TUFLOW Cloud Simulation Service

From Tuflow
Revision as of 17:03, 2 October 2019 by Chris Huxley (Talk | contribs) (Created page with "= TUFLOW Cloud Simulation Service = As of 2019, TUFLOW offer an on-demand cloud simulation service to assist when your company hardware or licensing becomes a limiting factor,...")

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

TUFLOW Cloud Simulation Service

As of 2019, TUFLOW offer an on-demand cloud simulation service to assist when your company hardware or licensing becomes a limiting factor, compromising your on-time project delivery.

This service is designed for bulk simulation of any TUFLOW product (Classic, HPC or FV). As such, it is typically used for design event or Monte Carlo modelling that requires 100’s to 1000’s of simulations of an established model. This service is less suited to model build activities, which often require regular modeller/model interactions. We recommend model build activities be completed using a standard client licence as it is typically a more cost effective option.

Key features of this service include:

  • Licences are scaled to meet your project demand (i.e. this is completely independent of any in-house licences you may own).
  • We have complete flexibility regarding the Microsoft Azure Cloud hardware available for your project. We tailor the hardware and storage resource specification to suit your project requirements (based on project and model size and also the simulation type).
  • This service has no limitations in terms of hardware type. We offer simulation on a range of CPU and GPU virtual machines.

How the Service Works

An overview of the TUFLOW Cloud Simulation Service steps is provide below:

  1. You provide us with simulation estimates for your project:
    • An estimate of the combined total run time for your project. To assist with this we will provide you with a TUFLOW benchmark model. It has a known simulation time on our cloud system (Azure). You can run the same model on your hardware and up or downscale the time as needed to calculate an equivalent run time on the cloud.
    • The maximum RAM required by a single simulation in your project. We require this information to assist in selection of the most cost effective cloud hardware.
    • How many separate simulation batches you require us to execute/manage. There is a service fee associated with each batch.
  2. We will provide you with a cost estimate for the model simulation project.
  3. Following acceptance and prepayment of the simulation cost estimate we will set up cloud storage for model inputs and outputs, providing access to our Microsoft Azure subscription or using your own subscription.
  4. You will upload your model files to the inputs storage and provide us with a list of the simulations you wish to run (a batch). The structure of your TUFLOW model folder and inputs is no different to how it would be configured on your own computer.
  5. We will test your model to ensure initialisation dependencies are error free. (Note: model debugging support is not included in this service. If this test identifies an initialisation error we will let you know. It will be your responsibility to fix it.)
  6. We manage the simulation of your models on the cloud. We will provide periodic updates of completion status during the project. Model results will be written to the cloud outputs storage as configured. Please note that the standard service includes the following:
    • The base cloud storage account is limited to 5TB per batch.
    • After each batch concludes two weeks of additional storage of the results is provided. For an extra fee the storage period can be extended.
  7. We offer a variety of result delivery options:
    • You will always have access to your result datasets during the project, whether it is within our Azure subscription or your own.
    • You can download the full project results directly from the cloud outputs storage with tools we provide you. Keep your company network data limits in mind when selecting this option.
    • We can upload the full project results to any cloud storage you may already have available.
    • We can request Microsoft Azure download the full project results dataset and ship it to you. That service comes at an additional cost (including the cost to purchase the required media).
  8. Repeat Step 4 though Step 6 as needed. There is a cost associated with each "batch", reflecting the amount of time required for us to manage your simulations.
  9. The project costs will be rationalised. If the final project costs are less than estimated you will be reimbursed. Alternatively, if the costs are more we will issue an invoice for the outstanding amount (Note: We monitor the project simulation status to ensure unexpected cost overrun does not occur. We provide you with regular updates so you are fully aware of the project status).

Price

Our software simulation service price includes the following line items:

  • Microsoft Azure compute, storage and download fees.
  • TUFLOW licence cost. This is calculated from an hourly rate. The rate has been derived from our standard TUFLOW rental rates.

TUFLOW SAS Rate Example 000.PNG

  • A service charge for our behind the scenes simulation/monitoring time.

There is a requirement for upfront payment of the estimated total project cost. If the actual total run time, data transfer and storage costs are less than the quote estimate we will provide a refund for the difference in cost. If the actual total run time, data transfer and storage costs are greater than the quote we issue an additional invoice for the difference in cost. We monitor the project simulation status to ensure unexpected cost overrun does not occur. We provide you with regular updates so you are fully aware of project status.

Model Optimisation Recommendations For Cloud Simulation

Depending on your available bandwidth past experience has found that model result download can be slow in some locations. As such, we recommend you reduce the data footprint of your model as much as practically possible. Some suggestions are:

  • Write time varying map results to XMDF format instead of DAT or WaterRide format. XMDF has a higher compression than the alternatives.
  • Consider an appropriate XMDF time interval to keep result file sizes small.
  • If high interval XMDF results are required consider limiting the extent of the results by using TUFLOW’s “Output Zone” feature.
  • Use maximum grid output (ASC, FLT or NetCDF) only and consider specifying a coarser resolution for your grid output (the default is half the 2D cell size). NetCDF is preferred as it is a compressed format. Don’t output grid results during a simulation at a defined interval (time varying 2D results are better written to the XMDF format).