Cisco UCS vNIC Placement vs RSS for CPU Offload
An engineer is experiencing performance issues on a Cisco UCS blade server. The B-Series blade server contains four CPUs, most of which are idle. The engineer notices that the CPU is suffering from too many requests sent by the NIC. Additionally, the number of queues appears to be insufficient and only a single CPU is processing the network traffic. Which policy must be used to alleviate these issues?
Community Votes
100% of anonymous learners picked answer C. Votes are pick records left by other test-takers — they are not the verified answer.
Community Insight
The core issue is single-CPU processing of network traffic, which is resolved by enabling Receive Side Scaling (RSS) within the Ethernet adapter policy.
This question addresses how to distribute network traffic across multiple CPUs on a Cisco UCS B-Series blade server using the Ethernet adapter policy.
Many candidates incorrectly select 'vNIC placement' (D), confusing physical pinning with software-based load balancing across available cores.
Community Discussion (4 comments)
Comments & Corrections
No comments yet — spotted an error or have a note? Share it below.
Expert Analysis
Why the Answer Is Correct
The correct answer is C (Ethernet adapter). The scenario describes a classic case where network interrupts are not being distributed across the available CPU cores, causing a bottleneck on a single processor despite idle resources. This is solved by enabling Receive Side Scaling (RSS) in the Ethernet adapter policy. As noted by user Fcpoultry, "RSS distributes network receive processing across multiple CPUs... enabled—Network receive processing is shared across processors whenever possible." By configuring the Ethernet adapter policy, the engineer enables RSS, allowing the NIC and OS to hash incoming packets to different queues mapped to different CPUs.Why the Other Options Are Wrong
Option D (vNIC placement) is a common distractor. While vNIC placement controls which physical PCIe slots or NUMA nodes a vNIC is pinned to, it does not inherently solve the problem of a single CPU handling all interrupts if RSS is not enabled. Option A (LAN connectivity) is too generic and refers to the general path rather than a specific configuration policy. Option B (dynamic vNIC connection) relates to how vNICs are assigned to service profiles dynamically, which has no bearing on CPU interrupt distribution.Community Comment Notes
The community consensus strongly favors C. User zeppie confirms that RSS is enabled on the ethernet adapter policy. User CoAsT_x emphasizes that this is a non-negotiable setting for such issues. One dissenting vote suggests D, but the technical reality of interrupt affinity points to the adapter's internal scaling features.Official Reference
Exam Strategy
When seeing "single CPU processing" or "interrupt storm" on multi-core systems, look for RSS (Receive Side Scaling) or IOMMU settings first. In UCS specifically, these are configured in the Adapter Policy, not just the vNIC definition.
Frequently Asked Questions
Why isn't vNIC placement the solution?
VNIC placement pins vNICs to specific hardware paths but doesn't split interrupt handling across cores unless RSS is also enabled.
Does RSS work on all Cisco UCS adapters?
Most modern VIC adapters support RSS, but it must be explicitly enabled in the Ethernet adapter policy within UCS Manager.