# 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="https://4272811953-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-LcWJ6pKCQH6l3rl9IQa%2Fuploads%2FedIMk4Mgfv2vVXwwdPSm%2FDALL%C2%B7E%202023-10-29%2005.56.22%20-%20Photo%20of%20a%20modern%20data%20center%20with%20rows%20of%20servers%20glowing%20in%20blue.%20Overlaying%20the%20scene%20is%20a%20leafy%20vine%2C%20symbolizing%20nature%2C%20winding%20its%20way%20through%20.png?alt=media&#x26;token=4b2cc467-47bf-46d4-9f0d-e2b08d15536a" alt="" width="375"><figcaption><p>Idle nodes enter power saving mode automatically</p></figcaption></figure> <figure><img src="https://4272811953-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-LcWJ6pKCQH6l3rl9IQa%2Fuploads%2FFDHlWfcAW2Wv2C6Xg8N7%2FScreenshot%202023-10-29%20at%206.52.18%20AM.png?alt=media&#x26;token=5e9ea0fd-0174-43b3-93f2-b94a87f3cb94" alt=""><figcaption><p>Unified Storage across all OIT data platforms</p></figcaption></figure> <figure><img src="https://4272811953-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-LcWJ6pKCQH6l3rl9IQa%2Fuploads%2FRHCuDiUMNFEYDyo9xAgh%2FScreenshot%202023-10-29%20at%206.12.44%20AM.png?alt=media&#x26;token=d36335e8-ab55-4704-b55b-a92fb2be5bb6" 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.