slurm queues: disable SMT on intel nodes and add m7a and m7i

* Disable SMT on c7i nodes by default

* Update _slurm_queues.tpl.yaml

Add another queue with memory-optimized nodes (m7a and m7i).

* Update _slurm_queues.tpl.yaml

Add another queue with memory-optimized nodes (m7a and m7i) in config-dev.

Author: jplanasc

hpc_provisioner/src/hpc_provisioner/config-dev/_slurm_queues.tpl.yaml

@@ -116,6 +116,42 @@
         - InstanceType: c7i.48xlarge  # compute optimized nodes
       MinCount: 0
       MaxCount: 20  # least number of nodes needed to simulate the full O1 circuit x2
+      DisableSimultaneousMultithreading: true
+      Efa:  # low-latency, high BW network
+        Enabled: true  # https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/efa-start.html#efa-start-security
+  Networking:
+    PlacementGroup:  # try to place nodes close to each other
+      Enabled: true
+    SubnetIds: [!config base_subnet_id]
+    SecurityGroups:
+      - !config base_security_group_id
+      - !config efa_security_group_id  # Efa
+  CustomSlurmSettings:
+    MaxNodes: 20
+    MaxTime: 720
+  Iam:
+    S3Access:
+      - BucketName: sboinfrastructureassets-sandbox
+
+# ==============================================================
+# prod-mpi-mem queue, for tightly-coupled, memory-intensive jobs
+# ==============================================================
+- Name: prod-mpi-mem
+  AllocationStrategy: lowest-price  # usually on-demand
+  ComputeResources:
+    - Name: cpu-m7a
+      Instances:
+        - InstanceType: m7a.48xlarge  # memory-optimized nodes, AMD arch
+      MinCount: 0
+      MaxCount: 20  # least number of nodes needed to simulate the full O1 circuit x2
+      Efa:  # low-latency, high BW network
+        Enabled: true  # https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/efa-start.html#efa-start-security
+    - Name: cpu-m7i
+      Instances:
+        - InstanceType: m7i.48xlarge  # memory-optimized nodes, Intel arch
+      MinCount: 0
+      MaxCount: 20  # least number of nodes needed to simulate the full O1 circuit x2
+      DisableSimultaneousMultithreading: true
       Efa:  # low-latency, high BW network
         Enabled: true  # https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/efa-start.html#efa-start-security
   Networking:

hpc_provisioner/src/hpc_provisioner/config/_slurm_queues.tpl.yaml

@@ -116,6 +116,42 @@
         - InstanceType: c7i.48xlarge  # compute optimized nodes
       MinCount: 0
       MaxCount: 20  # least number of nodes needed to simulate the full O1 circuit x2
+      DisableSimultaneousMultithreading: true
+      Efa:  # low-latency, high BW network
+        Enabled: true  # https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/efa-start.html#efa-start-security
+  Networking:
+    PlacementGroup:  # try to place nodes close to each other
+      Enabled: true
+    SubnetIds: [!config base_subnet_id]
+    SecurityGroups:
+      - !config base_security_group_id
+      - !config efa_security_group_id  # Efa
+  CustomSlurmSettings:
+    MaxNodes: 20
+    MaxTime: 720
+  Iam:
+    S3Access:
+      - BucketName: sboinfrastructureassets-sandbox
+
+# ==============================================================
+# prod-mpi-mem queue, for tightly-coupled, memory-intensive jobs
+# ==============================================================
+- Name: prod-mpi-mem
+  AllocationStrategy: lowest-price  # usually on-demand
+  ComputeResources:
+    - Name: cpu-m7a
+      Instances:
+        - InstanceType: m7a.48xlarge  # memory-optimized nodes, AMD arch
+      MinCount: 0
+      MaxCount: 20  # least number of nodes needed to simulate the full O1 circuit x2
+      Efa:  # low-latency, high BW network
+        Enabled: true  # https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/efa-start.html#efa-start-security
+    - Name: cpu-m7i
+      Instances:
+        - InstanceType: m7i.48xlarge  # memory-optimized nodes, Intel arch
+      MinCount: 0
+      MaxCount: 20  # least number of nodes needed to simulate the full O1 circuit x2
+      DisableSimultaneousMultithreading: true
       Efa:  # low-latency, high BW network
         Enabled: true  # https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/efa-start.html#efa-start-security
   Networking: