-

4.5.6E

Support FS Models N8610-32D and N8650-32OD

-

4.5.5E

Support FS Models N8510-24CD8D

-

4.5.4E

Support FS Model N8520-32D

-

4.5.3E

Support FS Model N9550-64D

-

4.5.2E

Support FS Model N9600-64OD

-

4.5.1E

Support FS Model N8550-24CD8D

-

4.5.3E

MOD CLI

A technology specifically designed to monitor packet loss during internal device forwarding of messages. Once MOD detects packet loss inside the device, it will immediately collect the time of packet loss, the reason for packet loss, and the characteristics of discarded messages, and report them to the remote collector so that administrators can timely understand the packet loss situation inside the device.

Currently, this feature supports three types of packet drop monitoring through CLI configuration: ingress-l3-dst-lookup-miss, ingress-l3-header-err, and ingress-l3-ttl-err.

-

4.5.3E

Differentiated Flow Scheduling for Elephant and Mice Flows

Network traffic can generally be categorized into high-volume (elephant) and low-volume (mice) flows based on the amount of data transmitted within a given time period.

This feature is supported on the following model:
N9550-64D

-

4.5.3E

CMIS Protocol Optical Module Information Reading

This feature introduces comprehensive CMIS protocol–based monitoring and diagnostic capabilities for 400G optical modules, enabling precise control and visibility into module performance and link stability.
It enhances optical module management efficiency and provides strong technical support for high-performance computing (HPC) and large-scale data center networks.

Feature Capabilities:

1. Port link-up time monitoring – Records and monitors the link-up time of each port.

2. Optical module register information reading – Supports reading detailed CMIS register data from optical modules.

3. Repeated laser toggling (on/off) – Allows repeated laser activation/deactivation for optical module stability testing.

4. Single-byte repeated register reading – Enables continuous single-byte read operations for debugging or validation purposes.

5. Port alarm status monitoring (DDM) – Reads and monitors port alarm and warning information such as temperature, voltage, and optical power.

Supported models:

• N9550-64D

-

4.5.3E

ROCE EasyDeploy

The switch collaborates with servers to enable one-click RoCE deployment, supporting both lossless and lossy modes (default: lossless), with the ability to switch between modes:

• Lossless mode: Enables PFC and ECN, with ECN configured for WRED (Weighted Random Early Detection) and QoS policies applied.

• Lossy mode: Only ECN is enabled (without PFC), with ECN configured for WRED and QoS policies applied.

This feature is supported on the following models:
N9550-64D

-

4.5.0E

IPv6 ND Inspection

IPv6 Neighbor Discovery (ND) Inspection is a security feature designed to enhance the protection of IPv6 networks by managing and validating Neighbor Discovery Protocol (NDP) messages, which are essential for the proper operation of IPv6 communication. Please have the details by reference document IPv6 Neighbor Discovery Inspection.

-

4.5.0E

IPv6 ND Snooping

IPv6 Neighbor Discovery (ND) Snooping is a security feature that safeguards IPv6 networks to prevent various types of attacks. It functions similarly to ARP (Address Resolution Protocol) Snooping in IPv4 networks. Please have the details by reference document IPv6 Neighbor Discovery Snooping.

-

4.5.0E

MPLS

MPLS (Multiprotocol Label Switching) operates between the link layer and the network layer in the TCP/IP protocol stack. It offers connectivity services to the IP layer while leveraging services from the link layer. Unlike traditional IP forwarding, MPLS uses label switching to direct traffic through the network. Please have the details by reference document MPLS Configuration.

-

4.5.0E

PIM BSR (Bootstrap Router)

Dynamic RP allows multiple PIM devices within a PIM domain to be configured as C-RPs (Candidate RPs). Among these C-RPs, an RP is determined through an election process. The BSR aggregates information from all C-RPs in the network into an RP Set using Bootstrap messages and distributes it to all PIM devices. Each PIM device uses the RP Set to calculate and compare based on consistent rules, ultimately selecting an RP from the available C-RPs. Please have the details by reference document PIM Configuration Guide.

-

4.5.0E

Ingress Buffer

Supports ingress buffer management, including guaranteed/shared/headroom management. Please have the details by reference document Configuring PFC Buffer.

-

4.5.0E

PFC Watchdog Manual Control

The PFC Watchdog feature detects and resolves PFC (Priority Flow Control) deadlocks. Recovery methods include both automatic and manual recovery, allowing users to choose the appropriate approach for resolving deadlock scenarios. Please have the details by reference document Configuring PFC Watchdog.

-

4.5.0E

PFC Deadlock Prevention

To avoid PFC deadlock issues, the DSCP value and corresponding Dot1p priority of the message can be modified so that the modified message can be forwarded using the new DSCP value in the new Dot1p priority queue, avoiding messages with the same DSCP value from remaining in PFC deadlock state. Please have the details by reference document Configuring PFC Deadlock Prevention.

