Jump to the main content block
:::

Host and Cloud Services

:::
  1. Home
  2. Services
  3. Host and Cloud Services
  4. AI Computing Host (Rented for Pay)

AI Computing Host (Rented for Pay)

Introduction to AI Computing Services

  • In order to meet the needs of the University for in-depth learning, machine learning, and big data analysis, the Center introduced AI supercomputers and related software and hardware facilities to build an AI platform to support teaching and research in the fields relevant to in-depth learning, machine learning, and big data analysis.
  •  AI computing services (hereinafter referred to as “Services”) refer to the service mode in which a user can obtain GPU computing, network, storage, and other resources approved and allocated by the Center through the network.
  • The Services are provided based on the principle of user payment. Users can prepare and list out use fees of AI computing services, computer use fees, program balance, department/institute management fees, and related fees in their research programs, and pay the use fees or store value.
  • The revenue from the Services is used for system development, equipment augmentation, troubleshooting, replacement of the old with the new, and employment of relevant personnel related to the Services under the principle of earmarking a fund for designated purpose only.

Application Process for AI Computing Services

  • After the AI Application Form is reviewed by the Computer and Network Center (handler -> team leader -> director), the AI resources are opened and users are notified. Users log in to the system to check the resources, and during the test period (7 days), inspect whether the environment is available. After the test period expires, the system will send an E-mail confirmation letter to notify users to confirm the situation and reply to the AI platform administrator.
  •  Large or special programs will be submitted to the Center for review.
  • The Computer and Network Center provides users with quotations, receipts, and other documents to make the application for purchase. After the receipt is issued, users can store value on the AI platform. If the above E-mail confirmation letter is not replied to after 3 days, or does not meet the requirements, the AI resources applied for by users will be closed.
  • Users must be a faculty or staff member of the University, and can use a log-in Cheng Kung account number to log in to the AI platform (http://ai.cc.ncku.edu.tw) to fill in the Application Form and raise demands for AI computing resources. The applicant is usually a program host, who pays the use fees with the relevant funds of the program.
  • Users shall use the AI resources as applied for, and the system will: 
    • Deduct the fees hourly, which shall be paid on a pay-as-you-go basis, according to GPU package resources; or
    • Deduct the fees on a lump sum basis after opening according to network storage services.
  • When the stored value in the system is less than 20%, users will be notified by E-mail.
  • The system will notify users to extend the resource application period on the 7th day and 1st day before its expiration date.
  • If the stored value in the system is zero, or the application period has not been extended after expiration, the resources still in operation will not be automatically terminated, and the system will continue to deduct the fees until the user closes the container. Those who have not paid the fees owed will not be allowed to create new resources. 

Application Process for AI Computing Services

Renting Notices for AIComputing Services

  • After opening the AI computing resources, the user (Ex.: the program host) can have other members’ account numbers (Ex.: laboratory students) join in this resource pool to share these AI computing resources.
  • Users must have knowledge relevant to operation and use of Linux, Container, Machine Learning, etc. Users can create a container computing environment based on a template provided by the system or by a customized method.
  • Users can create the container computing environment based on the template provided by the AI platform or by the customized method. The container provides users with full permissions to install packages by themselves.
  • This platform applies a multi-tenancy mechanism, that is, users can belong to multiple tenants. The tenant classification design allows users to be divided into different groups according to projects, programs, department development, or school competitions, etc., and configure different resources for different groups.
  • The Computer and Network Center is an AI computing resources provider, and provides users with containers created to use the GPU computing resources. Due to limited manpower, it is unable to assist individual users in making program development and debugging. 

Use Process for AI Computing Resources

Use Process for AI Computing Resources

Charging Methods 

  • Pursuant to Point 4 of the "Key Points of AI Computing Service Management of National Cheng Kung University", the AI computing services (hereinafter referred to as the Services) are provided based on the principle of user payment. Users can prepare and list out use fees of AI computing services, computer use fees, program balance, department/institute management fees, and related fees in their research programs, and pay the use fees or store value.
  • The charging standards for the Services are divided into "GPU computing services" and "network storage services". Users can select the number of GPU according to their computing demand; After the number of GPU is selected, the system sets the corresponding CPU, memory, and hard disk space capacity according to the Charging Table for GPU Computing Services, and defines the AI resource pool for users.
  • The GPU Computing Services applies a mechanism of pay-as-you-go. Users can generate containers to perform computing within the scope of the resource pool specifications applied for. After the containers are powered on, the billing starts and the fees are deducted from the stored value. After the containers are powered off, the deduction of the fees is stopped.
  • Users who need more than 200GB of storage space can apply for additional "network storage services". After the "network storage services" are opened, the fees will be fully deducted on a lump sum basis (not pay-as-you-go). After opening, the specifications and duration cannot be reduced.
  • Once the Services have been paid and entered into an account, no refund service can be provided.

 

Charging Table for v100 32GB GPU Computing Services
(updated on 11.25. 2020)

GPU (piece) CPU (core) Memory (GB) Hard Disk Space (GB) Pay-as-you-go (TWD/hour)
1 4 62 200 10
2 8 124 20
3 12 186 30
4 16 248 40
5 20 310 50
6 24 372 60
7 28 434 70
8 32 496 80

 

Charging Table for v100 16GB GPU Computing Services
(updated on 11.25. 2020)

GPU (piece) CPU (core) Memory (GB) Hard Disk Space (GB) Pay-as-you-go (TWD/hour)
1 4 62 200 5
2 8 124 10
3 12 186 15
4 16 248 20
5 20 310 25
6 24 372 30
7 28 434 35
8 32 496 40

Charging Table for Network Storage Services
Storage Specifications Hard Disk Space (GB) Unit Price (TWD)/day
All-Flash(SSD)
 10 5

 

Contact Person

  • Mr. Wang, Miss Chen
    • Extension: 61029, 61024
    • E-mail: 招先生的email