cert
cert-1
cert-2

Pass Cisco 350-601 Exam in First Attempt Guaranteed!

Get 100% Latest Exam Questions, Accurate & Verified Answers to Pass the Actual Exam!
30 Days Free Updates, Instant Download!

cert-5
cert-6
350-601 Exam - Verified By Experts
350-601 Premium Bundle
$29.98

350-601 Premium Bundle

$79.99
$109.97
  • Premium File 613 Questions & Answers. Last update: Nov 10, 2026
  • Training Course 143 Video Lectures
  • Study Guide 1923 Pages
 
$109.97
$79.99
accept 46 downloads in last 7 days
block-screenshots
Purchase Cisco 350-601 Exam Training Products Individually
Cisco 350-601 Premium File
Premium File 613 Q&A
$76.99$69.99
Cisco 350-601 Video Training Course
Training Course 143 Lectures
$21.99 $19.99
Cisco 350-601 PDF Study Guide
Study Guide 1923 Pages
$21.99 $19.99
Cisco 350-601 Practice Test 101

Question 1000

An orchestration tool creates a UCS server profile but omits its organization or policy association in the REST payload. The observed failure is: a syntactically valid request without required policy references can leave a profile incomplete. What is the BEST controlled recovery once this failure is recognized?

  1. The controlled remediation is to suppress all failed API responses and mark each server deploy successful.
  2. The controlled remediation is to use a switch FCID as an Intersight managed-object identifier.
  3. The controlled remediation is to assume every REST POST automatically attaches all necessary LAN and SAN policies.
  4. The controlled remediation is to create or reference the approved policy objects and retry with verified object IDs and deployment state.

Explanation

Correct Answer

D:

Validate against aPI error details, schema properties, object MOIDs, policy references and deployed server profile readback. Intersight automation works with typed managed objects and explicit relationships that must resolve correctly.

Incorrect Answers

A:

Suppress all failed API responses and mark each server deploy successful. This would not reliably restore service after the stated fault. A syntactically valid request without required policy references can leave a profile incomplete. The controlled recovery is to create or reference the approved policy objects and retry with verified object IDs and deployment state.

B:

Use a switch FCID as an Intersight managed-object identifier. This would not reliably restore service after the stated fault. A syntactically valid request without required policy references can leave a profile incomplete. The controlled recovery is to create or reference the approved policy objects and retry with verified object IDs and deployment state.

C:

Assume every REST POST automatically attaches all necessary LAN and SAN policies. This would not reliably restore service after the stated fault. A syntactically valid request without required policy references can leave a profile incomplete. The controlled recovery is to create or reference the approved policy objects and retry with verified object IDs and deployment state.

Question 1001

An RDMA-capable AI training network drops bursts of loss-sensitive traffic because the intended priority flow control class is absent. What is the BEST supported implementation or corrective action?

  1. Design and validate priority flow control on the designated lossless class with appropriate end-to-end QoS mapping.
  2. Enable link-level pause for every traffic class without congestion analysis.
  3. Route RoCE frames through a Fibre Channel zone as if it were native FC.
  4. Assume normal TCP retransmissions guarantee RDMA latency objectives.

Explanation

Correct Answer

A:

This directly addresses lossless Ethernet PFC for AI workloads because roCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking. Confirm the change against pFC-enabled priorities, queue pause and drop counters, RoCE class mapping and congestion measurements.

Incorrect Answers

B:

Treating all traffic as lossless or none of the RDMA class as loss-sensitive can degrade AI training performance. The relevant correction is to design and validate priority flow control on the designated lossless class with appropriate end-to-end QoS mapping.

C:

Route RoCE frames through a Fibre Channel zone as if it were native FC. The proposed change targets the wrong mechanism or lacks essential validation. RoCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking. Relevant evidence is pFC-enabled priorities, queue pause and drop counters, RoCE class mapping and congestion measurements.

D:

Treating all traffic as lossless or none of the RDMA class as loss-sensitive can degrade AI training performance. The appropriate action remains to design and validate priority flow control on the designated lossless class with appropriate end-to-end QoS mapping.

Question 1002

An RDMA-capable AI training network drops bursts of loss-sensitive traffic because the intended priority flow control class is absent. Which evidence BEST verifies the intended technical behavior?

