> For the complete documentation index, see [llms.txt](https://docs.ccv.brown.edu/oscar/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.ccv.brown.edu/oscar/sys-changes/rhel-9-migration.md).

# RHEL-9 Migration

1. **What is the reason for this maintenance?**

   Oscar's current operating system, RHEL-7, and its maintenance support phase will come to an end in June 2024. So Oscar is being upgraded to latest RedHat Enterprise Linux RHEL-9.2

   <div data-gb-custom-block data-tag="hint" data-style="info" class="hint hint-info"><p>Due to the new kernel and glibc majority of applications will break.</p></div>
2. **We are also introducing a bunch of new features:**

   * **Upgraded OS -** The OS has been upgraded with a newer kernel and improved security patches
   * **Power Saving Mode:** In the batch partition, idle nodes now enter a power-saving mode, consuming only about 40W. They seamlessly transition to high-performance mode just before job execution begins
   * **GPU Direct Storage:** GDS enables a direct data path for direct memory access (DMA) transfers between GPU memory and storage, which avoids a bounce buffer through the CPU. This direct path increases system bandwidth and decreases the latency and utilization load on the CPU
   * **SLURM Upgrade:** We have tuned the scheduler to provide much higher throughput. Now supports `json` and `yaml` formatting for all slurm commands
   * **SPACK & LMOD -** Newer industry standard for installing and managing applications on Oscar. We now support multple shells `bash,zsh & fish` etc
   * **Increased-core core count for GPU accounts:**

   <table><thead><tr><th width="226">Account</th><th width="100">Partition</th><th>Current core-limit</th><th>New core-limit</th></tr></thead><tbody><tr><td>Exploratory</td><td>gpu</td><td>4-cores</td><td>12-cores</td></tr><tr><td>Standard GPU Priority</td><td>gpu</td><td>16-cores</td><td>24-cores</td></tr><tr><td>Standard GPU Priority+</td><td>gpu</td><td>32-cores</td><td>48-cores</td></tr><tr><td>High-End GPU Priority</td><td>gpu-he</td><td>16-cores</td><td>24-cores</td></tr></tbody></table>

   <div align="left"><figure><img src="/files/83aqCmOBErvykqsapWZk" alt="" width="375"><figcaption><p>Idle nodes enter power saving mode automatically</p></figcaption></figure> <figure><img src="/files/32Vq2t70DTpOOFe7lFVM" alt=""><figcaption><p>Unified Storage across all OIT data platforms</p></figcaption></figure> <figure><img src="/files/LAX4aj2rHlIieGB2Vn8n" alt=""><figcaption><p>GPUDirect Storage - Lower laency &#x26; Higher Bandwidth for IO</p></figcaption></figure></div>
3. **What are exact version changes?**

   <table><thead><tr><th width="199.33333333333331">Component</th><th width="300">Current Version</th><th>New Version</th></tr></thead><tbody><tr><td>Operating System</td><td>RHEL-7.9</td><td>RHEL-9.2</td></tr><tr><td>Kernel</td><td>3.10.0-1160.76.1</td><td>5.14.0-284.11.1</td></tr><tr><td>GLIBC</td><td>2.17-326</td><td>2.34-60</td></tr><tr><td>SLURM</td><td>22.05.7</td><td>23.02.6</td></tr><tr><td>Nvidia Driver</td><td>535.54.03</td><td>535.113.01</td></tr><tr><td>Package Manager</td><td>PyModules</td><td>SPACK</td></tr></tbody></table>
4. **How to access the new cluster?**

   We will provide detailed instructions in coming weeks. Thank you for your patience.