XGS3700 High Available

Redundancy Design with Stacking,

This document introduces how to deploy a high available redundancy network in a SMB company by using the
ZyXEL XGS3700 series and GS2210 series switches. This document also brings the basic concept of the XGS3700’s
stacking, RSTP and Link Aggregation. It also focuses on two main purposes: 1. High redundancy for a failed single
link path. 2. High redundancy for XGS3700 stacking when a single XGS3700 goes down.

Modules:

1. Introducing the XGS3700’s stacking.

1. Introducing the XGS3700’s stacking.

The XGS3700 series support hardware stacking to bundle 2 to 8 XGS3700 series switches into a single logic
switch. Users are able to manage multiple stacked XGS3700 series switches by using a single IP address.

Stacking role:

- Master: A switch in a stack from which users can manage each stacked switch via
console/telnet/SSH/WebGU
- Backup: A switch that takes over the role of the master when the master in the stacking system is down
- Linecard: The rest of active switches

Stacking Channel:

When the XGS3700 series switch is operating in Stacking mode, the last 2 of 10G switch ports are reserved for
stacking.
XGS3700-24/24HP, port 27 and port 28 are reserved for stacking channel.
XGS3700-48/48HP, port 51 and port 52 are reserved for stacking channel.

Stacking Topology

- Daisy Chain: When a single stacking link fails, the oringinal stacking system will be separated into 2
stacking systems
- Ring (recommend): This provides stacking link redundancy when a single stacking link fails
Stacking benefit:

- Centralized management: Users can manage the stacking system by configuring the stacking master and
synchorinze the configuration on all active slots (stacked switches)
- Redundancy: This allows data and management redundancy when the stacking master goes down
- Hot Swappable: Users can add/remove/replace a stacking slot without shutting down the system
- More available switch port: Users will have more available switch ports

The correct steps to enable the XGS3700 series stacking mode:

1. Restore the switch to its factory default settings by using a login profile with privilege level 14 (e.g.
admin)

2. Log in to the XGS3700 with the default login profile (admin///1234) and configure its operating mode as
stacking mode.

3. Now the XGS3700 is operating in stacking mode.

4. Select the stacking master. There are two ways to decide a stacking system’s master.
- Stacking priority (0-63): the switch which has the highest priority will be the stacking master.
- Force master: you can ignore stacking priority setting and force a switch to be the stacking system
master directly.

5. Connect the rest slot(s) and start to configure the stacking system.
 2. Introducing Rapid Spanning Tree Protocol

In an early Ethernet network, IEEE 802.1d STP is used to provide a loop-free topology by blocking switch ports
when the network topology is a ring. There are some dis-advantages of STP. For example, a switch port needs to
move through multiple port states to converge the network and there are less switch port roles.

IEEE 802.1s RSTP improves STP's operation by simplifying switch port states and allowing faster convergence. It
also adds more switch port roles, such as alternative port, backup port and edge port.

RSTP port status

- Discarding: A switch port does not record MAC addresses and discard packets
- Learning: A switch port does not forward packets but building a MAC address table
- Forwarding: A switch port forwards packets

RSTP port roles

- Root port: A switch port which has the lowest cost to a root bridge from a non-root bridge

- Designated port: A forwarding port

- Backup port: A substitute switch port for DP and port status is discarding

- Edge port: A switch port exists at the edge of a spanning tree, which connects an end device but not a
switch. The Edge port does not join spanning tree process and move to forwarding state directly while
connecting to an end device. When the edge port receives BPDUs, it will join spanning tree process.

Fast convergence

Unlike STP's longer delay due to STP port state change and slow convergence, RSTP converges rapidly when
network topology changes occur. The main enhancement is that: RSTP changes switch port states by negotiating
with neighbor devices actively but not relies on timers. So a port can go to discarding states without the need to
wait for forwarding delay twice.

Topology Change

When a bridge receives BPDUs with Topology change, it will clear its MAC address table and re-learn MAC
addresses. It also sends BPDUs with Topology Change from RP and DP.
 3. Introducing Link Aggregation.

Users are able to bundle 2 to 8 switch ports into a single virtual link for bandwidth enlargement, link redundancy
and load sharing. Currently the ZyXEL Ethernet switches support 2 types of link aggregation.

- Static trunk: The switch is able to negotiate its port speed and duplex with neighbor devices.
- LACP: IEEE 802.3ad standard, which requires neighbor devices to run LACP. The switch is able to negotiate its
port speed and status with neighbor devices. If the neighbor device does not run LACP, the switch will block the
connected switch port itself.

4. Introducing High Available Redundancy Network Design

Nowadays, customers pay their attention on network performance and focus on redundancy design, so that
redundancy path and devices are able to perform service recovery in a short time when netowrk issues occur.
Link Aggregation cannot load-balance traffic across switches which do not support stacking, therefore it’s not
possible to do link redundancy between servers and these switches. With the XGS3700 stacking feature, users
are able to configure LACP across multiple switches for path redundancy.
When one of the links fails or one of stacking switches goes down, other stacking members can recover the
service in a short time.

Below is a simple example to demonstrate how to set up a stacking redundancy network by using RSTP and Link
Aggregation. Hosts should connect the GS2210 and configure the XGS3700-2 and XGS3700-3 as the default-
gateway of the distribution layer. The physical link between the GS2210 and the XGS3700-2 and the XGS3700-3
should be connected. Link aggregation is also enabled and configured to bundle switch ports. When one single
link fails, another path will be able to forward traffics.

The XGS3700-1 stacking switches play the root bridge role of RSTP, and the speed of link between the XGS3700
stacking switches is 10G. The physical link between the XGS3700 stacking switches should be a ring topology and
use RSTP for path redundancy.

The 3* XGS3700 stacking switches should remain the same interface for RSTP. When one of 10G links fails
between the XGS3700 stacking switches, RSTP will open the alternate port and forward packets. The XGS3700-2
and XGS3700-3 should have a default-route to the XGS3700-1. The XGS3700-1 needs to add a static routing
entry and set the XGS3700-2 or XGS3700-3 as the next hop.
Recovery time test

Test method: Perform a Ping test to continually send ICMP from host 1 to test a device.
Test items: 1. Remove a single link between the GS2210 and the XGS3700-2 in stacking mode.
2. Remove the 10G link between the XGS3700-1 and the XGS3700-2 in stacking mode.
3. Power off one of the XGS3700-2 stacking master.

Result
If you remove a single link between the GS2210 and the XGS3700-2, the data forwarding recovery time is 1-2
seconds. When the link between the XGS3700-2 and the XGS3700-1 fails, the data forwarding recovery time is 1-
2 seconds. When the XGS3700-2 master goes down, the data forwarding recovery time is 4-7 seconds. Please
note that the time value varies depending on how many features the switch has been configured to use. If you
configure many features, it might require more than 7 seconds to recover data forwarding.