Overview

In B2B networking, the MAC address table size—typically 16K or 32K entries—determines how many unique devices a switch can learn before flooding traffic. Choosing the wrong capacity leads to broadcast storms, packet loss, and core backbone congestion. This FAQ answers both pre-sales and post-sales engineering questions, covering optical module compatibility, maximum throughput, backplane architecture, CLI configuration, stacking interfaces, and redundant power failover. All answers are structured for immediate technical resolution.

Frequently Asked Questions

Q1: What is the real-world maximum switching throughput (Gbps) of a switch with 32K MAC address table?

The maximum switching throughput is typically 176 Gbps to 1.28 Tbps for a 32K MAC table switch, depending on port density and ASIC architecture. A 48-port Gigabit switch with 4 x 10G uplinks achieves 128 Gbps switching fabric, while a 32K table enables line-rate non-blocking forwarding because the table can hold all directly connected endpoints. For 10G deployments, ensure the switch fabric is at least 480 Gbps to avoid oversubscription.

Q2: Is my SFP+ optical module compatible with a switch that has a 16K MAC table?

Yes, optical module compatibility depends on the transceiver form factor (SFP, SFP+, QSFP28) and vendor coding—not on the MAC table size. However, a 16K table limits how many unique MAC addresses can be learned across those optical uplinks. For aggregation switches with 16K table, use a maximum of 16 downstream switches (each contributing ~1000 MACs) per optical module. Always verify DOM (Digital Optical Monitoring) support and DDMI alarms; they are unaffected by MAC table size.

Q3: How does MAC table size 32K compare to 16K for Layer 3 routing protocols like OSPF and BGP?

A 32K MAC table directly supports larger Layer 2 domains under a single VLAN or VXLAN segment, but it does not replace the Layer 3 FIB (Forwarding Information Base) for OSPF or BGP. For an enterprise core running OSPF, the MAC table holds ARP entries (IP-to-MAC mappings) for directly connected neighbors. With 32K, you can peer up to 32K adjacent routers on the same broadcast segment; 16K would flood after 16,000 peers. For BGP, the MAC table size has zero impact on route table limits (typically 128K-1M IPv4 routes).

Q4: What is the maximum backplane capacity required to support 32K MAC addresses without packet loss?

The backplane capacity must be at least twice the total port throughput to avoid head-of-line blocking. For 32K active MACs generating 10% line-rate traffic, a 48x1G + 4x10G switch requires a 176 Gbps backplane (non-blocking). If the backplane is undersized (e.g., 96 Gbps), the switch will drop frames when the MAC table is full because internal buffers cannot drain fast enough. As a rule: backplane capacity (Gbps) = (number of ports × port speed) × 1.2 oversubscription factor for a 32K table.

Q5: How do I troubleshoot redundant power supply failover on a switch with 32K MAC table?

First, verify PSU status using CLI command show power inline or show environment power. A healthy redundant PSU should show both as active or one as standby with no alarms. To test failover: remove primary AC feed while monitoring console logs; the secondary PSU must switch within 50 ms (typical hold-up time). If MAC table resets during failover, your PSU capacitors are degraded. Check syslog for "Power supply redundancy lost" followed by "MAC table clear"—this indicates insufficient hold-up capacitance. Replace PSU modules with hot-swappable 80+ Platinum units rated for continuous 32K MAC learning.

Q6: What is the correct CLI configuration to prevent MAC table overflow attacks on a 16K table?

Configure port security with maximum MAC addresses per interface. For a 16K global table, set switchport port-security maximum 64 on user-facing ports and maximum 512 on trunk ports. Then enable sticky MAC learning: switchport port-security mac-address sticky. To protect against CAM table overflow (MAC flooding), globally enable mac address-table aging-time 300 and storm-control broadcast level 10.00. For critical VLANs, implement mac address-table static entries for gateway MACs. Always monitor with show mac address-table count — action if usage exceeds 80% of 16K.

Q7: Can I stack multiple switches with 32K MAC tables to create a virtual chassis with 128K shared MAC capacity?

No, stacking does not aggregate MAC table sizes. In a stack of four 32K MAC switches, each unit maintains its own 32K table. The stack master synchronizes MAC addresses via a dedicated stacking interface (typically 40 Gbps), but each switch still learns a maximum of 32K locally. To achieve 128K shared capacity, you must use a distributed chassis system with a common backplane (e.g., Cisco VSS or Juniper VCF). For standard stacking, ensure stack bandwidth is at least 80 Gbps to avoid MAC sync latency; otherwise, you will see unicast flooding across stack members.

Q8: How do I identify if my 16K MAC table is causing throughput degradation? And what is the upgrade path to 32K?

The primary symptom is high CPU utilization (above 60%) due to constant MAC learning and aging, plus unknown unicast flooding visible via show interfaces | include unicast. Run show mac address-table aging-time — if age-out count exceeds 500 per second, your table is thrashing. For a software upgrade path: first check if your switch supports dynamic table expansion via license (e.g., from 16K to 32K). If not, replace with a switch featuring 32K as base hardware. For physical migration: use mac address-table static to pin critical MACs before swapping hardware. Downtime: plan 15 minutes for a 48-port access switch; zero downtime if using VRRP with dual switches.