Servers & Systems: The Right Compute

How HPE on-premises infrastructure compares to Amazon Web Services (AWS)

Part 1: How does on-premises infrastructure compare to public cloud models like AWS? To start our new blog series, we discuss the benefits of on-premises infrastructure, including TCO, workload throughput and price-performance ratio.

Blog_on-prem-vs-aws.jpgIn a recent study, Gartner estimates that public cloud IaaS will grow at an average of 30% CAGR* through 2020. Which raises the question: Why are some enterprises moving a portion of their infrastructure to the public cloud?

A quick search on the web can provide many of the common answers: Lower costs. Fast deployment times. Pay-per-use with no upfront payments. Ability to shift capital expenses to operational expenses. Greater elasticity. And better utilization rates.

These benefits have driven significant growth in public cloud adoption, and we sought to answer a few more questions:

  • Are all these benefits actual and only available through a public cloud model?
  • Does moving to the public cloud provide only upside benefits with no downsides?

We wanted to find out for ourselves, so we commissioned a detailed, objective analysis that compares HPE on-premises solutions to the Amazon Web Services (AWS) public cloud. We sized and purchased both the on-prem and public cloud infrastructures and optimized them for a cloud-scale advanced analytics workload. We wanted to compare the total monthly costs, workload performance, price-performance ratio and the elastic IT features that produce a flexible, pay-per-use experience. You can download the complete technical paper here: HPE On-Prem vs. AWS. (Registration is required.)

Summary of findings in the HPE On-Prem vs. AWS study

  • TCO: AWS IaaS TCO is 1.7 to 3.4 times higher than comparable HPE on-premises deployments. (Lower is better.)
  • Throughput: HPE on-premises deployments workload throughput is 45% higher than comparable AWS deployments. (Higher is better.)
  • Price-Performance: AWS price-performance ratio is 2.5 to 5.0 times higher than comparable deployments from HPE. (Lower is better.)
  • Elastic experience: HPE GreenLake Flex Capacity delivers pay-per-use consumption that is on-premises, which provides better control over latency, privacy, compliance and security, and it actually costs less per month than AWS solutions for varying workload throughput demands.

AWS Primer

Ordering infrastructure online through services like AWS EC2 is actually fairly simple and fast. We now present a brief primer of EC2 IaaS options available through AWS. The AWS customer’s first step is to determine the appropriate compute infrastructure for the workload.


AWS provides several choices of compute “families” including: General-Compute, Compute-Optimized, Memory-Optimized, Storage-Optimized and Accelerated-Computing. Each of these families is designed to provide optimization for workload sensitivities based on their name. For instance, the Memory-Optimized family provides a higher ratio of memory per vCPUs than other family options and the Storage-Optimized family provides local hard-drives or local SSDs rather than only network-attached storage.  Within each family, several types are typically available to enable the customer to tailor the infrastructure based on CPU generation, frequency, cost, etc.

Instances and dedicated Hosts

The customer next decides whether the compute can be purchased as a set of Dedicated Hosts, Dedicated Instances or Shared Instances. A Dedicated Host is a complete server node and is usually selected this way to minimize cost of existing server-bound software licenses. An instance is a slice of a node ranging from a few vCPUs (threads) per instance up to the maximum number of vCPUs available in a complete server node for that particular type. The other sub-systems, such as memory, local storage and networking bandwidth, also tend to scale proportionately with the number of vCPUs in the instance type at a ratio consistent within and across the family.

Dedicated and shared Instances

A customer may choose to save a little cost by selecting a shared instance, which may reside on a host that contains other customer’s instances. Alternatively, the customer, who is concerned about “noisy neighbor” effects and wants more control over throughput predictability, may select Dedicated Instances where there are no other AWS customers sharing the host hardware.

On-demand and reserved Instances

An attractive feature of public cloud suppliers, like AWS, is the on-demand offering, which allows customers to quickly scale their infrastructure up (or down) based on changing business requirements while only paying for what they use. These changing business requirements may be in the form of increased throughput demand due to seasonality or the need to execute short-term projects without the long-term infrastructure commitment. The on-demand purchase option provides the most flexibility, including the ability to cancel the infrastructure at any time with no further use charges. However, on-demand is the most expensive purchase option per hour used. Alternatively, AWS EC2 provides options of 3-year, paid all upfront, Reserved Instances and 1-year, paid all upfront, Reserved Instances, providing significant discounts from the on-demand pricing. From the AWS website, the 1-year commitment provides an average discount of 40% and the 3-year commitment provides an average discount of 60% and up to 75% for some instance types.

