HPE Advanced Compute Architect (HPE7-S01) Certification Sample Questions

01. A national lab partitions its HPE Cray EX system with HPE Cray System Management. Today an administrator creates and resizes each project's partition by hand in the management console. The lab's platform team wants its own automation pipeline to make these changes on request, with no console work.
Which capability should the design rely on?
a) The Slingshot Fabric Manager integrated in HPE Cray System Management
b) A documented console checklist that administrators follow for each partition change
c) HPE Performance Cluster Manager image groups, with one group per project
d) The REST APIs of HPE Cray System Management

02. A logistics company is buying HPE Private Cloud AI on subscription. It wants the statement of work to list what remains its own responsibility under the shared responsibility model.
Which two responsibilities stay with the customer?
(Select two.)
a) Installation and activation of the platform at the customer site
b) Ordering capacity beyond what the installed platform provides
c) Onsite hardware support when a platform component fails
d) Orientation for the customer's team after activation
e) Site readiness for the delivered racks

03. Researchers at an institute with an HPE Cray EX system want to run containerized AI training and bioinformatics tools pulled from public registries. The system's operators run Slurm and refuse both to grant users root privileges and to add a Kubernetes cluster to the machine.
Which approach meets both groups' requirements?
a) Install Kubernetes on a partition of the EX system and run the containers as pods
b) Give researchers a Docker daemon with root access on the compute nodes
c) Run the containers as Slurm jobs with Singularity, which needs no root privileges
d) Move the containerized work to HPE Private Cloud AI, leaving the EX for batch jobs

04. A legal-services firm wants to build its own generative AI inference cluster from HPE ProLiant servers and a software stack it selects itself, rather than buy a turnkey platform. Its architects want a pre-tested configuration with deployment guidance so that the build follows a design HPE has already validated.
Which HPE resource should the architect point them to?
a) An HPE Reference Architecture for the generative AI use case
b) The QuickSpecs of each proposed server and GPU
c) The NVIDIA NGC catalog documentation for the chosen inference containers
d) HPE Private Cloud AI QuickSpecs, used as a build guide for assembling the stack

05. An architect is finalizing a validated design for an HPE Cray EX system. The design assumes facility water of suitable quality, adequate floor loading and three-phase power, all provided by the customer before delivery. The account team wants the presentation to cover only performance.
How should these items appear in the design presentation?
a) Left out, as site preparation is a separate HPE deliverable that has its own documents
b) Listed as HPE responsibilities, since HPE installs and validates the system
c) Raised only if the customer asks, so that the performance case stays clear
d) As stated customer dependencies, each with an owner and a date before delivery

06. A media company is designing storage for an AI inference farm. Dozens of inference servers load models at startup and a vector database serves highly concurrent small random reads, and the company expects performance and capacity to grow at different rates over three years.
The incumbent proposal is an HPE MSA array with hard-drive tiering. Why should the architect recommend HPE Alletra Storage MP instead?
a) Its all-NVMe, disaggregated design lets performance and capacity grow separately
b) Alletra Storage MP contractually guarantees continuous data availability and capacity beyond any MSA configuration
c) Once its cache warms, the MSA hard-drive tiering would serve small random reads and model loads better than flash
d) The MSA cannot present block storage to Linux hosts, so the inference servers could not mount its volumes

07. An automotive company runs crash simulation and AI training on an HPE Cray XD cluster with an HPE Slingshot fabric, and its Lustre storage attaches directly to the same fabric. Large checkpoint writes are crowding out latency-sensitive MPI traffic, and the customer wants each type of traffic to receive a configured share of fabric bandwidth.
Which HPE Slingshot feature should the architect configure?
a) Adaptive routing that moves the checkpoint flows onto less-loaded fabric paths
b) LACP link aggregation across several NIC ports on each compute node
c) Quality-of-service traffic classes that allocate bandwidth by type of traffic
d) VLAN tagging applied by the host operating system on every node

08. A national weather service is installing HPE Cray EX2500 racks in a computer room on an upper floor of its headquarters building. The facilities manager has already confirmed that the chilled-water supply meets HPE’s facility-water requirements.
He asks what else must be settled during the site survey, before the design is finalized. Which facility inputs should the architect confirm now?
a) The facility water supply alone, with power and floor checks left to the installation crew
b) Floor loading for the liquid-cooled racks and 3-phase power feeds for their power shelves
c) Single-phase circuits for each rack and raised-floor airflow for the compute blades
d) The rack UPS runtime and the existing PDUs’ outlet types, with floor loading checked on delivery

09. A genome sequencing center is adding a 64-node HPC and AI cluster. Its 5 PB of raw sequencing data must be read concurrently by hundreds of MPI ranks and GPU training processes through one shared POSIX namespace, at the full bandwidth of the HPE Slingshot fabric.
Procurement will accept only products that HPE currently sells. Which storage should the design include?
a) HPE Alletra Storage MP X10000 object storage, read by every node through its S3 interface
b) HPE Cray ClusterStor E1000 running the Lustre parallel file system on the same fabric
c) A scale-out NFS service on HPE ProLiant file servers backed by HPE Alletra block storage
d) HPE Cray Supercomputing Storage Systems E2000 running the Lustre parallel file system

10. A bank's data science lead is evaluating HPE Private Cloud AI. The bank wants to customize an open LLM on its own anonymized complaint data so that responses use its terminology, and asks to see which NVIDIA AI Enterprise software would perform the customization.
What should the architect demonstrate?
a) The NVIDIA GPU Operator, which tunes model weights while scheduling pods
b) A NIM microservice that retrains itself on each complaint it answers
c) NVIDIA NeMo customizing the model on the bank's data before it is served
d) The NGC catalog, which customizes models as they are downloaded