Multi-Factor Performance Comparison of Amazon Web Services Elastic Compute Cluster and Google Cloud Platform Compute Engine

Multi-Factor Performance Comparison of Amazon Web Services Elastic Compute Cluster and Google Cloud Platform Compute Engine

Sanjay P. Ahuja, Emily Czarnecki, Sean Willison
Copyright: © 2020 |Pages: 16
DOI: 10.4018/IJCAC.2020070101
OnDemand:
(Individual Articles)
Available
$37.50
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

Cloud computing has rapidly become a viable competitor to on-premise infrastructure from both management and cost perspectives. This research provides insight into cluster computing performance and variability in cloud-provisioned infrastructure from two popular public cloud providers. A comparative examination of the two cloud platforms using synthetic benchmarks is provided. In this article, we compared the performance of Amazon Web Services Elastic Compute Cluster (EC2) to the Google Cloud Platform (GCP) Compute Engine using three benchmarks: STREAM, IOR, and NPB-EP. Experiments were conducted on clusters with increasing nodes from one to eight. We also performed experiments over the course of two weeks where benchmarks were run at similar times. The benchmarks provided performance metrics for bandwidth (STREAM), read and write performance (IOR), and operations per second (NPB-EP). We found that EC2 outperformed GCP for bandwidth. Both provided good scalability and reliability for bandwidth with GCP showing a slight deviation during the two-week trial. GCP outperformed EC2 in both the read and write tests (IOR) as well as the operations per second test. However, GCP was extremely variable during the read and write tests over the two-week trial. Overall, each platform excelled in different benchmarks and we found EC2 to be more reliable in general.
Article Preview
Top

2. Cloud Computing Platform

Interestingly, there is a significant similarity between many cloud providers (Sultan, 2010). The standard offerings like VM instances, elasticity, monitoring, logging, networking, etc. are just that, standard. Where the various cloud providers start to diverge is in the specialty offerings, like Google’s Vision service, and AWS RedShift. On those sorts of items, there is real value in choosing one provider over the other. For the commodity offerings, there are several distinctions on the tooling and SDK front, but not enough of an advantage is noticed between them to make any real difference (Amazon, n.d.; Google, n.d.).

2.1 Instance Types

Here we have VMs, Container Service and Lambda, for all three providers, and more or less the same OS, Compute, Memory, GPU and Disk customization options. Compute is a true commodity with a few exceptions. GPU/APU can vary insofar as chaining options and types (Sandbu, 2017). Further, the mechanisms for managing compute instances and compute clusters, and the means for networking the same can vary. But at the end of the day, there is no meaningful difference in cost or performance. There is statistically meaningful difference showing that AWS is faster insofar as latency and compute dedication, but not a practical operational meaningfulness (Patrizio, 2019).

2.2 Development Environment (SDKs)

Many developers prefer to do the majority of their work from the CLI. We are simply more productive when we don’t have to open a GUI for every task (Computer Hope, 2018). The majority of common management tasks, including provisioning, deployment, scaling, etc. can be achieved via the CLI tooling offered by all major providers. To this end, we can see having a complete CLI/SDK is, in fact, a commodity. AWS is one of the most complete CLIs (Amazon, n.d.). AWS more or less offers CLI management and interaction for all services and projects. The AWS CLI is complete, simple and highly intuitive making it an excellent resource for DevOps engineers. The gcloud CLI is also very complete, but the documentation is a little harder to follow (Google, n.d.-b). The gcloud CLI does have an interesting adaptation in that it is pre-loaded into the web-based command line tooling launchable from the gcloud web interface. We found the web CLI to be a very useful feature and it saves the user from having to be on a machine with the CLI installed to work on their IaaS hosted by GCP.

Complete Article List

Search this Journal:
Reset
Volume 14: 1 Issue (2024)
Volume 13: 1 Issue (2023)
Volume 12: 4 Issues (2022): 2 Released, 2 Forthcoming
Volume 11: 4 Issues (2021)
Volume 10: 4 Issues (2020)
Volume 9: 4 Issues (2019)
Volume 8: 4 Issues (2018)
Volume 7: 4 Issues (2017)
Volume 6: 4 Issues (2016)
Volume 5: 4 Issues (2015)
Volume 4: 4 Issues (2014)
Volume 3: 4 Issues (2013)
Volume 2: 4 Issues (2012)
Volume 1: 4 Issues (2011)
View Complete Journal Contents Listing