Add21 LFA and QoS
Add21 LFA and QoS
1
What is micro-loop assume link A-B fail, B
cost of links is equal
A
× C
route convergence
process once run SPF, best path from A
to B is A-E-D-C-B E D
A complete SPF(125ms)
2
LFA basic concept
Primary path
S figure out repair path S-N-D = Loop-Free
E
Primary NH Alternate (LFA) :
S Alternate NH D n Repair path not pass through link S-E ;
N n Next hop router N to D's path not pass
Repair path through back to S
S Source (local) router The router where all calculations are done
N Neighboring router The neighbor router which is the alternate next hop
router under investigation
E Another neighboring router The primary next hop router
3
Inequality 1,2,3 from RFC5286 Primary path
Loop Free Alternate (link protection)
E
Inequality 1: D(N,D) < D(N,S) + D(S,D) Primary NH
“Path is loop-free because N’s best path is not through local router.”
Traffic sent to backup next hop is not sent back to S.
S Alternate NH D
N
Down Stream Path Repair path
Inequality 2: D(N,D) < D(S,D)
“Neighbor router is closer to the destination than local router.”
Loop-free is guaranteed even with multiple failures (if all repair-paths are downstream path).
Node Protection
Inequality 3: D(N,D) < D(N,E) + D(E,D)
“N's path to D must not go through E.”
“The distance from the node N to the prefix via the primary next-hop is strictly greater than the optimum
distance from the node N to the prefix.“ © ZTE All rights reserved
4
Example 1
10 E 5 10 E 5 10 E 5
S D S D S D
12 N 11 12 N 20 12 N 27
5
Example 2:downstream path and micro-loop
n N and S are mutually LFA nodes, meeting the link protection
Inequality 1: 14<5+15 conditions but not the node protection conditions. When node E
fails, S will send the message to N. N will re-transmit the message to
Inequality 2: 14<15
S. The micro-loop of the message is generated until the whole
Inequality 3: 14<14 network is converged and synchronized.
n If downstream conditions are introduced at the same time, N is S's
5 E 10 LFA meeting the downstream conditions, but S is not N. So, after S
S 4 D forwards the packet to N, N will directly discard the packet and
avoid micro-loop.
5 N
n Select a LFA that meets the downstream condition to avoid micro-
Node failure with loop after the packet arrives at the LFA. However, as the downstream
only link protection condition is stricter, the fault coverage rate is reduced accordingly.
6
downstream path example
10 R4
source R2 12
4
7 11 R5 destination
R3
5
R1
source
considering inequality1,
• for R1,15<7+16,So R2 can be R1's LFA(red)
• for R2,16<7+15,So R1 can be R2's LFA(blue)
© ZTE All rights reserved
7
downstream path example
10 R4
source R2 12
4
7 11 R5 destination
R3
5
R1
source
8
LFA vs MPLS TE
LFA MPLS TE
Repair path Least cost Constraints based with bandwidth
guarantee and path control
Link protection Yes Yes
Minimal Significant
Provision
Configuration
Effective with No dependency. Always works.
Network
mesh
topology
coverage
9
Chapter QoS
10
Three Priority of the Switch
For a ZTE switch, the default priority, local priority, and drop priority are frequently
used in the port-based QoS configuration. Take ZXR10 8900E for example to
explain the three concepts. To explain the three concepts, first understand the basic
concepts of related terms.
• CoS value: 802.1p value of the received packet, also called packet priority
(for L2 Ethernet frame).
• Cos queue: Each physical port of ZXR10 8900E supports 8 output queues
(queue 0-7), called CoS queue, which is meaningful when processing
outgoing traffic on the physical port.
11
Three Priority of the Switch
8个COSqueue
ZXR10 Switch
COSqueue0
untag
frame COSqueue1
ingress port egress port
tagged COSqueue2
frame
COSqueue7
cos-local-map cos-drop-map
COS value(802.1P) Local-pri(COSqueue) COS value(802.1P) drop-pri(0G1Y2R)
0 1 0 2
1 2 1 2
2 3 2 2
3 4 3 1
4 5 4 1
5 6 5 1
6 7 6 0
7 0 7 0
12
Three Priority of the Switch
• Default priority: The priority of a specific port configured with the priority <interface><0-7> command is called the
default priority. When the packet received by the port is untag, the local priority of the packet is determined by the
default priority of the port (that is, the CoSqueue of the egress port to which the packet will enter when the packet is to
be forwarded).
