Spanning Tree, Link Aggregation , VLAN and First Hop Redundancy protocols are used in Campus, Service Provider Access and Aggregation and in the Datacenter environment. There are definitely other protocols which are common across the Places in the Networks but in order to keep this article short and meaningful I choose these four.
I will describe Spanning tree, link aggregation, 802.1q Vlan and First hop redundancy protocols at a high level since I will explained them in detail later in the separate articles.
For the more advanced layer 2 protocol information check this article.
Spanning Tree – IEEE 802.1d, 802.1w, 802.1s
Spanning tree is used to build a control path between the Ethernet switches in the campus , service provider and data center environment. It prevents data plane loops by creating a tree !
Loop preventation is very crirical for the Ethernet since there is no TTL value or any other loop mitigation mechanism encoded in the Ethernet header.
Loop prevention is achieved by blocking the link which has a higher cost to the root switch in the topology.
802.1d is also known as original spanning tree or legacy spanning tree has been extended by Cisco and named as PVST+.
PVST+ works in a similar way with 802.1d, only the link is blocked for each Vlans separately. Topology can be adjusted by placement of root switch.
Thus PVST+ provides flexibility compare to 802.1d , since with 802.1d legacy spanning tree, only one tree can be created and all the Vlans are mapped to that tree.
Redundancy can be provided as active/standby with 801.2 legacy spanning tree but with PVST+ which is Cisco’s implementation of 802.1d ; links can be used simultaneously by having Vlan based load balancing.
Even and Odd Vlan root switch placement at the distribution layer is very common deployment method and provides deterministic topology thus troubleshooting can be easier. But if the number of Vlans are high, configuration can be quite complex.
Rapid Spanning Tree is the IEEE work which has several enhancements to original 802.1d and called 802.1w. Rapid Spanning Tree is used for fast convergence commonly and author recommends it strongly for the Campus Networks.
MST which is 802.1s is the third implementation of spanning tree and use case is data center and service provider access networks which have thousands of Vlans and require large scale bridging support.
Down side of spanning tree is to block the links for any given Vlan and root switch placement for odd/even vlan is complex to manage. Alternative approach for the Campus Networks is Link Aggregation or Multi Chassis Link Aggregation which is the subject of next concept.
There are other drawbacks of spanning tree such as suboptimal traffic flow, lack of multpathing and so on and can be avoided by using other protocols such as SPB, TRILL, Fabricpath. You can find more information about large scale bridging from here.
To overcome to drawback of spanning tree which is the blocking links, link aggregation can be used between the switches. Link aggregation is used for both load sharing and redundancy.
At least two links is placed in a bundle and load sharing and depends on vendor implementation 4, 8, 16, 32 links can be placed in a bundle. Link aggregation might be used between the access-aggregation/distribution layers and distribution to core layers depends on capacity requirements.
Link aggregation works between two switches and if it will be run between three or more switches then it is called as multi chassis link aggregation.
Aggregation/distribution layer devices synchronize the state information through over the inter switch links. Over the inter switch link synchronization can be achieved with ICCP (Inter Chassis Communication protocol) or with vendor specific protocols.
VSS and VPC are the two technologies which are known by the network engineers who have experience on Cisco switches use multi chassis link aggregation to bundle two physical chassis into one logical device.
VLAN – Virtual Local Area Network / 802.1q
With adding another tag on the Ethernet frame, virtualization and segmentation can be achieved.
Similar to VSAN technology of Cisco in the storage area network, IEEE 802.1q adds a 4 byte header to the Ethernet frame and addresses the segmentation and multi tenancy in the campus,service provider and the datacenter networks.
VLAN concept is used to address either for broadcast/multicast or security concern. Broadcast, unknown unicast and multicast packets which also known as BUM traffic is flood everywhere in the network.
By separating the network with VLANs, BUM traffic is only flooded in that particular VLAN. What happens in Vegas stay in Vegas!.
Since the NICs of PCs and servers has to process every broadcast packets , limiting these traffic by diving users into separate VLAN is very common practice and recommended by the Author.
Another use case of VLAN is security. Inter VLAN traffic might be passed through the security devices such as Firewalls, IDS/IPS and unwanted communication or even zero day attacks can be prevented.
First Hop Redundancy Protocols – HSRP, VRRP, GLBP
If the requirement is to use the same Vlan on the many access switches, then layer 2 access design is used at the campus network.
Layer 2 access design will be explained later in the article. In the layer 2 access design , distribution layer switches is used as the default gateway for the end host.
For high availability purpose using at least two distribution switches is the best practice.
Question is which distribution layer switches will be the gateway of the hosts for the particular Vlan?.
There are three approaches for the first hop gateway redundancy.
HSRP ( Hot standby router protocol ) , VRRP ( Virtual router redundancy protocol ) and GLBP ( Gateway load balancing protocol ) .
HSRP is the Cisco propriety protocol and works as an active standby fashion. For a particular Vlan , only one of the switches will be active and the other switch will take the responsibility if the active device fails.
VRRP is the IETF’s answer to the problem and works in a similar way with HSRP. Common deployment method for HSRP and VRRP is to use one switch as default gateway for set of Vlans and the other switch for the remaining Vlans.
Depends on the number of Vlans , this approach might be complex from the configuration point of view.
GLBP is the Cisco propriety default gateway redundancy protocol as well. GLBP provides a redundancy for the hosts within the same Vlan. Two switches actively forward the traffic for the hosts in the same Vlan.
If one switch fails only half of the hosts on that Vlan is affected. With HSRP and VRRP, if active switch fails , all the hosts on that Vlan is affected from the failure.