Wednesday, March 4, 2020

End to end network latency

When it comes to testing connectivity and latency, I've noticed that many IT technicians don't seem to have any tools in their skill set that go beyond a ping. While that works in many situations, there are often situations where ICMP traffic (including ping) is blocked. At this point, connectivity testing skill set often falls down to a telnet command to the port to see if its open, instead of using many already available tools like the powershell test-netconnection cmdlet. Unfortunately, that commandlet and telnet only show that you can connect to a port, and it doesn't tell you how long it takes to get to it. If you're on a windows machine, you can use the Test-PortLatency function that I've written below. This will give a rough idea of the time to connect in milliseconds to a remote tcp port. If you're on a linux machine, the nmap suite of tools has several programs that give latency information, like nping, or just nmap itself. There are other options as well, and typically some programming languages like python or perl available, which should have some capability to create a simple script to provide this information.

While these tools provide latency from one source point to another, you may find that you need to run tests from multiple points. Connections can get complicated with multiple layers and applications that give you a full end to end experience through multiple servers and protocols. In these cases, you will need to work through to determine what your various points in the connection are, and at what points you need to test from. For example, you may be doing remote desktop through a jump server (bastion host). Your workstation doesn't have direct access to the final remote desktop server. Testing connectivity from the jump server to the final destination only gives you part of the total round trip end to end connectivity. You will need to test from the workstation to the jump server, and then add the latency of the jump server to the final destination to get a rough idea of your end to end latency. If you can run ping's at the different layers, it will help give an idea of packet loss as well.

function Test-PortLatency {

              param ( [parameter(mandatory=$true)][string]$Computer,
                              [parameter(helpmessage="Timeout in milliseconds")]$timeout=10000

               $starttime = get-date
               $Testconn = New-Object Net.Sockets.TcpClient
               $Testconn.BeginConnect( $computer, $Port, $Null, $Null ) | Out-Null
               $MaxTimeout = ( Get-Date ).AddMilliseconds( $timeout)
               $millisec = 0

               While ( -not $Testconn.Connected -and ( Get-Date ) -lt $maxTimeout )

                              Sleep -Milliseconds 10
                              $ms += 10          

               $endtime = get-date
               $result = new-object psobject
               add-member -input $result NoteProperty Connected $($testconn.connected)
               add-member -input $result NoteProperty Milliseconds $(($endtime - $starttime).totalmilliseconds)

               if ($testconn.client.remoteendpoint -eq $null) {
                              $resultstr = "Connection_Refused"

               } elseif ($result.milliseconds -gt $timeout) {
                              $resultstr = "Connection_TimedOut"
               } elseif ($result.connected) {
                              $resultstr = "Successful_connection"
               } else {
                              $resultstr = "status_unknown"

               add-member -input $result NoteProperty Result $resultstr


No comments:

Post a Comment