Meituan Network Engineer Interview Experience | Life / Interview Experience

Protocol Questions

Aging Times

MAC table aging time
- 300s (5 minutes)
ARP table aging time
- 120s
- Purpose: periodically clears useless IP-to-MAC address mappings to keep the table from growing too large and to reflect network topology changes
RIP table aging time
- 180s
TCP aging time
- 3600s (1 hour)
UDP aging time
- 300s

STP Protocol

Explain STP
- Layer 2 Spanning Tree Protocol
- Purpose: blocks redundant links to eliminate Ethernet loops while providing hot standby
Describe the STP working process
1. Step 1: Elect a single root bridge
  - All switches send BPDUs to compare BIDs; the one with the smallest value becomes the root bridge
    - BID composition: the first 2 bytes are the priority, and the last 6 bytes are the switch’s MAC address
  - Default initial priority: 32768 (the smaller the value, the higher the priority)
  - In automatic election, since the priorities are the same, MAC sizes are compared and the smallest is automatically elected
2. Step 2: Elect the root port
  - Calculate the cost from each port to the root bridge and select the lowest cost
    - The lower the bandwidth, the lower the cost
    - 100 Mbps cost 19; 1 Gbps cost 4; 10 Gbps cost 2
  - If ports have equal costs, compare the switch BPDUs; the smallest wins
  - If all of the above are identical, select the port on the local switch with the smallest port number, e.g., port 1
    - A fallback strategy for extremely rare cases
3. Step 3: Elect the designated port
  - All ports on the root bridge are designated ports
  - Non-root bridges have only one designated port
  - All ports except the root port go through the root port election mechanism again; the one with the lowest cost becomes the designated port
4. Step 4: Port state transitions
  - Non-designated ports enter the blocking state
  - Root ports and designated ports enter listening and learning, then enter the forwarding state
    - Listening: sends and receives BPDUs to confirm there is no loop conflict
      - If there is a loop conflict, enter blocking and begin re-electing a new port
      - If there is no loop conflict, enter learning
    - Learning: builds the MAC table and records paths

ARP

What does the ARP protocol do?
- Purpose: converts IP addresses into MAC addresses
- Process: broadcasts an ARP request on the LAN to look up the MAC address corresponding to the target IP. All devices on the LAN receive the broadcast frame, but only the device holding the target IP replies directly in unicast form to the original requesting device

Vlan

Is VLan division based on devices or ports?
- Based on switch ports
How do VLANs communicate with each other?
- Through Layer 3 forwarding using the VLANIF interface of a Layer 3 switch
How many configuration modes do VLAN interfaces have?
- Access
  - Used to connect end-user devices
  - Only accepts and sends untagged data frames of that VLAN
- Trunk
  - Used to connect network devices to each other
  - Can carry data frames of multiple VLANs, distinguishing different VLAN traffic by the VLAN tag added in the frame header
  - Strips the VLAN tag from the data frame when sending
    - The frame’s default tag is vlan1
  - When it receives an untagged frame, it tags it with its own tag
    - But this tag is only used for internal transmission within the switch; it is stripped again when the frame is sent out
- Hybrid
  - A combination of Access and Trunk that can be flexibly configured
  - Can decide whether a VLAN tags data frames

Switch Stacking Technology

What problems can switch stacking solve?
1. Expand the number of ports
  - Ports of added members immediately become usable ports
2. Simplify management
  - There is only one management IP; all configuration changes take effect and are automatically synchronized to member switches
3. Improve reliability
  - Has a master/standby redundancy mechanism
    - By default, one switch is elected as the master switch to handle control
      - First compare priority; if equal, compare MAC size; if still equal, compare the stack ID
        
        The stack ID defaults to 1; a later-added switch conflicts with it and changes to 2, and so on
        