-

4.5.0E

Easy ECN

Explicit Congestion Notification (ECN) is a congestion notification mechanism operating at the IP and transport layers, serving as an extension to the TCP/IP protocol. With Easy ECN, users can enable WRED (Weighted Random Early Detection) policies, configure WRED thresholds, and set the maximum packet loss probability to manage network congestion more effectively. Please have the details by reference document Configuring Easy ECN.

-

4.5.0E

DLB (Dynamic Load Balance) 

DLB breaks through the limitations of traditional static hash mechanisms by introducing timestamp and real-time load measurement factors (port bandwidth load, queue size) to optimize load balancing in both time and bandwidth space dimensions, providing a dynamic and intelligent hash mechanism. Please have the details by reference document Configuring Dynamic Load Balancing.

-

4.5.0E

Standby IP Address

In cases where the management port cannot connect to the DHCP server and no static IP has been set through CLI during the switch's startup, it will default to using the secondary management IP address 192.168.1.1. This IP address serves as a backup, allowing management of the device even if DHCP services are unavailable. It is primarily used when the management port is directly connected to a PC, ensuring uninterrupted device management via this IP address. Please have the details by reference document Default Settings for Out-of-band Management Interface.

-

4.5.0E

Perpetual PoE

Perpetual PoE (also known as hot-start uninterruptible power supply or permanent PoE) refers to the ability of Power Sourcing Equipment (PSE) to continue providing power during a system restart. This includes restarts initiated through CLI commands such as "request system reboot" or by rebooting under the Linux shell. Additionally, it supports uninterrupted power during system upgrades, including upgrades triggered via CLI or Linux-based upgrade methods. This feature ensures that PoE-powered devices remain operational even when the system is restarting or undergoing an upgrade. Please have the details by reference document Configuring Perpetual PoE.

-

4.5.0E

PFC/ECN

GRPC monitoring PFC and ECN, in conjunction with gRPC, can provide PFC pause frame counts, PFC deadlock monitoring and ECN-marked packet counts for statistical queries. Please have the details by reference document PFC and ECN Statistical Reporting through gRPC.

-

4.5.3E

Auto Negotiation/Link Training

• Support the ability to display the current interface link training status.

• Support for the N9550-64D platform.

-

4.5.3E

Port Breakout

Support three types of port breakout:

• 400G to 2*200G

• 400G to 4*100G

• 200G to 2*100G

-

4.5.0E

DHCP Server Enhancement

In versions prior to 4.5.0E, clients were unable to obtain an address in a DHCP relay scenario. However, starting from version 4.5.0E, this issue has been resolved, and the system now fully supports DHCP address assignment in relay scenarios.

-

4.5.0E

DHCP ZTP

After enabling the DHCP server with PicOS, address pools can be configured to allocate IP addresses to clients, along with additional network information such as gateway, DNS server addresses, log server addresses, TFTP server addresses, boot file names, and other options. These configurations are applied and synchronized with the clients as addresses are allocated. Please have the details by reference document Zero Touch Provisioning (ZTP).

-

4.5.0E

400G Port Splitting

The N9550-32D/AS9716-32D switches support the capability to split a 400G port into 2 * 200G and split to 4 * 100G ports, providing flexible bandwidth allocation for diverse network needs. Please have the details by reference document Configuring Port Breakout and Merge.

-

4.5.0E

PBR ECMP

PBR (Policy-Based Routing) action supports not only specifying a next-hop router or modifying DSCP values, but also enables the use of nexthop-group for ECMP (Equal-Cost Multi-Path) routing, allowing for more advanced and efficient traffic distribution across multiple paths. Please have the details by reference document Policy-Based Routing (PBR).

-

4.5.0E

L2L3 WEB Access Control

Support is provided for using the command set system services web disable to modify the status of Layer 2 and Layer 3 WEB access, allowing administrators to enable or disable web access.

The following switches support L2L3 WEB access, which is enabled by default:

• S5810-48TS-P

• S5810-28TS

• S5810-28FS

• S5810-48TS

• S5810-48FS

• S5860-20SQ

• S5860-24XB-U

• S5860-24MG-U

• S5860-24XMG

• S5860-48XMG-U

• S5860-48XMG

• S5860-48MG-U

The following switches support L2L3 WEB access, but WEB access is disabled by default:

• S5870-48T6S-U

• S5870-48T6S

• S5870-48MX6BC-U

• S5870-48T6BC-U

• S5870-48T6BC

21166

4.5.6E

[N8650-32OD] On N8650-32OD running PicOS 4.5.6E, the pica_lcmgr process could crash when an exception occurred during port speed downgrade operations.

18933

4.5.6E

[MLAG] In MLAG scenarios, MAC addresses were not aged out when a LAG went down and the peer-link did not include the corresponding VLANs.

18671

4.5.6E

[MLAG] In MLAG environments with large MAC tables (~30K entries), when deleting ae1 on DUT1, MAC addresses learned on ae1 were not correctly moved to the peer-link (ae10).

18594

4.5.6E

[MLAG] In MLAG, the peer status shown by the show mlag link summary command incorrectly displayed as up when the peer-link was down. The expected status is unknown.

