CCNA QoS Practice Questions (200-301)

How to work these questions

1. Identify the traffic type in the question: voice, video, interactive data, bulk data or scavenger traffic.
2. Decide which characteristic suffers: not enough bandwidth, too much delay, too much jitter, or packet loss. Voice and video suffer from delay and jitter; bulk TCP mostly from loss and bandwidth.
3. Pick the tool that fixes that characteristic: LLQ priority queue for delay and jitter, CBWFQ for a minimum bandwidth guarantee, WRED to avoid tail drop and TCP synchronization, shaping to buffer bursts toward a slower link, policing to enforce a hard rate by dropping or re-marking.
4. Check where the marking lives: DSCP (6 bits in the IP header) survives routing end to end, CoS (3 bits in the 802.1Q tag) exists only on trunk links.
5. Check placement: mark and trust as close to the source as possible (the IP phone), shape outbound toward the slower link, police where a contract must be enforced.

Practice questions

Pick an answer to see right away whether it is correct and why every option is right or wrong.

1. Question 1 of 10Concept

   You are writing a QoS design for a campus that is deploying IP phones. Which set of targets is commonly used for voice traffic?

   1. Up to 400 ms one-way delay, 100 ms jitter, 5% loss
   2. Up to 150 ms one-way delay, 30 ms jitter, 1% loss
   3. Up to 150 ms round-trip delay, 300 ms jitter, 10% loss
   4. Up to 50 ms one-way delay, no jitter limit, 3% loss
   Show the answer

   Answer: B

   Voice needs low delay (about 150 ms one-way), low jitter (about 30 ms) and low loss (under 1%). Bulk data tolerates delay and jitter; voice and video do not. Use these numbers to decide which tool a question is really asking for.

   • A. Callers notice 400 ms of one-way delay and talk over each other, 100 ms of jitter overflows typical jitter buffers, and 5% loss makes speech choppy.
   • B. These are the usual voice targets: about 150 ms one-way delay, about 30 ms jitter and under 1% packet loss.
   • C. The 150 ms figure is one-way, not round-trip, and 300 ms of jitter or 10% loss would make a call unusable.
   • D. Voice is very sensitive to jitter, so there is always a jitter target, and 3% loss is well above what voice tolerates.

2. Question 2 of 10Calculation

   A switch is configured to mark voice traffic with DSCP EF (decimal 46). Which 6-bit value is written into the DSCP field of the IP header?

   1. 101000
   2. 100110
   3. 101100
   4. 101110
   Show the answer

   Answer: D

   DSCP is 6 bits with place values 32, 16, 8, 4, 2, 1. EF 46 = 101110, CS values are the precedence followed by 000 (CS5 = 101000 = 40), and AFxy = 8x + 2y. Convert by subtracting place values from the left.

   • A. 101000 is 32 + 8 = 40, which is CS5, not EF.
   • B. 100110 is 32 + 4 + 2 = 38, which is AF43.
   • C. 101100 is 32 + 8 + 4 = 44, which is not a standard per-hop behavior used for voice.
   • D. 46 = 32 + 8 + 4 + 2, so the bits are 101110. The first three bits, 101, are IP precedence 5, which is why EF lines up with the old precedence value for voice.

3. Question 3 of 10Exhibit

   Refer to the exhibit. How does the router treat packets marked DSCP EF that leave GigabitEthernet0/0/1 while the interface is congested?

   class-map match-any VOICE
    match dscp ef
   class-map match-any ORDERS
    match dscp af31
   !
   policy-map WAN-EDGE
    class VOICE
     priority 512
    class ORDERS
     bandwidth 2048
    class class-default
     fair-queue
   !
   interface GigabitEthernet0/0/1
    service-policy output WAN-EDGE

   1. They use a strict priority queue served first, policed to 512 kbps
   2. They get a 512 kbps minimum but wait behind the ORDERS queue
   3. They are re-marked to DSCP 0 before entering the priority queue
   4. They are shaped to 512 kbps and any excess is buffered to send later
   Show the answer

   Answer: A

   In a policy-map, priority creates the LLQ strict priority queue (for voice), bandwidth gives a CBWFQ minimum guarantee, and class-default catches everything else. The priority queue is policed to its rate under congestion. service-policy output applies the policy outbound.

   • A. The priority command makes this LLQ: EF packets go into a strict priority queue that is always served first, and during congestion that class is limited to 512 kbps so it cannot starve the others.
   • B. A minimum guarantee without priority is what the bandwidth command gives ORDERS. The priority class never waits behind a CBWFQ class.
   • C. Nothing in the policy contains a set command, so markings are not changed.
   • D. The implicit limit on a priority class is a policer: traffic above 512 kbps during congestion is dropped, not buffered as shaping would do.

4. Question 4 of 10Scenario

   An IP phone marks voice frames with CoS 5 and DSCP EF. The access switch sends them over an 802.1Q trunk to a distribution switch, which routes them over a routed point-to-point link to the WAN router. Which marking can the WAN router use to classify the voice packets?