        Identical MACs basically never happen
    - Designate one or more standby switches
    - If the master switch fails, it switches over to the standby switch within milliseconds, basically imperceptibly
4. Eliminate internal bandwidth bottlenecks
  - Member switches connect through dedicated stacking links, allowing the internal link bandwidth to be aggregated
  - The stacking bandwidth is used only for data exchange between members and does not consume external bandwidth
  - Stacking bandwidth = single link bandwidth x number of links
    - The stacking bandwidth between different members must be calculated separately

Routing

What is a floating route?
- It is a backup route for the primary route
- When the primary route fails, the backup route is enabled
- It cannot be generated automatically; the route priority is configured manually
  - Huawei default priority 60, Cisco default 1
  - The smaller the value, the higher the priority

Wireless

What wireless configuration modes are available by default?
1. Fat AP mode
  1. The switch connects to the fat AP; the AP independently handles SSID broadcasting, user authentication, and data forwarding — all operations are done on the AP
  2. The switch is only responsible for power supply and data transmission
  3. Suitable for small home networks
2. Thin AP + AC mode
  1. A thin AP finds the AC via DHCP, downloads the configuration from the AC, and broadcasts the SSID issued uniformly by the AC
  2. When a user connects to an AP, authentication requests are handed to the AC for processing
  3. Can be configured to go directly through the AP or through the AC tunnel
  4. Suitable for schools and large companies, making it easy to centrally manage many Wi-Fi points
3. Agile AP networking
  1. A central AP connects to the switch, communicates with the AC, and manages one or more RUs
  2. RUs connect to the central AP via Ethernet cable, receive the configuration, and broadcast it
  3. User data is passed from the RU to the central AP, which then forwards it centrally to the switch
  4. Suitable for hotels and dormitories where an RU is placed in each room; convenient cabling and easy management
What are bypass mode and inline (direct) mode?
1. Bypass mode
  1. The AC is directly connected to the core switch
  2. AP data goes directly to the AC, and the AP registers with the AC
  3. Data forwarding: after receiving user data, the AP sends it directly to the AC through the CAPWAP tunnel
  4. After the AC finishes processing, it forwards the data from the core switch directly to the upper-layer network
2. Inline (direct) mode
  1. The AC is not directly on the data forwarding path of the core switch
  2. It hangs off a port of the core switch or an aggregation switch
  3. AP data first goes to the core switch, which then hands it to the bypass AC; after the AC finishes processing, it hands it back to the core switch to be forwarded out
  4. The AC is only responsible for control and management; it does not handle the forwarding path
In bypass mode, will an AC failure affect wireless internet access?
1. Without redundant backup, yes, it directly affects internet access
2. With redundant backup
  1. Dual-link hot standby
    1. Thanks to pre-connection and data synchronization, after startup the AP initiates connections to both the master AC and the standby AC, establishing two independent CAPWAP tunnels
    2. When the master AC goes offline, the standby AC takes over within milliseconds, with an imperceptible experience
  2. Dual-link cold standby
    1. Since the data on the master AC and the standby AC is not synchronized in cold standby, after the master AC fails you must wait for the following process
      1. AP discovers the AC
      2. Establish the tunnel
      3. Complete registration
      4. Obtain configuration and user information
    2. During the switchover, there will be a brief network outage of a few minutes, and existing connections will also drop
What are the differences in requirements on the AC device between bypass mode and inline (direct) mode? And what is the difference between the two modes?
1. In inline (direct) mode
  - Because the AC must directly process and forward all user data, performance requirements are high
  - Hard to scale and hard to replace; if redundant backup is lacking, replacement or upgrade causes a network outage
2. In bypass mode
  - The AC is mainly responsible for control; data forwarding is done by the core switch, so performance requirements are low
  - Easy to scale and replace; after replacement, ordinary PSK authentication and local forwarding are unaffected
  - However, if re-authentication or key renegotiation is enabled and depends on the AC; or if dynamic bandwidth scheduling requiring continuous AC interaction is enabled, existing users will also lose connectivity
  - If 802.1 authentication is enabled without the escape function configured, user authentication into the network is affected during device replacement