The breakeven point of on-demand vs. 3-year paid all upfront reserved instances

AWS instances.pngIf an AWS customer knows how long a portion of infrastructure will be used, it is fairly easy to determine whether to purchase that portion as paid all up-front reserved or on-demand. The example in Figure 1 illustrates the break-even point for the 3-year, paid all upfront Reserved Instances as compared to on-demand instances.

We used the average discount of 60% for the 3-year commitment. In the example, assuming the infrastructure is going to be used for a multi-year time period, it shows that if the total amount of utilization time is more than 40% (4.8 months per year,) it is more cost effective to purchase the infrastructure with a 3-year, paid all upfront reserved option. If the duration of the project is less than one year, with no other projects to use the infrastructure, or the usage requirement is quite “bursty,” then it probably makes sense to purchase the infrastructure using the on-demand option when only comparing AWS purchase options.

Storage and Elastic Block Store (EBS)

The AWS S3 EBS offering provides network-accessible, persistent block storage for certain EC2 compute instances. In fact, many of the instance families only support EBS for boot and data storage. The network-attached block devices range from cold storage to throughput-optimized rotational drives to general-purpose SSDs to provisioned IOPS SSDs. The configuration of these devices offers quite a bit of flexibility, including the ability to dial up IOPS to meet workload optimization requirements.

AWS Simple Monthly Calculator

In just minutes, a customer can price-out AWS EC2 and S3 configurations using the AWS Simple Monthly Calculator. This tool allows you to select the region, the number, size and types of instances, shared versus dedicated versus reserved, operating system and hypervisor, the EBS device types, IOPS specification for provisioned SSDs, monthly data transfer in/out assumptions and several other options we haven’t mentioned here. It is worth spending an hour to become familiar with the tool that provides monthly and upfront pricing for the prescribed infrastructure.

In the HPE On-Prem vs. AWS technical paper, we attempted to be objective and fact-based in our comparisons—and we hope this analysis comes across that way. In the next four blogs, we go deeper into each section of the analysis and also share some additional findings not presented in the paper.

Follow the blog series

In this blog series, we present details and insights around the HPE and AWS comparison for the following topics:

Part 1: Introducing the HPE on-prem vs. AWS primer

  • An overview of the study and the summary of findings in the comparison of HPE and AWS
  • An AWS primer to provide a brief overview of what is available in AWS EC2 IaaS capabilities

Part 2: Total Cost of Ownership (TCO)

  • The configuration options and selected configurations: HPE and AWS
  • The “all-in” cost analysis for the on-prem configuration, including costs for maintenance labor, data center infrastructure, energy and cooling, carbon footprint and warranty
  • The cost analysis for the AWS configurations based on reasonable configurations and purchase options

Part 3: The workload and throughput (performance) 

  • Overview of the cloud-scale advanced analytics workload
  • The throughput measurements for each configuration in total Queries per Minute (Qpm)
  • Analysis of each architecture and comparison of Local SSDs and EB

Part 4: Price-performance and control 

  • Price-performance for each configuration
  • A look at the cost of high-throughput EBS volumes
  • Analysis of ability to control attributes of performance, data sovereignty, privacy, security

Part 5:  Elastic IT experience

  • Pay per use with variable payments based on actual metered usage
  • Dynamic and instant growth flexibility
  • Onsite extra capacity buffer

* Gartner estimates in an October, 2017 article, a 30% average CAGR in public cloud IaaS adoption through 2020.

 Lou Gagliardi.jpgMeet Infrastruture Insights blogger Lou Gagliardi, Sr. Lab Director, Enterprise Solutions & Performance, Hybrid IT, HPE.

Lou joined Compaq Computer Corporation in 1988 and has held executive-level positions in server and storage development engineering and WW presales while at Compaq, Hewlett-Packard, Dell, Newisys and Spansion. He returned to HP in 2010 as Sr. Director of Integrated System Test. In 2013, he transitioned to his current role to deepen HPE’s understanding of the character of the new style of IT workloads.

Infrastructure Insights
Hewlett Packard Enterprise




About the Author


Our team of HPE and other technology experts shares insights about relevant topics related to artificial intelligence, data analytics, IoT, and telco.