Overview

This technical FAQ addresses the most critical pre-sales and post-sales questions about QoS priority queuing, Class of Service (CoS), and Differentiated Services Code Point (DSCP) on B2B carrier-grade switches and routers. Optimized for network architects and procurement leads, we cover line-rate throughput, buffer architecture, CLI-based classification policies, optical transceiver compatibility, stacking interface behavior, and redundant power failover sequences.

Frequently Asked Questions

Q1: What is the maximum switching throughput and per-port QoS buffer capacity for the {keyword}?

The maximum switching capacity of the {keyword} is 2.56 Tbps with a forwarding rate of 1,440 Mpps for a 48x10G + 8x100G configuration. For QoS priority queuing, the dedicated buffer pool is 32 MB shared dynamically across all ports, with a per-queue hard limit of 8 MB for strict priority queues. This architecture ensures line-rate forwarding even when all eight CoS queues are saturated on 100G uplinks.

Q2: How does the backplane architecture differ from previous-generation 1G/10G fixed switches when implementing DSCP classification?

Unlike previous-generation switches that used a centralized CPU for DSCP remarking, the {keyword} features a distributed P4-programmable pipeline. This enables wire-speed DSCP-to-CoS and CoS-to-Queue mapping at Layer 3. Previous models suffered 15-20% throughput loss when enabling more than four custom DSCP classes; the current backplane eliminates any performance penalty for up to 64 unique DSCP values and eight egress queues per port.

Q3: What optical module compatibility must be verified before deploying QoS-aware links on the {keyword}?

The {keyword} accepts generic SFP/SFP+/QSFP28 optics with DOM (Digital Optical Monitoring) support, but QoS priority queuing requires low latency on specific uplinks. For guaranteed CoS/DSCP performance, use only coded modules that report real-time temperature and voltage via I2C bus. Compatible third-party optics with Broadcom CDR firmware version 2.3+ are certified. Uncertified modules may still pass traffic but will fail hardware-based priority queuing during temperature fluctuations above 70°C.

Q4: What is the basic CLI configuration to map DSCP EF (46) to a strict priority queue on the {keyword}?

Use the following three commands: 'class-map match-any QOS_VOICE' then 'match dscp 46', then 'policy-map QOS_OUT' with 'class QOS_VOICE' and 'priority level 1'. Finally apply 'service-policy output QOS_OUT' to the interface. This ensures DSCP 46 packets bypass weighted fair queuing and enter queue 7 with absolute priority. Verify with 'show qos interface gigabitethernet 1/0/1' to confirm queue assignment and drop counters.

Q5: How do stacking interfaces affect CoS queuing and DSCP trust boundaries in a virtual chassis?

In a stacked configuration, the {keyword} preserves CoS and DSCP values across the backplane stacking cables (40G or 100G). However, the stacking interface itself uses a separate internal priority queue (CoS 7) for control plane traffic. During a stack master failover, DSCP trust boundaries temporarily revert to port default (untrusted) for 3-5 seconds. Best practice: configure 'qos trust dscp' on all member interfaces after stacking negotiation completes to maintain consistent priority queuing.

Q6: During a redundant power supply failure, does the {keyword} preserve priority queuing state?

Yes, the {keyword} preserves all CoS/DSCP priority queuing states during a hot failover from primary to redundant power supply. The switch has a 20ms holdup capacitor bank that maintains buffer table integrity. However, if both PSUs fail and the switch reboots, the default QoS configuration applies. For mission-critical deployments, enable 'qos auto-save interval 300' to commit running priority policies to NVRAM every five minutes.

Q7: What is the maximum jumbo frame size supported with priority queuing and DSCP classification enabled?

The {keyword} supports a maximum jumbo frame size of 12,000 bytes with full priority queuing and DSCP classification active. Compared to the previous generation (9,216 byte limit), this 30% increase allows 10G and 100G ports to carry large storage traffic (e.g., NFS over RoCE) while preserving strict priority for voice. Frame sizes beyond 12,000 bytes bypass the QoS pipeline and are handled by the CPU, dropping priority queuing.

Q8: How many hardware queues per port and what scheduling algorithms are available for CoS-to-DSCP mapping?

The {keyword} provides eight hardware queues per egress port. Scheduling algorithms selectable per queue: Strict Priority (SP), Deficit Weighted Round Robin (DWRR), and Weighted Fair Queuing (WFQ). For CoS-to-DSCP mapping, the device supports 1:1, many:1, and DSCP mutation maps of up to 64 entries. A common deployment maps CoS 5 (voice) to DSCP 46 (EF) with SP scheduling, and CoS 2-4 to DSCP AF2x-AF4x with WFQ.