Description

Your administrator is able to hard code the interface speeds as shown in vSRX: 

[edit]
root# set interfaces ge-0/0/0 speed ?
Possible completions:
  100m                 
  10m                  
  1g                   
[edit]

However, this configuration change does not affect the actual traffic throughput via the specific interface. 

This article explains the above behavior by demonstrating what happens in a virtualized environment. 

Symptoms

The above behavior is demonstrated with the help of the following example:

Topology

Client ==== ge-0/0/0 -- vSRX -- ge-0/0/1 ==== Server

Scenario 1

  • The interface speeds are left at default: 

root> show interfaces ge-0/0/0 | grep speed                        
  Speed: 10Gbps, BPDU Error: None, Loop Detect PDU Error: None,

root> show interfaces ge-0/0/1 | grep speed                        
  Speed: 10Gbps, BPDU Error: None, Loop Detect PDU Error: None,

root> show configuration interfaces | display set | match speed    
root>

  • The IPerf throughput result from the client to the server via vSRX is as follows: 

[ ID] Interval           Transfer     Bitrate         Retr
[  5]   0.00-3.00   sec   293 MBytes   819 Mbits/sec   38             sender
[  5]   0.00-3.04   sec   291 MBytes   802 Mbits/sec                  receiver

Scenario 2

  • The interface speeds of both ge-0/0/0 and ge-0/0/1 interfaces are hard coded to 100M: 

root> show interfaces ge-0/0/0 | grep speed 
  Speed: 100mbps, BPDU Error: None, Loop Detect PDU Error: None,

root> show interfaces ge-0/0/1 | grep speed    
  Speed: 100mbps, BPDU Error: None, Loop Detect PDU Error: None,

root> show configuration interfaces | display set | match speed 
set interfaces ge-0/0/0 speed 100m
set interfaces ge-0/0/1 speed 100m

  • The IPerf throughput result from the client to the server via vSRX is as follows: 

[ ID] Interval           Transfer     Bitrate         Retr
[  5]   0.00-3.00   sec   316 MBytes   884 Mbits/sec   34             sender
[  5]   0.00-3.04   sec   315 MBytes   868 Mbits/sec                  receiver

Scenario 3

  • The interface speeds of both ge-0/0/0 and ge-0/0/1 interfaces are hard coded to 10M: 

root> show interfaces ge-0/0/0 | grep speed                        
  Speed: 10mbps, BPDU Error: None, Loop Detect PDU Error: None,
root> show interfaces ge-0/0/1 | grep speed    
  Speed: 10mbps, BPDU Error: None, Loop Detect PDU Error: None,
root> show configuration interfaces | display set | match speed    
set interfaces ge-0/0/0 speed 10m
set interfaces ge-0/0/1 speed 10m

  • The IPerf throughput result from the client to the server via vSRX is as follows: 

[ ID] Interval           Transfer     Bitrate         Retr
[  5]   0.00-3.00   sec   320 MBytes   895 Mbits/sec  162             sender
[  5]   0.00-3.04   sec   319 MBytes   879 Mbits/sec                  receiver

You can see that the actual traffic throughput via vSRX is unaffected by changes to the interface speed configuration as demonstrated.

Solution

This is expected behavior. 

In a virtualized environment, the Hypervisor will not allow the VM to set the Physical Layer properties of a vNIC. 

Certain features that a VM cannot control in a virtualized environment include:

  • Auto-negotiation

  • Speed/Duplex/Link settings

  • Actual speed of operation, which depends on the CPU cycles available and the speed of the uplinks

In short, a VM (vSRX in this case) cannot control the physical properties of a vNIC that the Hypervisor is assigning to it. 

If you want to perform bandwidth policing on vSRX, use the COS / Traffic Shaping features instead.

Related Information