• Local priority: A mapping relationship between 802.1p packets and 8 CoSqueue s at the egress port. The CoS value and
local priority mapping table are created through the ingress port configuration command cos-local-map xxx. According
to the 802.1p value of the packet received by the ingress port (the default priority value of the port is used for the untag
packet), query the mapping table to obtain the local priority value of the packet. The local priority value is CoSqueue, that
is, the local priority value of the packet is known. When the packet is to be forwarded, Which CoSqueue will enter the
egress port, and then participate in the queue dispatching of the egress port.
• Drop priority: For each packet entering the port from the ingress direction, the CoS value and the local priority mapping
table of the port are queried according to its 802.1p value (if the packet entering the port is untag, then the CoS value
and the local priority mapping table of the port are queried according to the default priority of the port (that is, the local
priority of the untag packet). The CoS value of the untag packet is obtained, that is, the 802.1p value of the untag
packet). The packets with different priorities are internally painted according to the mapping table (that is, the CoS and
the drop priority mapping table). Three drop priorities are marked: Low (green, =0), medium (yellow, =1) and high (red,
=2). In addition to queue congestion, the coloring result can be used for other discarding policies.
13
Configure the mapping table corresponding to the priority
Configure the mapping relationship between cos and drop priority(cos-drop-map)
ZXR10(config-pm-qos)#cos-drop-map <0-2><0-2><0-2><0-2><0-2><0-
2><0-2><0-2>
ZXR10(config-pm-qos)#trust-cos-dp-map <interface>{enable | disable }
Configure the mapping relationship between local priority and outgoing queue(cos-local-map)
ZXR10(config-pm-qos)#cos-local-map <0-7><0-7><0-7><0-7><0-7><0-7><0-
7><0-7>
ZXR10(config-pm-qos)#trust-cos-lp-map <interface>{enable | disable }
14
Description of the Mapping Table Corresponding to the Priority
• CoS value and local priority mapping table: It is created through the cos-local-map xxx command. In this
command, the first <0-7> indicates the local priority value of the 802.1p=0 packet (that is, the CoSqueue No.
that the packet enters), and the second <0-7> indicates the local priority value of the 802.1p=1 packet.
• CoS and drop priority mapping table: Created through the cos-drop-map xxx command. In this command, the
first <0-2> indicates the drop priority value of the 802.1p=0 packets (namely, the internal coloring of the
packets), and the second <0-7> indicates the drop priority value of the 802.1p=1 packets.
• DSCP mapping table: Created through the conform-dscp <0-63><0-63><0-7><0-2> command. The first <0-
63> is the source DSCP value carried in the packet, the second <0-63> is the new DSCP value carried in the
packet, the <0-7> is the new priority value carried in the packet (that is, the local priority value), and the <0-2>
is the packet drop priority.
Note:
The CoS value and local priority mapping table (local priority mapping for short), CoS and drop priority mapping
table (drop priority mapping for short), and DSCP mapping table must be configured on the ingress port.
For the local priority mapping configuration, the ingress port and egress port must be on the same interface board
or switching chip. The local mapping relationship is valid only inside the local switching chip. Once a board or
chip is overridden (the ingress port for mapping configuration and egress port for scheduling are not on one
interface board or even one switching chip), this relationship does not exist.
15
DS-domain Introduction
• The QoS configuration of the ZXCTN 6100 product is different from that of the ZXR10. For example, the
configuration of the DS domain, the ZXR10 switch does not have the configuration of the DS domain, and only
the M6000 V3 version has the DS domain configuration (not V2). The configurations of the 6100 and the
M6000 DS domain are also somewhat different. Another example is HQoS. The understanding and
implementation of different products are also different.
• Let's first introduce the DS domain. The DS domain is a mapping table containing the mapping relationship
between the external priority (8021p, dscp and exp carried in the packet) and the internal priority (phb (service-
class)) plus the color . This procedure describes how to bind a DS domain to a specific port to implement the
following functions:
– In the incoming direction, according to the priority of the packets received by the port, query the mapping table to obtain the
service-class and color of the packets inside the equipment. Service-class is a type of phb, including 8 service levels in total
be/af1/af2/af3/af4/ef/cs6/cs7, indicating the service level (or internal priority) of the packet inside the equipment. It is used for
subsequent traffic processing (such as queue scheduling and traffic shaping). Color contains green, yellow and red (value: 123).
It indicates the discard priority of the packets in the equipment. It is used for subsequent traffic processing (for example,
setting drop policy to avoid congestion).