  1. A saved script file that has never been executed or tested.
  2. PFC-enabled priorities, queue pause and drop counters, RoCE class mapping and congestion measurements.
  3. A dashboard snapshot detached from actual target configuration and logs.
  4. A successful SSH session without script output, API responses or resulting state.

Explanation

Correct Answer

B:

These observations test the actual lossless Ethernet PFC for AI workloads mechanism rather than a superficial management status. RoCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking.

Incorrect Answers

A:

This information cannot independently verify lossless Ethernet PFC for AI workloads. The decisive records instead include pFC-enabled priorities, queue pause and drop counters, RoCE class mapping and congestion measurements.

C:

This information cannot independently verify lossless Ethernet PFC for AI workloads. RoCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking.

D:

This information cannot independently verify lossless Ethernet PFC for AI workloads. A successful SSH session without script output, API responses or resulting state.

Question 1003

An RDMA-capable AI training network drops bursts of loss-sensitive traffic because the intended priority flow control class is absent. Which principle BEST explains why the recommended solution is appropriate?

  1. Cloud or on-box automation removes the need to validate inputs and outcomes.
  2. A script saved on a controller always runs successfully on every target.
  3. RoCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking.
  4. An HTTP 200 connection proves the intended device state was changed correctly.

Explanation

Correct Answer

C:

Validation depends on pFC-enabled priorities, queue pause and drop counters, RoCE class mapping and congestion measurements. This explains why the most appropriate response is to design and validate priority flow control on the designated lossless class with appropriate end-to-end QoS mapping.

Incorrect Answers

A:

Cloud or on-box automation removes the need to validate inputs and outcomes. This is an incorrect generalization for lossless Ethernet PFC for AI workloads. The applicable technical behavior is that roCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking. The operator should verify pFC-enabled priorities, queue pause and drop counters, RoCE class mapping and congestion measurements.

B:

A script saved on a controller always runs successfully on every target. This is an incorrect generalization for lossless Ethernet PFC for AI workloads. The applicable technical behavior is that roCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking. The operator should verify pFC-enabled priorities, queue pause and drop counters, RoCE class mapping and congestion measurements.

D:

An HTTP 200 connection proves the intended device state was changed correctly. This is an incorrect generalization for lossless Ethernet PFC for AI workloads. The applicable technical behavior is that roCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking. The operator should verify pFC-enabled priorities, queue pause and drop counters, RoCE class mapping and congestion measurements.

Question 1004

An RDMA-capable AI training network drops bursts of loss-sensitive traffic because the intended priority flow control class is absent. Assume an engineer instead chooses to enable link-level pause for every traffic class without congestion analysis. What is the MOST significant failure introduced by that shortcut?

  1. Keeping API error responses is the underlying cause of failed automation.
  2. Reviewing a Terraform plan makes correct configuration objects disappear.
  3. Using a dry-run test inherently disables every network management API.
  4. Treating all traffic as lossless or none of the RDMA class as loss-sensitive can degrade AI training performance.

Explanation

Correct Answer

D:

RoCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking. That is the direct consequence of choosing the misleading shortcut.

Incorrect Answers

A:

This does not explain the actual lossless Ethernet PFC for AI workloads defect after the shortcut. The meaningful failure is that treating all traffic as lossless or none of the RDMA class as loss-sensitive can degrade AI training performance.

B:

This does not explain the actual lossless Ethernet PFC for AI workloads defect after the shortcut. The relevant evidence is pFC-enabled priorities, queue pause and drop counters, RoCE class mapping and congestion measurements.

C:

This does not explain the actual lossless Ethernet PFC for AI workloads defect after the shortcut. In the given case, an RDMA-capable AI training network drops bursts of loss-sensitive traffic because the intended priority flow control class is absent.

Question 1005

An RDMA-capable AI training network drops bursts of loss-sensitive traffic because the intended priority flow control class is absent. The observed failure is: treating all traffic as lossless or none of the RDMA class as loss-sensitive can degrade AI training performance. What is the BEST controlled recovery once this failure is recognized?

