Managing Projects

Projects in GPU Cluster Manager serve as logical groupings for your workloads, resource allocations, and operational priorities. Each Project is associated with a specific Department, ensuring that resources are organized according to your organization’s structure (e.g., Corporate, Research, Engineering). By configuring Projects, you define the CPU, memory, and GPU resources allocated to the workloads under that Project, as well as a priority level indicating how resources are scheduled when multiple workloads compete.

Key Concepts

1. Department Dependency
   Every Project is linked to a Department. You must have at least one Department created before you can create a Project.

2. Resource Allocation

   • CPU & Memory: Specify the amount of CPU and memory that your Project can use.
   • GPUs: Assign GPUs from one or more Node Groups, enabling AI or HPC workloads that require GPU acceleration.

3. Workload Priority
   When multiple Projects share resources, the Priority setting determines which Project’s workloads are favored in times of resource contention. Supported priority levels include:

   • Lowest
   • Low
   • Medium
   • High
   • Highest

   Use this feature to ensure mission-critical workloads receive the resources they need first.

---

Accessing the Projects Dashboard

1. Open the Projects View

   • The Projects Dashboard is the default view when logging in to the User Account.
   • In addition, you can navigate to the Projects Dashboard from other Dashboards by clicking on the Projects Button on the left-side navigation bar.
   • Below is a view of the Projects Dashboard:

   

Elements of the Projects Dashboard

The Projects dashboard gives you a clear overview of all your assigned projects and their current status within MemVerge.AI. This is your go-to place to monitor your project's resource usage and the activity of your workloads.

The screen displays a list of your projects in a table format, offering various ways to understand and manage them:

Search Projects

You can easily find a specific project by typing its name into the Search Projects field located at the top left of the project list.

Project List Table

Each row in the table represents one of your projects and provides the following information:

• Name: The name of your project (e.g., proj-blue). Click on a project name to view more detailed information about that specific project.
• Priority: Indicates the priority level assigned to your project (e.g., high), which can influence resource allocation.
• Department: Shows the department or team that the project belongs to (e.g., dev-bda).
• Fractional GPU: Displays specific details about fractional GPU usage, if applicable.
• Reserved / Allocated GPUs: Shows the total number of GPUs reserved for your project versus how many are currently allocated to active workloads (e.g., 0 / 0).
• Reserved / Allocated CPU Cores: Displays the total CPU cores reserved for your project versus how many are currently allocated (e.g., 0 / 0).
• Reserved / Allocated Memory: Shows the total memory (RAM) reserved for your project versus how much is currently allocated (e.g., 0G / 0G).
• Total Workloads: The total count of all workloads associated with your project, regardless of their status.
• Running Workloads: The number of workloads from this project that are currently active and executing (e.g., 0).
• Pending Workloads: The number of workloads from this project that are waiting to be started or are in a queued state (e.g., 0).
• Created At: The date and time when your project was originally created (e.g., 2025-06-12 10:47:34).
• Actions: This column is where you would typically find icons or buttons to perform actions on a specific project, such as editing project settings or viewing more details. (For users, there are no actions available. To ensure users are not taking other user resources, management of projects is performed at the administrator level)

Pagination Controls:

• At the bottom right of the table, you'll find navigation controls (e.g., 1 / 10 / page) that allow you to move between pages if you have many projects listed. You can also adjust how many projects are shown per page.

How You Can Use This Page:

• Monitor Resources: Quickly check how many GPUs, CPU cores, and memory are reserved for your projects and how much is actively being used.
• Track Workload Status: See at a glance how many of your project's workloads are running or waiting.
• Identify Project Details: Get a summary of your project's priority, department, and creation date.

• Navigate to Project Details: Click on a project's name to explore more in-depth information about it.

• View a Project

  • Look for the project in the table provided at the bottom of the dashboard.
  • Alternatively, you can use the Search Projects box at the top to find a specific project by name.

---

Viewing Detailed Project Information

To view more information on a Project, within the Projects Dashboard, click on a project within the Name Column. You will see the following screen for the Project you clicked on, whith the default, Workloads Tab, open:

The screen is organized to give you a clear picture of your project's status and its associated workloads.

Project Summary

At the top of the Project Details view, you'll see a summary of your selected project:

• Project Name: The name of your project (e.g., Project: proj-blue).
• Tags: Any labels or priorities associated with your project (e.g., High, dep-blue).
• Node Groups: The number of node groups assigned to your project, indicating the infrastructure it can use (e.g., 1).
• Created Date: The date and time your project was created (e.g., 2025-06-12 10:47:14).
• Reserved / Allocated Resources: This section shows the total resources (CPUs, Cores, Memory) that are reserved for your project versus how much is currently being used by its active workloads. In the example, it shows 0 / 0 for all resources, meaning no resources are currently allocated or in use.

Project Detail Tabs

Below the summary, there are tabs to navigate to different aspects of your project:

• Workloads (Active Tab): This tab, which you are currently viewing, lists all the individual tasks and jobs running or planned within this project.
• Metrics: This tab would typically show performance graphs and resource utilization trends specific to this project.
• Settings: This tab would allow you to view or potentially modify specific settings for your project (depending on your permissions).

Workloads Overview & List

Under the Workloads tab, you'll find a breakdown of all workloads in your project:

• Workload Status Summary: A quick count of workloads by their status:
• All: Total workloads.
• Pending: Workloads waiting to start.
• Running: Workloads currently executing.
• Succeeded: Workloads that completed successfully.
• Failed: Workloads that encountered an error and stopped.
• Evicted: Workloads that were stopped due to resource needs elsewhere.
• Preempted: Workloads that were stopped to make way for higher-priority tasks. (Note: In the example image, all counts are 0, indicating no active or completed workloads for this project currently.)
• Workloads Table: This table lists each individual workload with key details:
• Name: The name of the workload.
• UID: A unique identifier for the workload.
• Status: The current state of the workload (e.g., Running, Pending, Failed).
• Submitted: The date and time the workload was submitted. (Note: The message "No data" indicates that there are currently no workloads to display for this project.)

Metrics Overview & List

Under the Metrics tab, you'll find graphical reporting on GPU Utilization, GPU Memory Utilization, CPU Utilization, and CPU Memory Utilization for the specified project over time.

Settings

The Settings tab of Project Details provides information on resource limits that your project has set:

How You Can Use This Page

• Monitor Resources: Quickly check how many GPUs, CPU cores, and memory are reserved for your projects and how much is actively being used.
• Track Workload Status: See at a glance how many of your project's workloads are running or waiting.
• Understand Project Configuration: Review the project's name, priority, department, and creation date.
• Access Detailed Metrics: Navigate to the Metrics tab to view performance trends over time for your project's resources.
• Review Project Settings: Use the Settings tab to check the configuration details of your project.