– In the outgoing direction (only for the 6100 product), if the external priority field of the packet is newly added, the packet
priority is determined by the DS domain. That is, according to the service-class of the packet to be forwarded, query the
mapping table to obtain the external priority of the packet (uni and nni are not distinguished). If the packet carries the original
external priority field, it depends on whether egress remapping is configured (the remapping is only valid for uni). If yes, the DS
domain is used to find a new priority and overwrite the original priority of the packet. Otherwise, the original priority of the
packet remains unchanged.
16
DS-domain Introduction
8 service-class
ZXCTN 6100
802.1p=3 802.1p=3 service-class(EF)
ingress
service-class(AF4)
gei-0/0/0/1
service-class(AF3)
D
S
802.1p=5 802.1p=7
egress do
m
ai
service-class(BE)
n
bi
nd
in
te
ds-domain mapping table
rfa
ingress ce egress
external priority phb(service class) color external priority phb(service class) color
802.1p=3 AF4 yellow(AF42) 802.1p=5 AF4 green
802.1p=5 AF4 green(AF41) 802.1p=5 AF3 green
dscp=20 AF3 green dscp=46 AF4 yellow
dscp=30 AF4 green dscp=50 EF green
exp=5 EF green exp=2 AF1 yellow
exp=6 CS6 green exp=0 BE green
17
DS Domain Configuration Commands
On the 6100 product, the commands are as follows:
• Create a DS domain and configure intra-domain mapping relationships.
zte(config)#cfg-ds-field domain1
zte(config-domain1)#cfg-ds-ingress 802.1p<->phb PHB AF42 3
//The mapping from 8021p/dscp/exp to phb can be configured in the incoming direction. For example, the mapping
relationship here is 8021p=3 mapping to AF42 (where 2 indicates the yellow discard priority).
zte(config-domain1)#cfg-ds-egress 802.1p<->phb PHB AF31 5
//The mapping from phb to 8021p/dscp/exp can be configured in the outgoing direction. For example, AF31 is mapped to
8021p=5 (where 1 of AF31 indicates green discard priority).
• Binds the DS domain to the port
zte(config)#bind-ds-port gei-0/0/0/1 domain1 trust-priority 1
//This command is mandatory. The trust command only makes sense to the ingress direction. Trust-priority value: 1=802.1p,
2=dscp, 3=exp. At the ingress, determine the phb of the packet according to the trust value.
• Configure egress remapping
zte(config)#cfg-port-remap-8021p gei-0/0/0/1 enable
zte(config)#cfg-port-remap-dscp gei-0/0/0/1 enable
//This command is optional. The remap command is valid only for the egress direction and valid only for the uni egress. Uni If
the phb mapping of the egress needs to take effect, the egress remapping must be configured. Two remap commands can be
configured at the same time. The values of 8021p and dscp of the remapped packets can be configured.
© ZTE All rights reserved
18
DS Domain Configuration Commands
On the M6000-S product, the commands are as follows:
19
Differences of HQoS between different products——ZXR10 8900E
The HQoS of ZXR10 8900E is understood as hierarchical meter. Hierarchical meter can set two levels of rate limit:
MICRO METER and MACRO METER. MACRO METER supports two modes: Bandwidth borrowing and overall
bandwidth limit.
1. Bandwidth Borrowing
The total bandwidth macroflow is 100 M
(rate limit value), and the bandwidth
microflow1- 4 of each stream is 20 M, 30
M, 40 M and 10 M (rate limit value).
If the actual traffic is flow1=50 M,
flow2=30 M, flow4=20 M and flow3,
flow1 and flow4 can borrow flow3
bandwidth, but the total bandwidth
should not exceed macroflow 100 M.
20
Differences of HQoS between different products——ZXR10 8900E
2. Total Bandwidth Limit
The total bandwidth macroflow is 80 M
(rate limit value), and the bandwidth
microflow1- 4 of each stream is 20 M, 30
M, 40 M and 10 M (rate limit value).
If the actual traffic is flow1=50 M,
flow2=30 M, flow4=20 M and flow3, there
is no traffic at present. Since bandwidth
borrowing is not allowed in this mode, the
traffic that flow1 can pass is 20 M, the
traffic that flow2 can pass is 30 M, and the
traffic that flow4 can pass is 10 M. The
total bandwidth does not exceed 80 M.
21
Differences of HQoS between different products——ZXR10 M6000-S
The HQoS of ZXR10 M6000-S controls the total bandwidth of multiple queue s through multi-level
scheduler. These queue s may come from the same service or different services of the user, so the
QoS of one or multiple services can be controlled.
22
Thank You!