  1. Correct PFC scope and related QoS, then stress-test for loss and pause storms.
  2. Route RoCE frames through a Fibre Channel zone as if it were native FC.
  3. Assume normal TCP retransmissions guarantee RDMA latency objectives.
  4. Enable link-level pause for every traffic class without congestion analysis.

Explanation

Correct Answer

A:

RoCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking. This addresses the observable fault while preserving a check of the expected result.

Incorrect Answers

B:

Route RoCE frames through a Fibre Channel zone as if it were native FC. This would not reliably restore service after the stated fault. Treating all traffic as lossless or none of the RDMA class as loss-sensitive can degrade AI training performance. The controlled recovery is to correct PFC scope and related QoS, then stress-test for loss and pause storms.

C:

Assume normal TCP retransmissions guarantee RDMA latency objectives. This would not reliably restore service after the stated fault. Treating all traffic as lossless or none of the RDMA class as loss-sensitive can degrade AI training performance. The controlled recovery is to correct PFC scope and related QoS, then stress-test for loss and pause storms.

D:

That option would enable link-level pause for every traffic class without congestion analysis, but it addresses a different concern. This addresses the observable fault while preserving a check of the expected result.

Question 1006

A GPU cluster uses RoCEv2 transport, but congestion management applies best-effort queues to its loss-sensitive traffic class. Which configuration decision MOST directly resolves the issue?

  1. On this affected device, the engineer should route RoCE frames through a Fibre Channel zone as if it were native FC.
  2. On this affected device, the engineer should design and validate priority flow control on the designated lossless class with appropriate end-to-end QoS mapping.
  3. On this affected device, the engineer should assume normal TCP retransmissions guarantee RDMA latency objectives.
  4. On this affected device, the engineer should enable link-level pause for every traffic class without congestion analysis.

Explanation

Correct Answer

B:

This directly addresses lossless Ethernet PFC for AI workloads because roCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking.

Incorrect Answers

A:

Route RoCE frames through a Fibre Channel zone as if it were native FC. The proposed change targets the wrong mechanism or lacks essential validation. RoCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking. Relevant evidence is pFC-enabled priorities, queue pause and drop counters, RoCE class mapping and congestion measurements.

C:

Treating all traffic as lossless or none of the RDMA class as loss-sensitive can degrade AI training performance.

D:

Treating all traffic as lossless or none of the RDMA class as loss-sensitive can degrade AI training performance. Enable link-level pause for every traffic class without congestion analysis.

Question 1007

A GPU cluster uses RoCEv2 transport, but congestion management applies best-effort queues to its loss-sensitive traffic class. What should the engineer inspect to confirm the actual state?

  1. For this deployment, examine an old CI job name without run records or configuration diffs.
  2. For this deployment, examine a saved script file that has never been executed or tested.
  3. For this deployment, examine pFC-enabled priorities, queue pause and drop counters, RoCE class mapping and congestion measurements.
  4. For this deployment, examine a dashboard snapshot detached from actual target configuration and logs.

Explanation

Correct Answer

C:

These observations test the actual lossless Ethernet PFC for AI workloads mechanism rather than a superficial management status. They should be checked after design and validate priority flow control on the designated lossless class with appropriate end-to-end QoS mapping.

Incorrect Answers

A:

This information cannot independently verify lossless Ethernet PFC for AI workloads. An old CI job name without run records or configuration diffs.

B:

This information cannot independently verify lossless Ethernet PFC for AI workloads. In the given case, a GPU cluster uses RoCEv2 transport, but congestion management applies best-effort queues to its loss-sensitive traffic class.

D:

This information cannot independently verify lossless Ethernet PFC for AI workloads. A dashboard snapshot detached from actual target configuration and logs.

Question 1008

A GPU cluster uses RoCEv2 transport, but congestion management applies best-effort queues to its loss-sensitive traffic class. What technical behavior underlies the correct decision?

  1. For this condition, the applicable principle is that an HTTP 200 connection proves the intended device state was changed correctly.
  2. For this condition, the applicable principle is that cloud or on-box automation removes the need to validate inputs and outcomes.
  3. For this condition, the applicable principle is that the existence of a state file proves all remote resources match desired state.
  4. For this condition, the applicable principle is that roCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking.

Explanation

Correct Answer

D:

Validation depends on pFC-enabled priorities, queue pause and drop counters, RoCE class mapping and congestion measurements.

