the key factors that can affect their performance. Hence, in this paper, we present a detailed evaluation of throughput, scalability, and VMs size vs. VMs number for six modern databases in a hybrid cloud, consisting of a private cloud in Adelaide and Azure based datacenter in Sydney, Mumbai, and Virginia regions. Based on results, as the distance between private and public clouds increases, the throughput performance of most databases reduces. Second, MongoDB obtains the best throughput performance, followed by MySQL C luster, whilst Cassandra exposes the most fluctuation in through performance. Third, vertical scalability improves the throughput of databases more than the horizontal scalability. Forth, exploiting bigger VMs rather than more VMs with less cores can increase throughput performance for Cassandra, Riak, and Redis.

Motivation & Objective

• To investigate how inter-cloud distance affects throughput and scalability of distributed databases in hybrid cloud architectures.
• To compare the performance of six modern databases (MongoDB, Cassandra, Riak, CouchDB, Redis, MySQL Cluster) across varying distances in hybrid cloud deployments.
• To evaluate the trade-offs between vertical and horizontal scaling in hybrid cloud environments with geographically separated private and public cloud datacenters.
• To analyze the impact of VM size (cores) versus VM count on database throughput under different network latencies.

Proposed method

• Deployed six distributed databases across a hybrid cloud setup with a private cloud in Adelaide and public cloud datacenters in Sydney, Mumbai, and Virginia.
• Measured throughput performance under controlled workloads while varying the number of VMs burst into the public cloud and the size (core count) of VMs.
• Conducted systematic experiments across three geographic distances (Sydney: short, Mumbai/Virginia: long) to isolate the effect of latency due to distance.
• Evaluated both horizontal scaling (increasing VM count) and vertical scaling (increasing VM core count) for each database under identical workload conditions.
• Used VM packing strategies to compare performance when using fewer, larger VMs versus more, smaller VMs, focusing on inter-VM communication overhead.
• Analyzed performance trends across consistency mechanisms, replication strategies, and data models to identify patterns in scalability behavior.

Experimental results

Research questions

• RQ1 How does increasing geographic distance between private and public cloud datacenters affect the throughput performance of distributed databases in hybrid clouds?
• RQ2 Which database system achieves the highest throughput in hybrid cloud deployments, and how does this vary with distance and scaling strategy?
• RQ3 To what extent does vertical scaling (larger VMs) outperform horizontal scaling (more VMs) in terms of throughput for different distributed databases?
• RQ4 How does VM packing (fewer, larger VMs vs. more, smaller VMs) influence performance, particularly for communication-intensive or in-memory databases?
• RQ5 How do consistency mechanisms and replication strategies interact with network latency to affect database scalability in hybrid cloud environments?

Key findings

• As distance between private and public cloud datacenters increases, the throughput performance of most databases decreases, with MongoDB and MySQL Cluster showing the most stable performance.
• MongoDB achieved the highest throughput among all databases due to its full replication and eventual consistency model, outperforming others even at long distances.
• Cassandra exhibited the highest performance fluctuation, with throughput initially decreasing and then increasing as more VMs were added to the public cloud, particularly in the Sydney region.
• Vertical scaling (increasing VM core count) improved throughput more than horizontal scaling: for Cassandra, throughput increased by up to 800% when scaling from 2 to 8 cores.
• For Riak, throughput improved by 30% and for Redis by 10% when using larger VMs instead of more smaller VMs, indicating strong benefits of vertical scaling for memory-intensive databases.
• VM packing (fewer, larger VMs) significantly improved performance for Cassandra and Riak in the Sydney region, suggesting that inter-VM communication overhead is a key factor in performance degradation.