4A0-107 Nokia Quality of Service Questions and Answers

The primary contributors to latency are distance, scheduling, and congestion.

The weighted round robin scheduling algorithm will result in more delay for high priority traffic than the regular round robin algorithm.

The round robin scheduling algorithm will result in less jitter for low priority traffic than a strict priority scheduler.

Packet loss can be reduced by decreasing the amount of buffer space available to a queue.

Latency can be reduced by increasing the amount of buffer space available to a queue.

IS-IS

RSVP

ICMP

BGP

Access ingress, access egress, network egress

Access ingress, network ingress, network egress

Port ingress, port egress, shared bi-directional

Port ingress, port egress, network egress

Port ingress, network ingress, network egress

EF

H1

L1

H2

L2

Profile mode queues

Priority mode queues

1 queue

3 queues

3 forwarding classes

1 forwarding class, with 2 subclasses

1 forwarding class, with 3 subclasses

2

3

5

6

Profile/priority state of a packet

Forwarding class

Queue depth

HPO setting

Set the CBS to 0.

Set the MBS equal to the CBS.

Set the high-priority-only equal to the MBS.

Set the MBS to 0.

Set the high-priority-only to 0.

Classification rules must specify both an FC and a profile for the packets.

A network policy is applied at a physical port.

Classification is performed with behavior-aggregate statements only.

Match criteria can be chosen from dot1p, DSCP, and EXP fields in the packet.

The default forwarding class for ICMP traffic is BE and cannot be changed.

The SAP has been shut down.

Queue 3 has no PIR or CIR configured, and thus cannot hold any packets.

The SAP-ingress policy has not been applied to the SAP.

dot1p

dst-mac

lsp-exp

dscp

H2

NC

L2

EF

H1

BE

It will be discarded because the high-priority-only threshold is exceeded.

It will be queued but re-marked as " out-of-profile. "

It will be queued in the reserved portion of the buffer pool and remain " in-profile. "

It will be queued in the shared portion of the buffer pool but contend for shared buffer space with packets from other queues.

It will be queued in the shared portion of the buffer pool but may be dropped if the buffer utilization exceeds the start-average value of the low slope.

The default scheduler spends an equal amount of time servicing each queue that has traffic to be sent out.

Queue-type can be configured as best-effort, expedite, or auto-expedite.

A queue will be serviced as expedited if its type is configured as expedited, regardless of the forwarding classes mapped to it.

A queue configured as auto-expedited will be an expedited queue as long as no Best Effort FC is mapped to it.

Once before going through each of the redundant switching fabrics.

To protect against failure of either ingress or egress card.

Once before going through the switching fabric, and once more before going onto the egress port.

Once to decide if the packet is in-profile or out-of-profile, and once more to process the packet according to its profile.

The weight specifies the priority of the queue during scheduling up to its CIR.

The weight specifies the priority of the queue during scheduling above-CIR, up to its PIR.

The weight is used to determine how this queue and other queues or schedulers at the same CIR-level share available CIR bandwidth at that level.

The weight is used to determine how this queue and other queues or schedulers at the same level share available above CIR bandwidth at that level.

Defining the slope of the TCP slow-start mechanism for self generated traffic.

Defining the slope of the scheduling rate per forwarding class when using HQoS.

Defining the high and low slope values when deploying buffer admission control for UDP traffic in the committed buffer pool.

Defining the high and low slope values when deploying buffer admission control for TCP traffic in the shared buffer pool.

Defining the high and low slope values when deploying buffer admission control for TCP traffic in the committed buffer pool.

It will be stored in the reserved buffer space and remain out-of-profile.

It will be stored in the reserved buffer space and become in-profile.

It will contend for shared buffer space with packets from other queues.

It will be discarded because the high-priority-only threshold has been reached.

Change the CIR-level value for the best-effort queue to 8.

Do not change anything; it is already guaranteed to receive its CIR

Change the CIR-level value for the best-effort queue to 1.

Change the CIR-weight value of the AF queue to 0.

It specifies the maximum rate at which a queue can be serviced.

It specifies the maximum depth of a queue beyond which packets will be dropped.

It specifies the portion of a queue that is taken from the reserved buffer pool.

It specifies the threshold beyond which only in-profile packets will be queued.

It specifies the peak bandwidth that can be attained by the incoming traffic.

FC H1, high

FCH1, low

FCAF, high

FCAF, low

1st

8th

9th

17th

Router interfaces

Ingress MDA

Ingress Ethernet network ports

Egress Ethernet network ports

Service SAPs

CIR

CIR level

CIR weight

Rate

Level

Weight

8

16

32

48

DSCP value

802.1p value

IP DF bit

TCP/UDP port numbers

HTML version number

High-1

Assured

Best Effort

Expedited

DiffServ reserves resources dynamically using the RSVP protocol.

