Adding Up the Price/Performance Advantages Of Bare-Metal vs. Virtualized Servers in the Cloud

Close

 

 

Research shows that bare-metal servers can deliver significant price/performance benefits versus virtualized servers in cloud-based Infrastructure-as-a-Service (IaaS) environments. The advantages of bare-metal servers are attained through reduced software licensing fees, lower data transfer costs and performance improvements enabled by the elimination of overhead penalties associated with virtualization.

For IT professionals evaluating bare-metal vs. virtualized servers for their cloud environments it is important to understand where those price/performance advantages are and how significant they can be for specific workloads and use cases.

Toward that end, the IBM Competitive Project Office (CPO)1 conducted independent lab-based technical analysis of solutions from SoftLayer, an IBM company, and Amazon Web Services (AWS). It is worth noting that IBM SoftLayer offers both bare-metal and virtualized approaches to dedicated servers.

The IBM CPO research examined four workloads and configurations: A Web application; analytics; message passing, and Web application consolidation. In every case, SoftLayer was able to outperform AWS on both performance and price/performance. In this article we look at the results of two Web application scenarios.

Comparing SoftLayer vs. AWS For Web Applications
IBM CPO tested a Web application in a SoftLayer bare-metal environment versus an AWS virtualized environment, and also tested a SoftLayer virtualized environment versus an AWS virtualized environment.2

In the bare-metal-vs-virtualization comparison, the IBM solution consisted of two dedicated servers, comprised as follows:

App Server

  • 4x3.4GHZ Xeon 1270 cores, 8 Hyperthreads
  • 12 GB RAM
  • Four 32GB SSDs (RAID0)
  • Redhat Linux 6, WebSphere Application Server

Database Server

  • 4x3.4GHZ Xeon 1270 cores, 8 Hyperthreads
  • 12 GB RAM
  • Four 32GB SSDs (RAID0)
  • Redhat Linux 6, DB2

AWS doesn’t offer a bare-metal alternative, so the comparable solution was based on two 8 vCPU virtual machines, housed in a U.S. East Coast data center, with 3-year commit instances and heavy usage. The configuration consisted of:

App Server

  • 8 vCPUs
  • 15 GB RAM
  • 2x 80GB SSDs
  • Redhat Linux 6, WebSphere Application Server

Database Server

  • 8 vCPUs
  • 15 GB RAM
  • 2x 80GB SSDs
  • Redhat Linux 6, DB2

One of the reasons the Web application was chosen is because it is a common business scenario for cloud environments. It also requires high-performance workloads and database back-end systems. The specific test conducted by CPO was modeled after a typical banking scenario whereby a user would follow a number of steps during a typical transaction, from logon to sign out.

SoftLayer: An Overview

Get to know SoftLayer, and learn how its unified platform can offer your team cloud infrastructure at Internet scale.

The performance differentials between the SoftLayer bare-metal configuration and the AWS virtualized configuration were as follows:

  • The SoftLayer solution was able to process a maximum of 19,453 requests per second, versus a maximum of 7,329 requests per second for the AWS solution.
  • The SoftLayer solution was 45% lower in price than the AWS solution, based on three-year total cost of acquisition.
  • The SoftLayer solution was 2.6 times faster than the AWS solution for compared systems.
  • The SoftLayer solution delivered 4.8x better price performance than the AWS solution.3

IBM CPO conducted similar tests on a comparably configured SoftLayer virtualized solution versus an AWS virtualized solution. In that comparison, the AWS solution was the same as the one described earlier, while the SoftLayer solution consisted of:

App Server

  • 8 vCPUs at 2.0GHZ each
  • 8 GB RAM
  • 25GB SAN storage
  • Redhat Linux 6, WebSphere Application Server

DB Server

  • 8 vCPUs at 2.0GHZ each
  • 8 GB RAM
  • 25GB SAN storage
  • Redhat Linux 6, DB2

IBM CPO deployed the same banking scenario on the systems and then tuned each environment to provide the fastest throughput (best performance), measured by requests per second. In this case, the SoftLayer solution also surpassed the AWS solution in price, performance and price/performance:

  • The SoftLayer solution performed a maximum of 8,169 request per second, versus 7,329 for the AWS solution.
  • The SoftLayer solution was 26% lower in price than the AWS solution, based on a three-year total cost of acquisition.
  • The SoftLayer solution was 1.1x faster than the AWS solution for compared systems.
  • The SoftLayer solution delivered 1.5x better price performance than the AWS solution.4

As more IT professionals increasingly explore IaaS and dedicated servers through the cloud, it is important to understand all of your options, and to know how your choices can impact the overall costs of your solution as well as the performance you deliver to your organization. The evidence shows potential for saving money and driving improved performance by working with providers that deliver bare-metal solutions in the cloud.5

1Based on an internal IBM study of a basic Web application workload. Customer applications, differences in the stack deployed and other systems variations may produce different results. Pricing based on published U.S. list prices for a U.S. East Coast data center as May 8, 2014 and is calculated by projecting costs measured over a fixed testing period to a three-year total cost of acquisition including hardware, software and data transfer monthly costs.
2Ibid, footnote #1
3This is an IBM internal study of a basic web application workload. Results obtained under controlled conditions testing resources of a publicly available cloud infrastructure. Customer applications, differences in the stack deployed, and other systems variations or testing conditions may produce different results. Pricing based on published US list prices available for data center location specified as of 5/8/2014 and calculated by projecting costs measured over a fixed testing period to a 3 year TCA, including all hardware, software and data transfer monthly costs. Option for dedicated servers not available in Competitor Cloud environment.
4Ibid, footnote #3
5The examples shown illustrate potential results and savings. They are based on specific workloads and specific cloud environments optimized for maximum results. Results and pricing will vary depending on cloud environment, location and other factors. Generally expected results cannot be provided as each client’s results will vary depending on the systems and services ordered, workload, location and other factors. Contact IBM to see what SoftLayer can do for you.

Advertisement