Netmiko vs Ansible: Does It Matter?

We launched device configuration with Netmiko in the September 2026 netlab release and made it easier to use in the October 2026 release. It will become the default configuration method for several popular platforms (Arista EOS, Cisco IOS, Cisco Nexus OS, Cisco IOS XR, FRRouting virtual machines) in a few months, but before doing that, I wanted to check whether the change matters.

TL&DR: Yes, it does.

You can’t fairly compare these two tools; Netmiko does one thing (handling SSH sessions and downloading device configurations), while Ansible tries to be everything for everyone. It’s like comparing a chainsaw to a Swiss Army knife. However, since netlab only needs configuration downloads, it made sense (to me) to compare performance on that task.

I did a very unscientific test:

  • I started an EVPN lab because there’s plenty to configure there (configuration normalization, initial configuration, VLANs, OSPF, BGP, VRFs, VXLAN, EVPN)
  • I measured how long it takes to start the lab (Arista EOS and Linux containers) and how long it takes to configure it.
  • The configure the lab part on Arista EOS was done with bash scripts, Netmiko and Ansible. netlab always configures Linux containers with bash scripts.

Here are the results:

Operation netlab CPU time1 Elapsed time
Starting the lab 1.6 s 23.5 s
Configuration with bash scripts 0.7 s 4.3 s
Configuration with Netmiko 0.9 s 6.2 s
Configuration with Ansible 9.8 s 33.1 s

The difference between Netmiko and Ansible configuration time is significant, and while it doesn’t matter much if you’re using VMs (which start much slower than containers), just watching the Ansible playbook generating noise while slowly progressing through the configuration steps makes me cringe.

Speaking of noise: here’s the complete printout from a Netmiko-based configuration of a lab topology with two switches and four hosts; you really don’t want to see the corresponding Ansible printout2.

┌──────────────────────────────────────────────────────────────────────────────────┐
│ CHECKING Are lab devices ready to be configured?                                 │
└──────────────────────────────────────────────────────────────────────────────────┘
[INFO]    Checking SSH server(s) on dut_s1,dut_s2
[SSH] Waiting for 1 devices (1.0 seconds)
┌──────────────────────────────────────────────────────────────────────────────────┐
│ CONFIG Normalizing device configurations                                         │
└──────────────────────────────────────────────────────────────────────────────────┘
[INFO]    Executing normalize configuration for node dut_s1
[INFO]    Executing normalize configuration for node dut_s2

┌──────────────────────────────────────────────────────────────────────────────────┐
│ CONFIG Deploying device configurations                                           │
└──────────────────────────────────────────────────────────────────────────────────┘
[INFO]    Executing initial configuration for node dut_s1
[INFO]    Executing initial configuration for node dut_s2
[INFO]    Executing initial configuration for node h1 (namespace clab-evpn-h1)
[INFO]    Executing initial configuration for node h2 (namespace clab-evpn-h2)
[INFO]    Executing initial configuration for node h3 (namespace clab-evpn-h3)
[INFO]    Executing initial configuration for node h4 (namespace clab-evpn-h4)
[INFO]    Executing routing configuration for node h1 (namespace clab-evpn-h1)
[INFO]    Executing routing configuration for node h3 (namespace clab-evpn-h3)
[INFO]    Executing routing configuration for node h4 (namespace clab-evpn-h4)
[INFO]    Executing routing configuration for node h2 (namespace clab-evpn-h2)
[INFO]    Executing vlan configuration for node dut_s1
[INFO]    Executing vlan configuration for node dut_s2
[INFO]    Executing ospf configuration for node dut_s1
[INFO]    Executing ospf configuration for node dut_s2
[INFO]    Executing bgp configuration for node dut_s1
[INFO]    Executing bgp configuration for node dut_s2
[INFO]    Executing vrf configuration for node dut_s2
[INFO]    Executing vrf configuration for node dut_s1
[INFO]    Executing vxlan configuration for node dut_s2
[INFO]    Executing vxlan configuration for node dut_s1
[INFO]    Executing evpn configuration for node dut_s2
[INFO]    Executing evpn configuration for node dut_s1

Results of non-Ansible configuration deployments
====================================================================================================================================
dut_s1           Script:   normalize,initial,vlan,ospf,bgp,vrf,vxlan,evpn
dut_s2           Script:   normalize,initial,vlan,ospf,bgp,vrf,vxlan,evpn
h1               Script:   initial,routing
h2               Script:   initial,routing
h3               Script:   initial,routing
h4               Script:   initial,routing


[SUCCESS] Lab devices configured

Finally, I prefer SSH-based configuration over hacks like the FastCLI scripts that netlab can use with Arista EOS containers. The performance difference isn’t huge, and we can use the same method with virtual machines and containers.

  1. Includes the CPU time spent in child processes, for example, ansible-playbook. ↩︎

  2. Yeah, I know Ansible has stdout callbacks, and some of them are pretty terse. However, I haven’t found one that would produce the concise information netlab generates when used with Netmiko, and I definitely won’t waste time writing my own. ↩︎

Recent posts in the same categories

Add comment
Home
Sidebar