SQL Transaction Processing, Price-Performance Testingv2.0

Microsoft SQL Server Evaluation: Azure Virtual Machines vs. Amazon Web Services EC2

Table of Contents

  1. Summary
  2. Cloud IaaS SQL Server Offerings
  3. Field Test Setup
  4. Field Test Results
  5. Price Per Performance
  6. Conclusion
  7. Disclaimer
  8. About Microsoft
  9. About William McKnight
  10. About Jake Dolezal

1. Summary

The fundamental underpinnings of most organizations are their transactions. These must be done well, with integrity, and with good performance. Not only has transaction volume soared of late, but the level of granularity in the transaction details has also reached new heights. Fast transactions greatly improve the efficiency of a high-volume business. Performance is incredibly important.

Recent trends in information management indicate organizations are shifting their focus to cloud-based solutions. In the past, the only clear choice for most organizations was an on-premises database using on-premises hardware. However, costs of scale are chipping away the notion that this is still the best approach for all of a company’s transactional needs. The factors driving operational and analytical data projects to the cloud are many, and the advantages–like data protection, high availability, and scale–are realized with infrastructure as a service (IaaS) deployment. In many cases, a hybrid approach serves as an interim step for organizations migrating to a modern, capable cloud architecture.

This report outlines the results from a GigaOm Transactional Field Test, derived from the industry-standard TPC Benchmark™ E (TPC-E), to compare two IaaS cloud database offerings:

  1. Microsoft SQL Server on Amazon Web Services (AWS) Elastic Cloud Compute (EC2) instances
  2. Microsoft SQL Server on Microsoft Azure Virtual Machines (VM)

Both are installations of Microsoft SQL Server, and we tested using Red Hat Enterprise Linux (RHEL).

The results of the GigaOm Transactional Field Test are valuable to all operational functions of an organization, such as human resource management, production planning, material management, financial supply-chain management, sales and distribution, financial accounting and controlling, plant maintenance, and quality management. The data underlying many of these departments today is in SQL Server, which is also frequently the source for operational interactive business intelligence (BI).

With the Azure local cache, Microsoft SQL Server on Microsoft Azure Virtual Machines showed 2.9x better performance over AWS when tested on RedHat Enterprise Linux 8.2. Moreover, SQL Server on Microsoft Azure Virtual Machines had up to 64% better price-performance when comparing both on-demand and pay-as-you-go rates.

Testing hardware and software across cloud vendors is very challenging. Certain configurations favor one cloud vendor over another in feature availability, virtual machine processor generations, memory amounts, storage configurations for optimal input/output, network latencies, software and operating system versions, and the benchmarking workload itself. Our testing demonstrates a narrow slice of potential configurations and workloads.

As the sponsor of the report, Microsoft selected the particular Azure configuration it wanted to test. GigaOm chose the AWS instance configuration closest to it in terms of CPU, memory, and disk configuration. There were tradeoffs, which resulted in an input-output operations per second (IOPS) disadvantage for AWS (which we discuss in the report).

We leave the issue of fairness for the reader to determine. We strongly encourage you, the reader, to look past marketing messages and discern for yourself what is of value. We hope this report is informative and helpful in revealing some of the challenges and nuances of platform selection.

In the same spirit as the TPC, price-performance is intended to be a normalizer of performance results across different configurations. Of course, this has its shortcomings, but at least you can determine that “what you pay for and configure is what you get.”

The parameters to replicate this test are provided. We used the BenchCraft workload driver, which was audited by a TPC-approved auditor who reviewed all updates to BenchCraft. All the information required to reproduce the results are documented in the TPC-E specification. BenchCraft implements the requirements documented in Clauses 3, 4, 5, and 6 of the benchmark specification. There is nothing in BenchCraft that alters the performance of TPC-E or this TPC-E derived workload.

The scale factor in TPC-E is defined as the number of required customer rows per single transaction per second. We did change the number of initial trading days (ITD). The default value is 300, which is the number of 8-hour business days to populate the initial database. For these tests, we used an ITD of 30 days rather than 300. This reduces the size of the initial database population in the larger tables. The overall workload behaves identically with an ITD of 300 or 30 as far as the transaction profiles are concerned. Since the ITD was reduced to 30, any results obtained would not be compliant with the TPC-E specification and, therefore, not comparable to published results. This is the basis for the standard disclaimer that this is a workload derived from TPC-E.

However, BenchCraft is just one way to run TPC-E. All the information necessary to recreate the benchmark is available at TPC.org (this test used the latest version 1.14.0). Just change the ITD, as noted above.

We have provided enough information in the report for anyone to reproduce this test. You are encouraged to compile your own representative queries, data sets, and data sizes, and to test compatible configurations applicable to your requirements.

Full content available to GigaOm Subscribers.

Sign Up For Free