   1. Only CoS 5, as Layer 2 markings are copied into each new frame
   2. Both CoS 5 and DSCP EF, as both markings are kept end to end
   3. Only DSCP EF, as CoS is in the 802.1Q tag lost at the routed hop
   4. Neither, as routers reset all QoS markings to zero when they route
   Show the answer

   Answer: C

   CoS (3 bits, 802.1Q tag) lives only on trunks. DSCP (6 bits, IP header) survives routing end to end. Mark DSCP close to the source and classify on DSCP anywhere past the first routed hop.

   • A. CoS is not copied anywhere by default. When the distribution switch routes the packet, it builds a new frame, and an untagged routed link has no field to carry CoS.
   • B. Only the IP-header marking survives routing. CoS exists only while the frame carries an 802.1Q tag.
   • C. DSCP is in the IP header and travels with the packet across routed hops. CoS is 3 bits inside the 802.1Q tag, so it disappears when the tag is removed at the routed hop.
   • D. Routers do not reset DSCP when forwarding unless a policy tells them to. That is why marking once near the source works.

5. Question 5 of 10Concept

   Which two statements about traffic shaping are true? (Choose two.)

   1. It buffers traffic above the configured rate and sends it later
   2. It drops or re-marks excess traffic at once, without buffering
   3. It is usually applied outbound toward a slower WAN or contracted rate
   4. It is supported equally inbound and outbound on Cisco routers
   5. It lowers delay for voice because shaped packets never wait in a queue
   Show the answer

   Answer: A and C

   Shaping = buffer and delay excess, outbound only, used toward a slower link. Policing = drop or re-mark excess, no buffering, inbound or outbound, used to enforce a contract.

   • A. Shaping holds excess packets in a buffer and releases them at the configured rate, smoothing bursts.
   • B. That is policing. A policer does not buffer; it drops or re-marks traffic that exceeds the rate.
   • C. Shaping is typically configured on the egress interface toward a slower link, for example to match a provider's contracted rate so the provider's policer does not drop traffic.
   • D. Shaping is an outbound tool because it needs an output queue to hold packets. Policing can be applied inbound or outbound.
   • E. Buffering adds delay and can add jitter, which is the cost of shaping. Voice is protected by a priority queue, not by shaping itself.

6. Question 6 of 10Scenario

   Each desk has a Cisco IP phone connected to an access switch port, with the user's PC connected to the phone's PC port. You want the network to trust voice markings but not let PC applications mark their own traffic as high priority. Where should the trust boundary be?

   1. At the user's PC, so each application can set its own DSCP
   2. At the IP phone, so the switch trusts it and it re-marks PC traffic
   3. At the WAN router, so markings are checked only before leaving
   4. At the distribution switch, so access ports pass markings unchecked
   Show the answer

   Answer: B

   The trust boundary is where the network starts believing markings. Put it as close to the source as possible: at the IP phone for voice, with PC traffic re-marked. Traffic from untrusted devices is re-marked at the edge.

   • A. A trust boundary at the PC means any application or user can mark traffic EF and get priority treatment, which is exactly what you want to prevent.
   • B. The access switch trusts markings from the detected IP phone, and the phone re-marks traffic from the attached PC (typically to 0), so the trust boundary sits at the phone, as close to the source as possible.
   • C. Checking markings only at the WAN edge lets falsely marked traffic get priority on every campus link along the way.
   • D. Moving trust further from the source lets PC markings cross the access layer unchecked and misuse the priority queues there.

7. Question 7 of 10Exhibit

   Refer to the exhibit. A policer on the branch router marks traffic from an order-entry application according to its rate. All three markings share one queue that uses DSCP-based WRED. When the queue starts to fill, which packets are most likely to be dropped first?

   Rate condition      Marking   DSCP (decimal)   DSCP (binary)
   ------------------  --------  ---------------  -------------
   Conforming          AF31      26               011010
   Exceeding           AF32      28               011100
   Violating           AF33      30               011110

   1. AF31 packets, as the lowest DSCP value has the lowest priority
   2. All three equally, as WRED drops at random regardless of marking
   3. AF32 packets, as the middle value is chosen to balance the loss
   4. AF33 packets, as a y digit of 3 is the highest drop precedence
   Show the answer

   Answer: D

   AFxy: x = class (1-4), y = drop precedence (1-3), decimal value = 8x + 2y. Inside one class, higher y is dropped first by WRED. EF is 46, CS values are multiples of 8, default is 0.

   • A. Within one AF class, the lower y digit has the lowest drop precedence, so AF31 is protected the most.
   • B. DSCP-based WRED uses a different drop profile per marking, starting to drop high drop-precedence packets at a lower queue depth.
   • C. AF32 is dropped before AF31 but after AF33, so it is not the first to go.
   • D. In AFxy, x is the class (queue) and y is the drop precedence (1 low, 3 high). AF33 = 8 x 3 + 2 x 3 = 30, and WRED drops it earliest as the queue grows.

