JamesRunnalls / hydrosim

Repository for information related to the Hydrosim project.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

hydrosim

Repository for information related to the Hydrosim project.

Compute Requirements

For the project two distinct servers are required with different characteristics:

Web Server

Server to host the node.js API that will make the data available to the front end & for direct download of the simulation results.

Requirements

  • Large storage capacity (>10TB)
  • The ability to read from NetCDF files and return JSON files (rules out static file providers like S3?)
  • Multiple cores to handle API requests in parrallel
  • Limited computation requirements

Compute Server

Server to run the hydrodynamic simulations and then transfer the results to the web server

Requirements

  • Large RAM, high performance CPU
  • Ability to run computationally intensive simulations in parrallel
  • Limited storage requirements (< 100GB)
  • May need to be runnning windows (possible)

Hardware Options

Eawag

Option 1 is to purchase the two machines and have the Eaway IT department run and maintain them.

Pros:

  • Full access to the machines
  • Unlimited customisation
  • No monthly costs
  • Storage is very cheap
  • Easy to align with existing procurement proceedures and include in project budget
  • Full Windows support

Cons:

  • Large upfront cost (>$10K)
  • Maintenance and repairs limited to office hours
  • Limited scalability (if server spec is wrong little option to make changes)
  • Catastrophic hardware issues could stop the project
  • Long setup time (define server specs, procurement, etc.)
  • Eawag is responsible for the backups
  • Dependant on the Eawag network, any Eawag outages will effect the performance of the API

Cloud Provider (e.g. AWS)

Option 2 is to use pay by use servers from a cloud provider such as AWS.

Pros:

  • Rapid deployment (< 1hr)
  • Scalability (fast and easy to increase compute power/ storage)
  • Option to spawn multiple low power servers rather than running simulations in parrallel
  • No upfront cost
  • Many services in one place (database, frontend (AWS amplify), S3 etc.)
  • 24hr service and maintenace

Cons:

  • Monthly costs can grow rapidly (hard to align with project budget)
  • Elastic block storage is expensive

Dedicated Cloud Server (e.g. OVHcloud)

Option 3 is to purchase dedicated cloud servers.

Pros:

  • Quick deployment (< 24hrs)
  • No upfront cost
  • Many services in one place (database, frontend (AWS amplify), S3 etc.)
  • 24hr service and maintenace
  • Monthly cost is predictable and lower than for virtual machines

Cons:

  • Difficult to scale
  • Costly to have multiple servers
  • May not be possible to get the exact server specs you want

About

Repository for information related to the Hydrosim project.