|
Standard |
Distributed |
|
- Provides virtual switch configuration for single host - Management Plane and Data Plane both runs on ESXi Host - Boundary of Standard switch is limited to ESXi Host. - Supports Trunk Port and Access ports. - If you want to use trunk, then you must allow from 0-4095. You do not have option to trunk specific VLANs. - Disadvantage: Scalability is the problem in standard switch. If we need to increase the number of hosts, we must create new switch configuration for each ESXi hosts. - Supports Layer 2 switch, VLAN segmentation, IPv6, 802.1Q tagging, NIC teaming, Outbound traffic shaping. - You can configure policies at switch level and port group level. |
- Provides virtual switch configuration for all hosts in a vCenter Server. - Segregation of Management and Data Plane. - Management Plane on vCenter and Data Plan on ESXi Host. - Boundary of vDS is Logical Datacenter - Supports Trunk Ports and Access ports - Supports up to 2000 ESXi host to the same distributed switch. - Trunk is supported on vDS (i;e 101, 103, 200) - Supports NIOC, Layer 2 Switch, VLAN Segmentation, IPv6, load based teaming, per-port policy, NetFlow, Port Mirroring, Private VLANs, VM network port block, Inbound traffic shaping, Outbound traffic shaping, NIC teaming, 802.1Q tagging. - You can configure policies at each per port level.
|
Notes: -
When port belongs to one L2 network than it’s Access port.
Group of VLANs, more than one L2 network. If port belongs to more than one L2 network, than its trunk port.
When you are using native VLAN
- Cisco: 1
- VMware: 0
0-4095 (0 is native, and you can use VLAN upto 1-4094)
Tagging is done through the Guest OS.
Tagging is done by kernel. Guest OS is not doing the VLAN tagging. Virtual switch is doing tagging, that’s why it is known as Virtual Switch tagging.
When frame received by the hypervisor
VLAN is only configured to external switch. Complete virtual switch is on one VLAN. But when the traffic will go out. In this scenario, any of device is on VLAN 10(example) then only it will be communicated. If it’s on VLAN 11, then it will not be communicated. Because the entire virtual switch is on VLAN 10.
Physical switch does the tagging.
Tagging done on L2 frames.
|
|
|
12 Bits |
- Provides connectivity between Virtual and Physical environment.
- One VMNIC can be associated with one vSwitch.
- For Guest – it’s vnic
- For Host – It’s vmnic
- Vmnic is also work as uplink.
- We can have multiple VLANs on one vSwitch.
- But we cannot have multiple VLAN on one Port group.
- NIC teaming of multiple vmnic
- Same L2 swtich for multiple broadcast domains.
- Reduce number of physical L2 switches.
Allocate bandwidth during the bandwidth contention. Below ports does exist where we can configure NIOC.
- Management
- NFS
- iSCSI
- vSAN
- vSphere vMotion
- Fault Tolerance
- vSphere Replication
- vSphere Data Protection
- VTEP
Scalability Problems with Distributed Switch
End to End L2 Physical Network
- Feature of dvSwitch.
- When teaming will be based on the load exist on physical NIC card. Your ESXi host check the workload on NIC in every 30 secs. If load is more than 75% on NIC1, then it will move some traffic from NIC1 to NIC2.
When you have connected more than one Physical NIC to the switch then Load Balancing will come into picture.
We have multiple options in Load Balancing.
- Load avoidance policy not a load balancing policy.
- Each virtual port has IDs.
- We connected 2 NICs with this switch.
- ID of the port will be associated with the uplink.
- Port ID will be associated with orginal port.
- In case we shutdown VM, port are useless.
- VM will power on in same port ID.
- Generate MAC hash value
- Hypervisor will calculate hash value on vnic MAC.
- Has value will be associated with uplink.
- VM can maximum consume the bandwidth of single physical NIC and associates with the vSwitch.
- If VM has two NIC then it can use of two NICs.
- Configure EtherChannel (for Cisco Switch) or LACP (for Non-Cisco Switch) at physical switch
- Works on Source IP and Desitation IP.
- Additional workload on hypervisor to get source IP and destination.
- Need LACP at(EtherChannel) because destination is different always but source is same. So need LACP.
- Consider all physical port as logical port.
- When you manually configure the failover order, then we use Explicit failover.