8. Question 8 of 10Concept

   Which queuing approach gives each traffic class a configured minimum share of bandwidth during congestion, but does not give any class strict priority?

   1. CBWFQ with a bandwidth command per class
   2. LLQ with a priority command for each class
   3. FIFO with one queue on the interface
   4. WRED with drop thresholds per class
   Show the answer

   Answer: A

   FIFO = one queue, no differentiation. CBWFQ = per-class queues with bandwidth guarantees. LLQ = CBWFQ plus a strict priority queue for voice and video. WRED manages drops inside a queue.

   • A. CBWFQ creates a queue per class and the bandwidth command guarantees each one a minimum during congestion, with no queue served ahead of the others unconditionally.
   • B. The priority command creates a strict priority queue that is always served first, which is the opposite of what the question describes.
   • C. FIFO has a single queue, so there are no classes and no bandwidth guarantees: packets leave in arrival order.
   • D. WRED is congestion avoidance, deciding which packets to drop early, not a scheduler that allocates bandwidth.

9. Question 9 of 10Scenario

   On a congested WAN link, the output queue regularly fills and then drops every arriving packet. Users report that file transfers all slow down at the same moment, recover together and then slow down again. Which tool addresses this behavior?

   1. LLQ, by placing the TCP flows in the strict priority queue
   2. Shaping on the LAN interface to slow down every host directly
   3. WRED, by dropping some packets early before the queue is full
   4. CoS marking on access ports so the frames carry a priority
   Show the answer

   Answer: C

   Tail drop hits all TCP flows at once, so they back off and ramp up together (global synchronization). WRED randomly drops early, more aggressively for higher drop precedence, keeping queues shorter and throughput steadier.

   • A. Putting bulk TCP in the priority queue would starve other traffic and does nothing about synchronized backoff.
   • B. Shaping on the LAN side does not prevent the WAN queue from tail-dropping, and it buffers rather than signalling individual flows.
   • C. This is tail drop causing TCP global synchronization. WRED drops a few packets from some flows early, so only those flows back off and the queue never reaches the point of dropping everything.
   • D. Marking only labels traffic. Without a queuing or drop policy that acts on it, the tail drop continues, and CoS does not even survive to the routed WAN link.

10. Question 10 of 10Scenario

    A branch router connects to the provider with a 1 Gbps Ethernet handoff, but the contract is 100 Mbps and the provider polices anything above that. Bursts from the branch are being dropped by the provider. What should you configure on the branch router's WAN interface?

    1. Police outbound at 100 Mbps so excess packets are dropped locally first
    2. Shape outbound to 100 Mbps so bursts are buffered and queued locally
    3. Use FIFO queuing so packets keep their order on the 1 Gbps link
    4. Re-mark all traffic to DSCP 0 so the provider policer ignores it
    Show the answer

    Answer: B

    When the physical link is faster than the contracted rate, shape outbound to the contracted rate toward the slower link. Shaping buffers and smooths bursts; policing drops them. A shaper also creates the congestion point where LLQ and CBWFQ can work.

    • A. A local policer would drop the same excess the provider is dropping, just one hop earlier, so the bursts are still lost.
    • B. Shaping to the contracted rate buffers the bursts and sends them no faster than 100 Mbps, so the provider's policer stops dropping, and your own queuing policy decides what waits.
    • C. Without shaping the interface never congests at 1 Gbps, so no queuing method changes anything; the drops happen at the provider.
    • D. A provider policer enforces the rate on all traffic. Re-marking to 0 only removes your priority information.

Common mistakes

• Confusing policing and shaping: policing drops or re-marks excess traffic immediately, shaping buffers it and sends it later.
• Expecting CoS to survive a routed hop or an access port, when it only exists inside the 802.1Q tag on trunks.
• Thinking a higher AF drop-precedence digit means better treatment, when AF33 is dropped before AF31 in the same class.
• Putting the trust boundary at the user's PC, which lets any application mark its own traffic as EF.
• Assuming the LLQ priority queue is unlimited, when the priority command also polices that class to its configured rate during congestion.

Questions people ask

More CCNA topics

• CCNA Subnetting Practice Questions
• CCNA OSPF Practice Questions: Single-Area OSPFv2
• CCNA VLANs and Trunking: Practice Questions
• CCNA ACL Practice Questions
• CCNA NAT, PAT and Port Forwarding Practice Questions
• CCNA Spanning Tree Protocol: Practice Questions
• CCNA IPv6 Practice Questions: Addressing, EUI-64, SLAAC and NDP
• CCNA EtherChannel: Practice Questions
• CCNA First-Hop Redundancy Practice Questions: HSRP, VRRP and GLBP
• CCNA Wireless Fundamentals Practice Questions
• CCNA Static Routing Practice Questions: ip route and the Routing Table
• CCNA IP Services Practice Questions
• CCNA Layer 2 Security Practice Questions
• CCNA Security Fundamentals Practice Questions (200-301)
• CCNA Automation and Programmability Practice Questions
• CCNA practice test: all domains