Incorrect Answers

A:

An HTTP 200 connection proves the intended device state was changed correctly. This is an incorrect generalization for lossless Ethernet PFC for AI workloads. The applicable technical behavior is that roCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking. The operator should verify pFC-enabled priorities, queue pause and drop counters, RoCE class mapping and congestion measurements.

B:

Cloud or on-box automation removes the need to validate inputs and outcomes. This is an incorrect generalization for lossless Ethernet PFC for AI workloads. The applicable technical behavior is that roCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking. The operator should verify pFC-enabled priorities, queue pause and drop counters, RoCE class mapping and congestion measurements.

C:

The existence of a state file proves all remote resources match desired state. This is an incorrect generalization for lossless Ethernet PFC for AI workloads. The applicable technical behavior is that roCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking. The operator should verify pFC-enabled priorities, queue pause and drop counters, RoCE class mapping and congestion measurements.

Question 1009

A GPU cluster uses RoCEv2 transport, but congestion management applies best-effort queues to its loss-sensitive traffic class. Assume an engineer instead chooses to enable link-level pause for every traffic class without congestion analysis. Which resulting risk or fault should the reviewer identify?

  1. The resulting defect is that treating all traffic as lossless or none of the RDMA class as loss-sensitive can degrade AI training performance.
  2. The resulting defect is that using a dry-run test inherently disables every network management API.
  3. The resulting defect is that storing execution logs causes the original script to use invalid parameters.
  4. The resulting defect is that reviewing a Terraform plan makes correct configuration objects disappear.

Explanation

Correct Answer

A:

RoCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking. The proper corrective action is to design and validate priority flow control on the designated lossless class with appropriate end-to-end QoS mapping.

Incorrect Answers

B:

This does not explain the actual lossless Ethernet PFC for AI workloads defect after the shortcut. In the given case, a GPU cluster uses RoCEv2 transport, but congestion management applies best-effort queues to its loss-sensitive traffic class.

C:

This does not explain the actual lossless Ethernet PFC for AI workloads defect after the shortcut. Storing execution logs causes the original script to use invalid parameters.

D:

This does not explain the actual lossless Ethernet PFC for AI workloads defect after the shortcut. Reviewing a Terraform plan makes correct configuration objects disappear.

Question 1010

A GPU cluster uses RoCEv2 transport, but congestion management applies best-effort queues to its loss-sensitive traffic class. The observed failure is: treating all traffic as lossless or none of the RDMA class as loss-sensitive can degrade AI training performance. Which remediation most directly restores the intended behavior?

  1. The controlled remediation is to assume normal TCP retransmissions guarantee RDMA latency objectives.
  2. The controlled remediation is to correct PFC scope and related QoS, then stress-test for loss and pause storms.
  3. The controlled remediation is to enable link-level pause for every traffic class without congestion analysis.
  4. The controlled remediation is to route RoCE frames through a Fibre Channel zone as if it were native FC.

Explanation

Correct Answer

B:

RoCE-based data center traffic may use PFC for selected priorities, while careless pause design can cause head-of-line blocking. Validate against pFC-enabled priorities, queue pause and drop counters, RoCE class mapping and congestion measurements.

Incorrect Answers

A:

Assume normal TCP retransmissions guarantee RDMA latency objectives. This would not reliably restore service after the stated fault. Treating all traffic as lossless or none of the RDMA class as loss-sensitive can degrade AI training performance. The controlled recovery is to correct PFC scope and related QoS, then stress-test for loss and pause storms.

C:

Enable link-level pause for every traffic class without congestion analysis. This would not reliably restore service after the stated fault. Treating all traffic as lossless or none of the RDMA class as loss-sensitive can degrade AI training performance. The controlled recovery is to correct PFC scope and related QoS, then stress-test for loss and pause storms.

D:

Route RoCE frames through a Fibre Channel zone as if it were native FC. This would not reliably restore service after the stated fault. Treating all traffic as lossless or none of the RDMA class as loss-sensitive can degrade AI training performance. The controlled recovery is to correct PFC scope and related QoS, then stress-test for loss and pause storms.

Implementing and Operating Cisco Data Center Core Technologies (DCCOR) (350-601) Practice Tests