Thursday, October 17, 2013

10-17-2013 - VLT - Virtual Link Trunking - Data Center


Virtual link trunking (VLT) allows physical links between two chassis to appear as a single virtual link to the network core or other switches such as Edge, Access or Top of Rack (ToR). VLT reduces the role of Spanning Tree protocols by allowing LAG terminations on two separate distribution or core switches, and by supporting a loop free topology. (A Spanning Tree protocol is still needed to prevent the initial loop that may occur prior to VLT being established. After VLT is established, RSTP may be used to prevent loops from forming with new links that are incorrectly connected and outside the VLT domain.) VLT provides Layer 2 multipathing, creating redundancy through increased bandwidth, enabling multiple parallel paths between nodes and load-balancing traffic where alternative paths exist.

1. Virtual link trunking offers the following benefits:

2. Allows a single device to use a LAG across two upstream devices

3. Eliminates Spanning Tree protocol (STP) - blocked ports

4. Provides a loop-free topology

5. Uses all available uplink bandwidth

6. Provides fast convergence if either the link or a device fails

7. Optimized forwarding with Virtual Router Redundancy Protocol (VRRP)

8. Provides link-level resiliency

9. Assures high availability

*** Dell does not recommend enabling stacking and VLT simoultaneously!!!

VLT Terminology

The following are key VLT terms.

•    Virtual link trunk (VLT) — The combined  port channel  between  an attached  device and the VLT peer switches.

•    VLT backup link — The backup link monitors  the health of VLT peer switches. The backup link sends configurable, periodic keep alive messages between VLT peer switches.

•    VLT interconnect (VLTi) — The link used to synchronize  states between  the VLT peer switches. Both ends must be on 10 Gb or 40 Gb interfaces.

•    VLT domain — This domain includes  both VLT peer devices, the VLT interconnect, and all of the port channels in the VLT connected to the attached devices. It is also associated to the configuration mode that must be used to assign VLT global parameters.

•    VLT peer device — One of a pair of devices that are connected with the special port channel known as the VLT
interconnect  (VLTi).

VLT peer switches have independent management planes. A VLT interconnect between the VLT chassis maintains synchronization of Layer 2 and Layer 3 control planes across the two VLT peer switches. The VLT interconnect uses either 10 Gb or 40 Gb ports on the switch.

A separate backup link maintains heartbeat messages across an out-of-band (OOB) management network. The backup link ensures that node failure conditions are correctly detected and are not confused with failures of the VLT interconnect. VLT ensures that local traffic on a chassis does not traverse the VLTi and takes the shortest path to the destination via directly attached links.

VLT Fabric Terminology

The following terms are unique to the design and deployment of a Layer 2 VLT fabric.

Leaf — A switch that connects switch, servers, storage devices, or top-of-rack (TOR) elements. The role of the leaf switch is to provide access to the fabric. The leaf switch connects to all of spines above it in the fabric.

Spine — A switch that connects to leaf switches. The role of the spine is to provide an interconnect to all the leaf switches. All the ports on the spine switches are used to connect the leaves, various racks together. The spines provides load balancing and redundancy in the distributed core. There are no uplinks on the spines.

Edge ports — The uplinks on the spines and downlinks on the leaves.

Uplinks — An edge port link on the first two spines in the VLT fabric that connects to outside the fabric.

Downlinks — An edge port link that connects the leaves to the data access layer. For example, servers or ToR elements. You must specify an even number of uplinks. The minimum number of uplinks is 2. One uplink is for redundancy.

Fabric Interlinks — Links that connect the spines to the leaves. The fabric interlink bandwidth is fixed: 10 Gb or 40 Gb.

VLT Components


Typical VLT TopologyThe VLT domain has VLTi (ICL) links connecting between VLT peers and VLT port-channels connecting to a single access switch, to a switch stack, a server supporting LACP on its NIC, or to another VLT domain as shown in the following illustration. The backup-link connected through the out-of-band (OOB) management network. Some hosts can connect through the non-VLT ports.

VLT configuration requires that you enable the feature and then configure the VLT domain, backup link, and VLT interconnect on both peer switches.

Configuration Notes
1. VLT port channel interfaces must be switch ports.
2. If RSTP is included on the system, it must be configured before VLT.
3. Dell Force10 strongly recommends that the VLTi (VLT interconnect) must be a static LAG and that LACP should be disabled on the VLTi.
4. If both VLT peers are rebooted in JumpStart mode and VLT LAGs are static, the DHCP server reply to the DHCP discover offer may not be forwarded by the ToR to the correct node. To avoid this scenario, configure the VLT LAGs to the ToR and the ToR port channel to the VLT peers with LACP. If supported by the ToR, enable the lacp-ungroup feature on the ToR using the command lacp ungroup member-independent port-channel.
5. If the lacp-ungroup feature is not supported on the ToR, VLT peers should be rebooted one at a time. After rebooting, verify that VLTi (ICL) is active before attempting DHCP connectivity.
6. When IGMP snooping is enabled on the VLT peers, ensure the value of the delay-restore command is not less than the query interval.
7. When Layer 3 routing protocols are enabled on VLT peers, make sure the delay-restore timer is set to a value that allows sufficient time for all routes to establish adjacency and exchange all the L3 routes between the VLT peers before the VLT ports are enabled.
8. The lacp ungroup member-independent command should only be used if the system connects to nodes using BMP to upgrade or boot from the network.
9. Ensure all port channels where LACP ungroup is applicable are configured as hybrid ports and as untagged members of a VLAN. BMP uses untagged DHCP packets to communicate with the DHCP server.
10. If the DHCP server is located on the ToR and the VLTi (ICL) is down due to a failed link when a VLT node is rebooted in JumpStart mode, it will not be able to reach the DHCP server, resulting in BMP failure.

No comments: