FAQ

12. What are redundant top-of-rack topologies and
associated benefits?
For midsize server installations, two half-width M4300 10GbE
models can be paired in a single rack space for redundant top-of-rack.
Compared with single top-of-rack switch installation, such two-unit
horizontal stacking is cost-eective yet highly ecient for HA.
•Management unit hitless failover and nonstop forwarding ensures no
single point of failure for servers and storage equipment
•All devices can connect to both redundant top-of-rack switches
using link aggregation (L2/L3/L4 LACP) with load-balancing and
failover
•Variety of 10 Gigabit copper and fiber ports - all backward
compatible with 1G speeds - enable any type of virtualization
•iSCSI Flow Acceleration and Automatic Protection / QoS enhance
server room networks containing iSCSI initiators and iSCSI targets
•Any 10 Gigabit copper and fiber ports can be used for stacking,
depending on inter-switch links oversubscription requirements
•Within the stack, a switch is elected (or chosen based on priority
settings) as the “management unit”
•The other switch is designated (or chosen based on priority settings)
as an alternate, backup management unit
•The Nonstop Forwarding (NSF) feature enables the stack to secure
forwarding server and storage trac when the management unit
fails:
- Power failure of the management unit
- Other hardware failure causing the management unit to hang or to
reset
- Soware failure causing the management unit to hang or to reset
- Failover initiated by the administrator
- Loss of cascade connectivity between the management unit and
the backup unit
•Instant failover from management unit to redundant management
unit is hitless for the servers and storage equipment connecting both
•As the backup management unit takes over, data streams may lose
a few packets, but do not lose their IP sessions, such as iSCSI, NFS,
CIFS etc...
•Other lower end solutions are causing service interruptions across
the stack for servers and storage without NSF and hitless failover
•Back to normal production conditions, hitless failback requires a
command in CLI or in GUI, for more control
•Hitless failback is automatic in case of new management unit
(triggered or accidental) failure
13. What are spine and leaf topologies and associated
benefits?
For typical collapsed core installations, with a variety of 1G and
10G access ports in branch oces, server rooms or campus high
performance labs: M4300 10G models can stack with M4300 1G
models, enabling innovative “spine and leaf” topologies (other ring
topologies are also possible).
•Spine and leaf architectures deliver highest performance with every
leaf switch (1G) connecting to every spine switch (10G)
•Every 1G “leaf” access switch connects to both 10G “spine”
distribution switches
•Any 10G port (copper, fiber) and any media type (RJ45, SFP+,
DAC) can be used for stacking on any M4300 model
•On 1G models, up to (4) 10G ports per switch can be used for
stacking, hence allowing for line-rate aggregation to their spine
•On 10G models, up to (16) 10G ports per switch can be used for
stacking, depending on inter-switch links oversubscription requirements
•Up to (8) M4300 switches can be aggregated using a virtual
backplane and a single console or web management interface
•While reducing the number of logical units to manage, stacking also
brings network resiliency with distributed uplinks in aggregation to
main core
•In this architecture, both 10G “spine” switches connect to main core
using 10G LACP link aggregation
•Using adequate priorities in the stack, both 10G “spine” switches are
meant to handle “management unit” and “backup management unit”
roles
•This way, management unit hitless failover and nonstop forwarding
ensures no single point of failure:
- Nonstop Forwarding (NSF) enables the stack to secure forwarding
end-user trac on all other switches when the management unit
fails
- Instant failover from management unit to backup management
unit is hitless for the rest of the stack
- Since both the management unit and the backup unit connect
to the core using distributed link aggregation (LACP), there is no
possible service interruption while the backup management unit
takes over
- All other switches in the stack keep forwarding L2 and L3 trac in
and out, while backup unit guarantees connectivity to the core
•Back to normal production conditions, hitless failback requires a
command in CLI or in GUI, for more control
•Hitless failback is automatic in case of new management unit
(triggered or accidental) failure
•M4300 Virtual Chassis stacking technology upsurges overall
network availability, providing better resiliency in network
architectures, and better performance with advanced load balancing
capabilities between network uplinks
14. What is active-active teaming?
Any server, storage equipment or any type of host can connect to
two dierent M4300 switches in a stack, using simple LACP Ethernet
channeling. A server for instance, will use its two network cards in
active-active teaming mode instead of standard failover mode. It
means more performance, more bandwidth for the server for same
level of redundancy: in case of a switch failure, nonstop forwarding
architecture ensures no service interruption for the other switches
in the stack. For the server, it means the trac will continue to flow
through the other link without noticeable eect.