A.  

The server is assigned a static IP address with connectivity to OV4VC.

B.  

The cluster’s server profile assigns NIC0 to a network with connectivity to the provisioning server.

C.  

The server has HPE drivers and HPE Agentless Management Software (AMS) pre-provisioned on it.

D.  

The server has UEFI optimized mode with Secure Boot enabled on it.

A.  

Support for 4 processors and 4 Lovelace GPUs per 5U chassis, combined with a choice of iLO or BMC management

B.  

Support for up to 8 NVIDIA Blackwell GPUs per 5U chassis, combined with iLO management

C.  

Scale-out design with up to 8 single-processor nodes per 5U chassis

D.  

Scale-up design with 8 processors and 8TB of memory per chassis

A.  

External access is enabled by default at the Ray dashboard UI. The external client should direct the ray.init() requests to TCP port 10001.

B.  

External access is not supported with the built-in KubeRay cluster. The customer should deploy a new cluster to achieve this access.

C.  

External access is not available by default, but can be achieved by, for example, changing the kuberay-head-svc to a NodePort service.

D.  

External access is enabled by default at the Ray dashboard UI. The external client should direct the ray.init() requests to TCP port 8265.

A.  

Restart the pod that generated the alert

B.  

Configure alert forwarding rules

C.  

View logs from the pod associated with the alert

D.  

Correlate the alert with other alerts from the HPE GreenLake cloud workspace

A.  

From within the HPE GreenLake cloud User management page, assign the Airflow Admin role to HPE Private Cloud AI Administrators.

B.  

Configure the Airflow framework and change the value for the access_control line to “Admin.”

C.  

Within the Airflow UI, go to the Security > List Roles page and assign the “trigger” permission to the Admin role.

D.  

Add an access control line to the DAG code that permits “Admin” the “can_edit” privilege.

A.  

It replaces IP ECN with PFC to provide more sophisticated congestion control for large AI flows.

B.  

It uses ECMP to provide fine-grained load balancing and to saturate redundant links.

C.  

It uses dedicated per-port buffers to ensure losslessness for RDMA traffic.

D.  

It uses targeted Flow Metering to manage congestion for bursty AI traffic.

A.  

Go to the Identity and access management page in the HPE AI Essentials UI. Manage the privileges of individual users to grant them access to the object store.

B.  

Go to the User Management page within HPE GreenLake cloud. Select the user and add privileges to access the data source from there.

C.  

Configure the data source settings and select the users permitted access within those settings.

D.  

Manually create an “access-token” secret with the data store credentials within the namespace of each user who should be permitted access to the store.

A.  

Receiving ingested data and preparing it for storage in a RAG database

B.  

Submitting queries to an unstructured database

C.  

Determining which vectors retrieved from the RAG database best match a query

D.  

Generating responses to queries submitted by users

A.  

Integrating the file and object storage on the AI-optimized node

B.  

Using CPU-only AI-optimized nodes

C.  

Reducing the number of software tools integrated on the system

D.  

Eliminating support for NVIDIA AI Enterprise

A.  

It uses InfiniBand NDR to reduce the average end-to-end latency for MPI communications.

B.  

It uses bandwidth efficiently with packet-by-packet adaptive routing for MPI and Lustre.

C.  

It provides advanced congestion management based on DiffServ.

D.  

It provides the end-to-end non-blocking architecture required for typical HPC applications.

A.  

MLflow

B.  

NeMo Customizer

C.  

Katib

D.  

NeMo Curator

A.  

Dynamically moving VMs away from a host that is experiencing hardware failures

B.  

Selecting the local drives used for the vSAN capacity tier

C.  

Provisioning a datastore from a connected SAN array

D.  

Applying workflows that incorporate vCenter and OneView tasks