Datasheet

Alcatel-Lucent Page 73
OmniSwitch 9000
every port. This feature greatly minimizes the number of VLANs required.
Port mapping feature is supported on OS6800/6850/9000. Following are the limitations for the feature.
• 8 sessions supported per standalone switch and stack
• An aggregable port of a link aggregation group cannot be a mapped port and vice versa
• A mirrored port cannot be a mapped port and vice versa
• A mobile port cannot be configured as a network port of a mapping session
STP convergence time (flat, 1x1, 802.1s) 30 sec
802.1w rapid reconfiguration Less than 1 sec
Learned MAC addresses per port Up to 16 K MAC Addresses is supported
Learned MAC addresses per system Up to 64 K MAC Addresses is supported
In synchronized mode (default), up to 16K MAC Addresses is supported per system
In Distributed mode, up to 64 K MAC Addresses is supported per system (no more than 16K per NI).
Layer-2 forwarding on Ethernet ports Wire-speed (64 Bytes packets)
Layer-2 forwarding GigE, known MAC Wire-speed (64 Bytes packets)
Broadcast per Ingress port Programmable
Loopback Interface The loop-back interface allows you to uniquely identify a router in the network with one IP address.
The advantage of the loop-back interface is to be independent of the physical ip interfaces. In a
redundant routing network, the loop-back interface is always accessible when routing topology
changes or ip interfaces go down.
The main advantage of Loop-back interface is a more reliable Network Management path through
OmniVista or an NMS station.
Also, you can use the loop-back interface to uniquely identify the router within OSPF and BGP if you
set the router-id to the same as the loop-back address.
The loop-back can also be used for the RP (Rendezvous Point) in PIM-SM.
The loop-back address is also used for the sFlow Agent IP address.
The Loopback address is used for source IP of RADIUS authentication.
User Definable Loopback Interface Loopback0 is the name assigned to an IP interface to identify a consistent address for network
management purposes (including SNMP/sFlow datagrams). The Loopback0 interface is not bound to
any VLAN; therefore it always remains operationally active. This differs from other IP interfaces, such
that if there are no active ports in the VLAN, all IP interfaces associated with that VLAN are not
active. In addition, the Loopback0 interface provides a unique IP address for the switch that is easily
identifiable to network management applications.
Sever Load Balancing (SLB) There are 2 kind of server clusters:
-Server Farm: The traffic is truly destined to the Server Farm and the destination IP is the Virtual IP of
the Server Farm. Each server is also configured with a Loopback Interface for the Virtual IP
-Advanced Clustering: the traffic is not necessarily destined to a Virtual IP, instead, it is matching a
user defined QoS condition, allowing L1-L4 classification. The most common application is Firewall
clustering where packets are load balanced to several firewall for inspection and sent back if accepted.
The following values are the tested limits with the functionality verified (stress test):
Tested limit of clusters (on a per switch basis) is 16.
Tested limit of servers (on a per cluster basis) is 16.
Tested limit of Probes: 20 Probes
The following values are the maximum limits enforced by the Code:
Maximum number of clusters: 16
Maximum number of physical servers: 75
Maximum number of probes on a switch: 20
Sever Load Balancing (SLB) Health monitoring is performed by the CPU of the Primary Management.
LOAD BALANCING HASHING
In both “VIP” and “Condition” SLB, the traffic is balanced among the servers using an hash algorithm
based on IPSA and IPDA.
Internally, each active server is seen as a host ECMP route to reach the cluster.
Therefore, the load balancing is the same than the ECMP load balancing.
Layer-3 Routing Unicast (IPv4)
Large L3 table support
Hardware:
• Maximum number of active flows in the hardware: 12K
One active flow is usually one “remote-subnet” flow (not a per destination ip flow based)
Now with the ARP table enhancement, one active flow can also be a “host routed” flow
The table is shared for
- IPV4 active flow (remote ipv4 network): 1 entry
- IPV6 active flow (remote ipv6 network): 2 entries
- Host active flow (ARP entry): 1 entry
• Maximum number of active “ARP entries” flows: 8K
• Maximum number of ECMP Next-hops that can be stored: 512
Software: