Red Oak Consulting is a leading independent HPC and Cloud consultancy based in the UK. Our services cover the whole of the HPC and cloud lifecycle: strategy planning, expansion & optimisation, procurement, implementation and managed services.

Understanding and performing TCO analyses allows you to better manage your HPC investments and ensure they align with your organisations financial and operational goals. When considering Cloud Solutions a HPC TCO analysis will compare the costs of on-premises Vs cloud options, factoring in variables like scalability, flexibility, storage and long-term expenses.

To assist customers Red Oak Consulting has developed a unique, specialised TCO model to help companies achieve accurate results and insights, allowing them to make the right business investments tailored to their specific needs.

Understanding your total cost of ownership (TCO) when running HPC is Traditional total-cost-of-ownership (TCO) models tend to adopt a cash-accounting approach, focusing purely on “money out of the door.” Often these models focus exclusively on local costs, ignoring costs elsewhere in the business.

The traditional approach however leads to a false (often optimistic) view of total cost. They certainly do not reflect the total cost to the organisation. Our approach to TCO modelling is a two phased approach. First, we seek to identify and establish an accurate account of the general accounting principles associated to the full running of HPC:

approach ongoing costs holistically across the organisation, capturing all costs;
use full capitation rates for staff, etc;
include estimates of inflation (for forward looking models);
seek to measure depreciation charges for infrastructure;
explore costs of capital;
factor in R&D tax credits if appropriate;
attempt to measure opportunity costs;
inform results in the light of corporately agreed targets for internal rate of return (IRR).

This analysis will provide a clearer base cost for the existing, HPC service, in totality.

We then progress to phase two which is where we then follow an analysis of the service outputs to provide a comparable metric cost (typically cost per delivered core-hour). This typically includes:

system availability;
system utilisation;
distribution of job sizes and run times;
job wait times;
job failure rates (due to system instability).

To deliver a more cost-effective output we look to take a risk-balanced approach to the analysis, iteratively improving the model through repeated error-reduction, i.e. focusing effort on those areas which will lead to the most reduction in numerical uncertainty of the result.

The outcome and deliverable from completing the above two phases will be:

A catalogue of cost input items including provenance and uncertainty estimates;
Comparative whole-life TCO model providing cost for the indicative workload.

Please be advised the results will be presented with supporting documentation in the form of a slide deck Model assumptions

The assumptions used in the model will be clearly documented. The key assumptions (with a largest impact on the study outcomes) are likely to be concerned with:

the evolution of the target workload, its mix and the Azure instances best suited to it;
storage costs;
the presumed rate of decrease of Azure costs over the period of study;
the impact of Azure pricing tiers (on-demand, spot, reserved, etc)

As with all projects there are dependencies which are associated to a TCO project, these include:

Red Oak Consulting will need access to appropriately empowered staff able to:

Customer to provide historic information about the test workloads used at the basis of the study;
Confirm assumptions about the evolution of the workloads over the period of the study;
Provide answers to questions about extant cost items;
Represent the diverse Customer stakeholder community and, if appropriate, facilitate up to four stakeholder interviews.

Please be advised that the project can only commence once all the data and information has been confirmed and been made available to Red Oak Consulting. Any delays in supplying the necessary scheduler logs will result in Red Oak Consulting applying industry averages.

This project has been timeboxed to 12 days of time to complete. If further / extra days are required, then an additional PO will be required and Red Oak will charge the daily rate of the project consultant(s).

High Performance Computing (HPC) Total Cost of Ownership

bởi Red Oak Consulting