---
title: "Partitions and HPC Resources"
canonical: "https://kb.uconn.edu/space/SH/28752281629/Partitions%20and%20HPC%20Resources"
format: markdown
---
# Partitions

Nodes on the Storrs HPC are organized into **partitions** — groups of nodes with shared features and usage limits. For instance, partitions with the -gpu suffix contain nodes with GPUs, allowing for accelerated performance for programs designed to utilize them. All users have access to **general** partitions, while additional access to **priority** partitions can be purchased.

### General partitions

| **Partition** | **Time limit** | **Resource limit** |
| --- | --- | --- |
| general | 12 hours | 8 node limit per job |
| lo-core | 7 days | 2 node limit per job |
| hi-core | 6 hours | 16 node limit per job |
| general-gpu | 12 hours | 6 GPUs per user |

### Priority partitions

| **Partition** | **Time limit** | **Resource limit** |
| --- | --- | --- |
| priority | unlimited | [QoS limit*](https://kb.uconn.edu/space/SH/26698678562/Priority+Access) |
| priority-gpu | unlimited | [QoS limit*](https://kb.uconn.edu/space/SH/26698678562/Priority+Access) |

# HPC resources and availability

There are two main classes of standard nodes on the HPC cluster.

| **Name** | ***Available***** cores per node**** | **RAM (GB)** |
| --- | --- | --- |
| Epyc64 | 62 | 487 |
| Epyc128 | 126 | 487 |

** Two cores are reserved per node for OS and storage processes.


Multiple types of GPU nodes are available on the Storrs HPC as well. The majority are Epyc64 nodes with between 1 and 3 GPUs.

| **Name** | **Available cores per node** | **RAM (GB)** | **GPU** | **Number of GPUs** |
| --- | --- | --- | --- | --- |
| Epyc64, a100 | 62 | 487 | NVIDIA-A100 | 1 or 3 |
| Epyc64, a30 | 62 | 487 | NVIDIA-A30 | 2 or 3 |

### Checking available resources

To check what nodes are present/available for jobs, use the `nodeinfo` command and specify the partition with the `-p` flag.

Example:

```
[jth12345@login5 ~]$ nodeinfo -p general-gpu
PARTITION      NODES  STATE  TIMELIMIT   CPUS    GRES MEMORY   ACTIVE_FEATURES   NODELIST
general-gpu        4   mix-   12:00:00    64+   gpu:3 515000+  epyc64,a100,gpu   gpu[13,35-36,38]
general-gpu        1  maint   12:00:00     64   gpu:1 515000   epyc64,a100,gpu   gpu25
general-gpu        1   unk*   12:00:00     64   gpu:3 515000   epyc64,a100,gpu   gpu15
general-gpu        2   unk*   12:00:00     64   gpu:1 515000   epyc64,a100,gpu   gpu[16,31]
general-gpu        7    mix   12:00:00     64   gpu:3 515000   epyc64,a100,gpu   gpu[14,20-22,29,37,39]
general-gpu        8    mix   12:00:00     64   gpu:1 515000   epyc64,a100,gpu   gpu[17,19,24,26-28,32-33]
general-gpu        4    mix   12:00:00     64   gpu:4 514500   gpu,l40           gpu[41-44]
general-gpu        5    mix   12:00:00     64   gpu:3 514900   epyc64,a30,gpu    gpu[46-50]
general-gpu        2    mix   12:00:00     64   gpu:2 773000   epyc64,a30,gpu    gpu[52,54]
general-gpu        4  alloc   12:00:00     64   gpu:1 515000   epyc64,a100,gpu   gpu[18,23,30,34]
general-gpu        1  alloc   12:00:00     64   gpu:3 515000   epyc64,a100,gpu   gpu40
general-gpu        2  alloc   12:00:00     64   gpu:2 773000   epyc64,a30,gpu    gpu[51,53]
general-gpu        1   idle   12:00:00     64   gpu:4 514500   gpu,l40           gpu45
```

Here, the GRES column reads out the number of GPUs, the ACTIVE_FEATURES column specifies the node architecture, NODELIST lists the nodes, and STATE tells you whether the node is available or not.

A state of “idle” means the node is free and ready for a job. “alloc” means it is completely allocated. Nodes in “mix” are partially allocated and may be open for smaller jobs. Statuses with an asterisk (*), i.e. unk*, indicate the node is unresponsive and therefore unavailable. 

> ℹ️ **Next:** [https://uconn.atlassian.net/wiki/spaces/SH/pages/28751527994](https://uconn.atlassian.net/wiki/spaces/SH/pages/28751527994)