DiffServ only requires packets to be classified into forwarding classes at the first network hop.

DiffServ provides granular differentiation and superior fairness because it processes traffic streams individually.

DiffServ is more scalable it treats traffic streams as aggregates.

It will be discarded because the high-priority-only threshold is exceeded.

It will be stored in the queue’s reserved buffer space and become out-of-profile.

It will be stored in the queue’s reserved buffer space and remain in-profile.

It will contend for shared buffer space with packets from other queues.

Microflow reservations using RSVP

DSCP to EXP translation

Traffic classification

Automatic profiling and policy configuration

Buffer memory management

Traffic scheduling

1 multipoint and 8 unicast.

8 queues, regardless of traffic type.

8 unicast and 24 multipoint.

8 unicast and 8 multipoint.

16 queues, distributed as desired.

Network ingress port

Epipe ingress SAP

ES egress SAP

Ingress MDA

VPRN ingress interface

Shared buffer space

The IOM3 adds more memory to both forwarding complexes which increases the total number of queues available.

The IOM3 only contains a single forwarding complex which reduces the number of queues used.

The IOM3 allows multiple forwarding classes to be assigned to a single queue.

The IOM3 allows queues to be dynamically allocated between ingress and egress.

The IOM3 has a larger CPU capable of tracking active hardware queue utilization and adjusting memory allocation accordingly.

Shapeless policing at SAP-ingress only.

Shapeless policing at SAP-ingress and egress.

Biased round robin queuing at ingress only.

Soft shaping at all network interfaces.

Soft shaping at network ingress only.

2

3

5

6

The EXP field in the bottom MPLS header will be marked.

The EXP field in both the top and bottom MPLS headers will be marked.

For Layer 2 services, the top MPLS header will be marked; for Layer 3 services, the bottom MPLS header will be marked.

The EXP field in the top MPLS header will be marked.

Neither the top nor the bottom EXP field will be marked since EXP remarking can only be done on the network ingress.

The policy should be placed in a " shutdown " state before changes are made.

A new network policy should be created and applied to the interface, overwriting the previous policy.

The policy must first be removed from the interface, then modified and re-applied to the interface.

Before the policy can be changed, it must be deleted with the no network qos < policy-id > command.

Before any changes are made to the policy, all network interfaces where the policy is applied have to be shut down.

Scheduler policies cannot be children of an egress port-scheduler.

HQoS provides a static method of allocating scheduling resources to each queue and/or child.

Scheduler policies can be applied to the MDA to schedule network ingress traffic.

Every time a scheduler-policy is assigned to a SAP, a new hardware scheduler is created to service the SAP.

A scheduler applied to multiple SAPs within the same service will result in a single hardware scheduler operating across all SAPs within the service.

TCP port numbers

LSP EXP

DSCP

Source and destination IP addresses

Dot1p priority

1

2

3

4

5

8

All in-profile packets in round-robin, then all out-of-profile packets in biased round-robin.

In-profile EF packets, then in-profile H2 packets, then in-profile AF packets, then out-of-profile EF packets, then out-of-profile H2 and AF packets, and then out-of-profile BE packets.

In-profile EF, H2 and AF packets, then out-of-profile H2, AF and BE packets in biased round-robin.

In-profile EF packets, then in-profile H2 and AF packets, then out-of-profile H2, AF and BE packets in biased round-robin.

In-profile EF and H2 packets, then in-profile AF packets, then out-of-profile H2, AF and BE packets in biased round-robin.

WRED monitors the shared buffer space utilization over a period of time.

WRED disables the TCP slow-start congestion control mechanism.

WRED uses two configurable slopes for each buffer pool: a high slope and a low slope.

WRED uses the shared buffer utilization, instead of the individual queue depth, to get a better picture of the average resource utilization of the shared buffer space.

WRED polices the shared buffer pool and marks all packets above a certain rate as " discard eligible. "

DSCP

PREC

Dot1p

ToS

Q-in-Q

When using profile-mode queues, traffic that is marked as in-profile on ingress and scheduled in the above-CIR loop will still be considered as in-profile at egress.

When an HQoS strategy using an egress port-scheduler has been deployed, Tier 1 receives its scheduling rates based on the amount of egress bandwidth available for each forwarding class.

The greater the difference between PIR and CIR rates in a queue, the less likely a CIR will be provided.

Packets scheduled out of high-priority queues in the default scheduler will have a lower chance of packet loss at the expense of incurring a higher probability of delay.

When using priority-mode queues, traffic that is marked as in-profile on ingress and scheduled in the above-CIR loop will be considered as in-profile at egress.

It maps customer traffic to forwarding classes on ingress.

It can use customer QoS markings to do the mapping.

The default SAP-ingress policy classifies all traffic into FC “be” with low priority.

It serves the classification, queuing, and marking functions of QoS but does not serve scheduling

QoS is not configured on the router.

