Dell Precision R5400 Remote Access Device: Networking Considerations - Page 10

VLAN and QoS Considerations

Page 10 highlights

Rack workstations -> rack workstation switch links Each rack workstation receives two Gigabit Ethernet connections to the rack workstation switch, one for the PC and one for the PCoIP traffic. Planning Bandwidth for each PCoIP link is: 60 * 1.1 = 66 Mbps Rack workstation switch -> datacenter switch links Each rack workstation switch has a single Gigabit Ethernet connection to each datacenter switch. Planning Bandwidth for the PCoIP traffic on the link is: 10* 60 * 1.1 = 660 Mbps, leaving 340 Mbps for PC traffic. Datacenter switch -> workgroup switch links There are three switch configurations: A: 2 switches supporting 36 knowledge workers and 5 performance users B: 1 switch supporting 40 task workers and 8 knowledge workers C: 1 switch supporting 40 knowledge workers and 8 task workers The datacenter to workgroup switch links would then need to be sized: A: (36 * 15 + 5 * 60) * 1.1 = 924 Mbps B: (40 * 3 + 8 * 15) * 1.1 = 264 Mbps C: (8 * 3 + 40 * 15) * 1.1 = 686 Mbps In each case, less than 1 Gbps is required, so a single Gigabit uplink to the datacenter switches is sufficient. Workgroup switch -> Portal links In all cases, the planning bandwidth for each user is less than the 100 Mbps available for each link, so the standard 100-base-T connections are sufficient. Datacenter switches In this network configuration, only a single datacenter switch would be required to meet all the traffic needs. To allow for network fault tolerance, two datacenter switches are connected to the network. Since each connection to a datacenter switch requires less than the available Gigabit link rates, a simple active/passive redundancy scheme is acceptable, where one switch takes all of the network traffic unless there is a failover. Where required, active/active redundancy can be implemented, using protocols such as Multiple Spanning Tree Protocol (MSTP), Cisco VSS, or Nortel SMLT. Using these protocols is transparent to the PCoIP connections. Each switch would have an identical configuration. The final per-switch configuration would be: • 1 Gigabit Ethernet port between the two switches, to allow for setup and failover communication • 12 Gigabit Ethernet ports to the blade chassis • 1 Gigabit Ethernet port to the workstation rack • 4 Gigabit Ethernet ports to the workgroup switches • Plus a number of ports for other server and internet connections VLAN and QoS Considerations PCoIP technology was developed to work over standard IP networks without requiring special configurations. As such, virtual LAN (VLAN) and Quality of Service (QoS) settings are not required for system operation. In cases where the network is shared between PCoIP traffic and other data traffic, and the IT administrator wishes to guarantee a certain amount of bandwidth to PCoIP traffic, VLAN and QoS settings can be used within the network switches. TER0806005 Issue 1 10

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13

TER0806005 Issue 1
10
Rack workstations -> rack workstation switch links
Each rack workstation receives two Gigabit Ethernet connections to the rack workstation switch, one for the PC
and one for the PCoIP traffic.
Planning Bandwidth for each PCoIP link is: 60 * 1.1 = 66 Mbps
Rack workstation switch -> datacenter switch links
Each rack workstation switch has a single Gigabit Ethernet connection to each datacenter switch.
Planning Bandwidth for the PCoIP traffic on the link is: 10* 60 * 1.1 = 660 Mbps, leaving 340 Mbps for PC
traffic.
Datacenter switch -> workgroup switch links
There are three switch configurations:
A: 2 switches supporting 36 knowledge workers and 5 performance users
B: 1 switch supporting 40 task workers and 8 knowledge workers
C: 1 switch supporting 40 knowledge workers and 8 task workers
The datacenter to workgroup switch links would then need to be sized:
A: (36 * 15 + 5 * 60) * 1.1 = 924 Mbps
B: (40 * 3 + 8 * 15) * 1.1 = 264 Mbps
C: (8 * 3 + 40 * 15) * 1.1 = 686 Mbps
In each case, less than 1 Gbps is required, so a single Gigabit uplink to the datacenter switches is sufficient.
Workgroup switch -> Portal links
In all cases, the planning bandwidth for each user is less than the 100 Mbps available for each link, so the
standard 100-base-T connections are sufficient.
Datacenter switches
In this network configuration, only a single datacenter switch would be required to meet all the traffic needs. To
allow for network fault tolerance, two datacenter switches are connected to the network.
Since each connection to a datacenter switch requires less than the available Gigabit link rates, a simple
active/passive redundancy scheme is acceptable, where one switch takes all of the network traffic unless there
is a failover. Where required, active/active redundancy can be implemented, using protocols such as Multiple
Spanning Tree Protocol (MSTP), Cisco VSS, or Nortel SMLT. Using these protocols is transparent to the PCoIP
connections.
Each switch would have an identical configuration. The final per-switch configuration would be:
1 Gigabit Ethernet port between the two switches, to allow for setup and failover communication
12 Gigabit Ethernet ports to the blade chassis
1 Gigabit Ethernet port to the workstation rack
4 Gigabit Ethernet ports to the workgroup switches
Plus a number of ports for other server and internet connections
VLAN and QoS Considerations
PCoIP technology was developed to work over standard IP networks without requiring special configurations.
As such, virtual LAN (VLAN) and Quality of Service (QoS) settings are not required for system operation. In
cases where the network is shared between PCoIP traffic and other data traffic, and the IT administrator wishes
to guarantee a certain amount of bandwidth to PCoIP traffic, VLAN and QoS settings can be used within the
network switches.