20888

4.5.6E

[N8650-32OD] On N8650-32OD, error logs could be generated after the switch configuration was split and the system was rebooted.

20239

4.5.6E

[N8650-32OD] On N8650-32OD, only one power module was displayed after a system reboot.

20960

4.5.6E

[AN_LT] On N8610-32D, when a 200G DAC cable was connected to a 400G port and split into four lanes, auto-negotiation was incorrectly disabled.

20922

4.5.5E

When 200G or 400G DAC cables were downgraded to 100G, the affected ports could fail to come up. This issue has been resolved.

19901

4.5.5E

The switch could fail to load when running rc_hw.sh due to the mdio_gpio module not being enabled in the kernel.

20266

4.5.5E

In ONIE mode, after selecting Uninstall OS and then performing a rescue ONIE installation, the system could fail to boot into PicOS after installation.

20860

4.5.5E

When a 400G port was split into four 100G ports and forced to operate at 100G, the ports could go down and require manual LT configuration to recover.

19955

4.5.4E

[System Service] Fixed an issue where executing sudo service picos stop after enabling real-time printing resulted in the error: "Error: /tmp/system/old_nslcd.conf does not exist".

20014

4.5.4E

[Port Mirroring] Resolved an issue where the number of packets received on the mirrored port was incorrect after traffic transmission.

20029

4.5.4E

[ACL] Fixed an issue where disabling or enabling the Spanning Tree command triggered ACL-related error logs in the BCM shell mode.

19944

4.5.4E

[IPv6 RA] Fixed a failure when adding an IPv6 RA filter rule to the line card.

20187

4.5.4E

[CoPP] Resolved an issue where the show command for the CoPP node was not functioning.

20208

4.5.4E

[Port Security] Fixed an issue where the port-security function only allowed known unicast packets to pass, while blocking multicast, broadcast, and unknown unicast packets despite configuration.

20389

4.5.4E

[NETCONF] Fixed an issue where configuring interface-related commands via NETCONF would fail.

20459

4.5.4E

[PFC] Resolved a false "Only two mappings are allowed" error that occurred after deleting one mapping and adding a new one, which caused submission failure.

20516

4.5.4E

[CMIS] Fixed an issue where the N8520-32D model does not support the command for modifying registers.

13698

4.5.3E

[BGP] Default BGP Weight Inconsistency
The configured IPv6 addresses on the connected interfaces of the two switches show different BGP weights in the routing table. The expected behavior is that both should use the same default weight value.

15507

4.5.3E

[MLAG + Anycast GW]

After MLAG status is full, changing a VLAN interface from the default VRF to VRF1 and configuring IPv6/Anycast IPv6 addresses results in a neighbor table synchronization failure between MLAG peers.

18960

4.5.3E

[N8550-24CD8D/N9550-64D]

The ACL matching condition is mismatched - packets without TCP/UDP headers are mistakenly matched against the rule source-port 0..0.

A large number of sequential ACL rules were configured. Among them, filter f1 sequence 4000 specifies ICMP type 0 and ICMP code 57. When ICMP packets with type 0 and code 57 (which do not contain TCP/UDP headers) are sent, they are incorrectly matched by sequence 1300, which defines the condition source-port 0..0.

13587

4.5.3E

[FRR: MLAG]

Fixed an issue where, after MLAG reached the full state, deleting a static MAC address on one peer caused the corresponding dynamic MAC address on the other peer to fail to synchronize.

18922

4.5.3E

[N9550-64D 4.5.3E]

Fixed an issue where, after inserting and activating a 400G module, changing the port rate to 100G did not update the port status correctly - the port remained displayed as 400G up.

17889

4.5.3E

[FRR: OSPF6]

The ifindex value of the interface that introduces the route is inconsistent before and after the port UP/DOWN operation.

After configuring static OSPF6 routes on DUT1, the ifindex value associated with the static routes changes when the interface is toggled (UP/DOWN), resulting in an inconsistency compared to the initial state.

16367

4.5.3E

[FRR] The Hostname Display is Inconsistent across Different Platforms

When running in FRR mode, some switches display hostname PICOS in the show running-config output, while others do not.

18820

4.5.3E

[MLAG]

The Partner Port Number information displayed for the LACP port under MLAG is incorrect.

18911

4.5.3E

[N9550-64D | 4.5.3E]

The QDD-SR4-2x100G module could not be recognized on the Tomahawk4 platform.

19706

4.5.3E

[Tomahawk4]

When a split port is configured as 200G, the headroom size is calculated based on the 10G port configuration, resulting in incorrect headroom allocation.

19743

4.5.3E

[MLAG Trident4]

If VLAN 1 is included in the peer-link VLAN and additional links exist between the two MLAG devices other than the peer-link, MLAG synchronization packets may loop between the devices.

18217

4.5.3E

[S6860-24CD8D]

When MLAG and L3 Anycast are configured simultaneously, ARP entries cannot be learned properly.

18197

4.5.3E

[S6860-24CD8D | MLAG]

Traffic received from the peer-link should be dropped on the peer-link port in an MLAG environment.

17752