Packets in forwarding classes " be " and " 12 " are being queued.

Traffic belonging to forwarding class " ef " is not experiencing queuing delays.

Traffic belonging to forwarding class " 11 " will never make use of the shared buffer pool.

Out-of-profile traffic in the shared buffer pool is being dropped.

Signal degradation, attenuation, line loss

Latency, FIFO, WRED

Delay, jitter, packet loss

SNR, queue depth, latency

Attenuation, dispersion, latency

When applied at the FP (Forwarding Path) level for access egress

When applied at an access port for ingress or egress

When applied at the FP (Forwarding Pass) level for access ingress

When applied at the port level for network egress

The slope-policy

The SAP-ingress policy

The network-queue policy

The scheduler-policy

The network-policy

The WRED-policy

The arbitrator-policy

Set the queue’s CBS to 0

Set the queue’s MBS equal to the CBS

Set the queue’s high-priority-only equal to the MBS

Set the queue’s MBS to 0

Set the queue’s high-priority-only to 0

DSCP is part of the Layer 3 header.

DSCP is a 6-bit field.

DSCP specifies eight different priorities.

DSCP specifies four precedence levels.

DSCP is part of the Layer 2 header

DSCP specifies 128 different per-hop behaviors.

The CIR has no influence on the forwarding rate of the policer traffic.

The CIR influences whether a packet is considered high-profile or low-profile.

The policer discards all incoming packets when PIR bucket reaches the MBS level.

PIR bucket forwards packets that arrive with rates above PIR as out-of-profile.

It will be marked as DSCP BE and classified into forwarding class BE.

It will be marked as DSCP NC1 and classified into forwarding class BE.

It will be marked as DSCP NC1 and classified into forwarding class NC.

It will be marked as DSCP NC1 and classified into forwarding class AF.

TRUE

FALSE

It may exceed PIR of the queue.

It may be exceeded by the average servicing rate of the queue.

It varies depending on the traffic characteristics.

It is the rate at which packets are forwarded out of the queue.

Ingress traffic is stored in IOM queues and egress traffic is stored in CPM queues.

Ingress traffic is stored in CPM queues and egress traffic is stored in IOM queues.

It is always stored in CPM queues.

It is always stored in IOM queues.

A queue will be serviced as expedited if all of the forwarding classes mapped to it are expedited, regardless of its configured type.

A queue will be serviced as best-effort if all of the forwarding classes mapped to it are best-effort, regardless of its configured type.

A queue will be serviced as expedited if at least one of the forwarding classes mapped to it is expedited, regardless of its configured type.

A queue will be serviced as expedited if its type is configured as expedited, regardless of the forwarding classes mapped to it.

Packets marked with dscp = af33

Packets marked with dot1p = 4

All packets mapped to queue 2 have the same priority to enter this queue.

There are no packets mapped to queue 2.

prec

lsp-exp

src-ip

do1p

When applied at the port level for access ingress.

When applied at the FP (Forwarding Pass) level for access ingress.

When applied at the FP (Forwarding Pass) level for network ingress.

When applied at the port level for network egress.

Queue 6 within-CIR packets, then queue 5 within-CIR packets, then queue 5 and queue 6 above-CIR packets in round-robin.

Queue 6 within-CIR packets, then queue 5 within-CIR packets, then queue 5 and queue 6 above-CIR packets in biased round-robin.

Queue 6 within-CIR packets, then queue 5 within-CIR packets, then queue 6 above-CIR packets, and finally queue 5 above-CIR packets.

Queue 6 and queue 5 within-CIR packets in biased round-robin, then queue 5 and queue 6 above-CIR packets in round-robin.

The egress port scheduler can have up to 17 children spread across 8 levels.

Multiple egress port schedulers can be assigned to a single port with many HQoS children.

Multiple HQoS policies can be children of a single port scheduler.

Multiple egress port schedulers can be children of a single HQoS scheduler.

8, 6, 3

8, 8, 6

16, 8, 4

8, 6, 4

24, 16, 8

0

3

6

22

H-QoS capabilities must be enabled globally on the router

A scheduler-policy must be configured.

The queues in the SAP-ingress and SAP-egress policies do not require additional configuration attributes to make use of the hierarchical schedulers.

The network queue policy must be configured to make use of the scheduler-policy.

The scheduler-policy must be referenced in the SAP-ingress and SAP-egress policies.

The scheduler-policy must be applied on the service SAP.

It sets both the forwarding class and the packet’s dot1p, lsp-exp, precedence, or dscp field to the same value.

It groups all packets together into a single aggregate stream, and gives all packets the same treatment.

It copies the three most significant bits in the packet’s dscp field into the forwarding class.

It assigns a forwarding class based on the packet’s dot1p, lsp-exp, precedence, or dscp value.