Dynamic Routing

What dynamic routing protocols are there?
- OSPF
- IS-IS
- BGP
- RIP
- EIGRP (Cisco proprietary protocol)
- IGRP (Cisco proprietary protocol, obsolete)
What are the OSPF network types?
1. Broadcast
  1. Supports multicast
  2. Supports automatic neighbor discovery
  3. Elects DR/BDR
2. Non-Broadcast Multi-Access (NBMA)
  1. Does not support multicast
  2. Neighbors must be configured manually
  3. Elects DR/BDR
  4. Supports Frame Relay
3. Point-to-Point (P2P)
  1. Only two nodes; automatic neighbor discovery
  2. Does not elect DR/BDR
  3. Like a PPP link with a dedicated line
4. Point-to-Multipoint (P2MP)
  1. One node connected to multiple nodes
  2. Neighbors must be configured manually
  3. Does not elect DR/BDR
5. Virtual Link (VL)
  1. A logical point-to-point link
  2. Neighbors configured manually
  3. Does not elect DR/BDR
  4. Used to connect to a non-adjacent area 0
What is the basis for OSPF path selection?
1. Computes the shortest path tree based on the link-state database
2. Compares cost values, which are determined by bandwidth; the larger the bandwidth, the lower the cost
3. The router triggers the SPF algorithm, using itself as the root to compute the total cost of all paths from the root to each destination network
  - Establishing a neighbor and synchronizing the link-state database triggers the router to run the SPF algorithm, which runs once
  - When the network topology changes, the router synchronizes the link-state database and triggers SPF again
4. Finally, the path with the lowest total cost is selected as the optimal route
5. If several paths have equal cost, OSPF enables load balancing and uses all these paths to forward data simultaneously
When the OSPF cost value affects OSPF path selection, is it based on the incoming interface’s cost or the outgoing interface’s cost?
- Incoming interface
Why the incoming interface?
- Because the calculation is performed from the local router's perspective
- The router only knows the cost of its own interfaces; when it receives an LSA from a neighbor, it adds the cost of the interface through which the LSA enters the local router to the total cost
How does OSPF prevent loops?
1. Area partitioning: non-backbone areas can only communicate directly with backbone area 0, avoiding routing loops between areas
2. SPF algorithm: each router computes the shortest path with itself as the root, generating a loop-free routing table
3. Rules for propagating routing information
  1. LSAs within an area are flooded only within that area and never cross areas, preventing inter-area information confusion that could form loops
  2. When an ABR passes routes to a non-backbone area, it filters out routing information from other backbone areas, ensuring that non-backbone areas can only obtain routes of other backbone areas through backbone area 0
What is the difference between a Stub area and an NSSA area in OSPF?
1. A Stub area mainly filters external LSAs and only allows one default route in
  1. All routers in the area must be configured as Stub; ASBRs are not allowed
2. NSSA is an extension of Stub: besides filtering external LSAs, it also allows external routes to be introduced into the area, propagated via Type7 LSA, then converted into Type5 LSA at the ABR and propagated to the other areas
  1. Routers in the area need to be configured as NSSA and allow ASBRs to exist, for introducing external routes into the OSPF domain
How many types of LSAs does OSPF have, and what are they used for?
1. Type 1
  1. Generated by all routers, propagated only within the local area
  2. Describes the router's directly connected links
2. Type 2
  1. Generated by the DR, propagated only within the local area
  2. Describes the multicast capability of an MA network
3. Type 3
  1. Generated by the ABR, propagated throughout the entire OSPF domain
  2. Used for inter-area route propagation
4. Type 4
  1. Generated by the ABR, propagated throughout the entire OSPF domain except the area where the ASBR resides
  2. Used to discover and locate the ASBR
5. Type 5
  1. Generated by the ASBR, propagated throughout the entire OSPF domain
  2. Used to describe external routes
6. Type 6
  1. Generated in multicast OSPF; rarely used in practice
  2. Used to describe multicast group membership
7. Type 7
  1. Generated by the NSSA area, propagated within the NSSA area
  2. Used to propagate external routes within the NSSA area, which are then propagated to other areas by the ABR's Type5 LSA