AI Infrastructure Deployment and Data Management

DCAI · 50 questions

• A city public-health GPU cluster carries RDMA training traffic on dedicated AI leaf ports. Which Cisco Data Center congestion-control capability should the fabric team enable on those priorities so drop-sensitive RDMA frames are paused instead of discarded under congestion?
• After an all-reduce stall on the municipal AI fabric, engineers suspect Priority Flow Control contributed to a pause-frame deadlock. What operational step best hardens the civic leaf-spine for future trainer bursts?
• Civic spine queues that carry RoCEv2 GPU traffic are filling during wastewater-model training. How should the municipal fabric team use Explicit Congestion Notification so endpoints can slow down before hard drops?
• County GPU-to-GPU training flows use RoCEv2 across the civic leaf-spine. Which congestion approach correctly pairs fabric signaling with endpoint reaction?
• Overnight storage replication and daytime training RDMA share the same municipal AI leaf uplinks. Which ETS configuration goal should the fabric team enforce?
• A city AI fabric must carry RDMA training, storage I/O, and out-of-band management on the same Nexus leafs. What ETS design best matches that multipurpose civic role?
• Public-health GPU servers need RDMA across the civic Ethernet leaf-spine rather than a separate InfiniBand fabric. Which capability should be enabled on the GPU server NICs facing those leaves?
• City training jobs must span GPU racks in more than one Layer-2 domain inside the municipal data center. Why should the fabric team prefer RoCEv2 over classic RoCE for that design?
• Before declaring RoCEv2 ready for 911 model-training bursts, what should municipal fabric engineers verify end to end?
• A vendor proposes InfiniBand as the only acceptable RDMA answer for the city's Ethernet AI POD. How should the DCAI-aligned municipal team respond for the Cisco Data Center path?
• Clinic inference APIs enter north-south over HTTP while trainers flood east-west RDMA on the same municipal fabric. What QoS outcome should policies deliver?
• Transit digital-twin trainers depend on high-priority RDMA across leaf, spine, and GPU NICs. Which QoS practice correctly protects that class?
• Latency-sensitive clinic triage inference shares uplinks with overnight best-effort dataset backups. What QoS judgment should the municipal team apply?
• Trainers use PFC-enabled priorities on Nexus leaves. How should queuing policy relate to those lossless AI classes?
• Wastewater-model all-reduce traffic concentrates on one spine uplink during peak training. What load-distribution action should the municipal fabric team take?
• A few large RDMA elephant flows from the housing-authority trainers appear stuck on one spine link despite multiple ECMP paths. What load-balancing concern should engineers address?
• The civic hall adds another GPU rack and existing equal-cost paths become oversubscribed. What load-distribution change fits that growth?
• UCS GPU servers use NIC teaming toward dual civic leaves. After enabling teaming, what should fabric engineers validate about load distribution?
• For RoCEv2 public-safety trainers, how should Priority Flow Control and Explicit Congestion Notification be positioned relative to each other on the municipal fabric?
• Operations asks which markings identify AI RDMA versus management on the city fabric. What should the municipal team document and maintain?
• A contractor suggests disabling all congestion control on public-health AI leafs "to go faster" during a training crunch. What is the correct municipal response?
• The city plans citywide RoCEv2 cutover for GPU trainers. What deployment sequencing best reduces risk?
• During peak training windows, storage sync for the courts records lake must still make progress. How should ETS be used?
• Telemetry shows rising PFC pause frames on AI leafs during a many-to-one microburst toward a few GPU receivers. What is the most accurate interpretation?
• Engineers are setting ECN marking thresholds on civic AI spine queues that carry RoCEv2. What threshold intent best supports early congestion signaling?
• A municipal AI fabric team applies QoS policing on the lossless RDMA class used by clinic training. Which risk should they avoid?
• A city AI fabric adds a second GPU rack that skews leaf uplink utilization under ECMP. What load-distribution action best restores balance?
• Before RoCE performance tests on a county public-health fabric, what readiness check should the AI network team confirm fabric-wide?
• A municipal AI fabric carries synchronous training RDMA and asynchronous AI storage replication. How should QoS classes be mapped?
• A transit AI team evaluates ECMP load distribution for a collective that is sensitive to packet reordering. Which judgment is appropriate?
• A city NOC configures ETS on the municipal AI fabric. Which use of strict priority is appropriate?
• A housing-authority GPU cluster enables RoCEv2 for training. What fabric pairing is required rather than treating RDMA as best-effort IP?
• A port-authority AI fabric reviews all-reduce hot spots on leaf-to-spine links. How can wider ECMP help?
• A municipal AI team needs DSCP/CoS markings from UCS GPU servers to survive into the Nexus AI fabric. What alignment is required?
• A city AI fabric enables PFC for lossless RDMA. What operational readiness practice belongs with that design?
• A public-health GPU chassis must inherit a known-good UCS baseline through Intersight. Which construct should the city team apply?
• A city separates 911 inference compute from research training compute. How should UCS domain profiles be used?
• Before pushing a UCS domain profile update to production GPU racks, what should the city Intersight team do?
• During a municipal facility power constraint, how can a UCS power policy help GPU servers?
• Critical clinic inference servers need defined redundant PSU behavior under UCS. Where is that primarily configured?
• Training nodes trip municipal PDU budgets under peak GPU draw. What UCS control should the city adjust?
• A city UCS/Intersight team must present boot and scratch LUNs to GPU servers consistently. Which construct defines that attachment?
• As the municipal AI cluster scales out, how should storage layout stay consistent on new GPU nodes?
• Checkpoint writes are slow on the city training farm. How should the UCS storage policy evolve?
• A municipal UCS design must isolate GPU RDMA traffic from management. What LAN connectivity approach is correct?
• GPU servers in a county AI rack must survive a single leaf failure. How should vNIC policies be aligned?
• RoCEv2 readiness on municipal GPU servers depends on adapter settings. Where should MTU and related adapter attributes be set in UCS?
• A city maps UCS QoS so GPU markings match the Nexus AI fabric design. What must align?
• Clinic GPU RDMA must receive lossless treatment from UCS into the fabric. What UCS QoS choice is required?
• Best-effort municipal traffic threatens to starve AI priority classes on UCS. What should the Intersight QoS policy ensure?