Merge pull request #26 from vanderwb/network-fixes

Update ib0 references to ext
NCAR · Oct 7, 2024 · f084591 · f084591
2 parents 54a76a6 + cf5ec01
commit f084591
Show file tree

Hide file tree

Showing 2 changed files with 3 additions and 3 deletions.
diff --git a/notebooks/05-dask-hpc.ipynb b/notebooks/05-dask-hpc.ipynb
@@ -105,7 +105,7 @@
     "    resource_spec = 'select=1:ncpus=1:mem=4GB',\n",
     "    queue = 'casper',\n",
     "    walltime = '30:00',\n",
-    "    interface = 'ib0'\n",
+    "    interface = 'ext'\n",
     ")"
    ]
   },
@@ -120,7 +120,7 @@
     "\n",
     "* The `cores` and `memory` parameters are used by Dask to define workers, while the `resource_spec` is used by PBS to define jobs. In this single-worker config, they should match!\n",
     "* PBS uses *GB* to mean 1024-based storage units. `dask-jobqueue` accurately calls these `GiB`.\n",
-    "* We use `interface='ib0'` to instruct Dask to use TCP over Infiniband (the high-speed network on Casper), instead of the slower ethernet network.\n",
+    "* We use `interface='ext'` to instruct Dask to use TCP over the high-speed ethernet instead of other, slower, ethernet devices.\n",
     "\n",
     "Note also that we are using one worker per PBS job. This is a reasonable default on Casper, but it is possible to group workers together on one or more PBS jobs as well by increasing the `cores` and `ncpus`. Here are some considerations:\n",
     "\n",

diff --git a/notebooks/06-dask-chunking.ipynb b/notebooks/06-dask-chunking.ipynb
@@ -105,7 +105,7 @@
     "    resource_spec = 'select=1:ncpus=1:mem=10GB',\n",
     "    queue = 'casper',\n",
     "    walltime = '30:00',\n",
-    "    interface = 'ib0'\n",
+    "    interface = 'ext'\n",
     ")"
    ]
   },