The complaint never arrives while you’re looking. “The internet was terrible around lunchtime yesterday” is useless unless something was measuring at lunchtime yesterday. That’s the job of continuous latency monitoring, and two very different tools come up when admins go looking for it: SmokePing, the long-lived open-source Perl daemon whose “smoke” graphs many network engineers can read at a glance, and Obkio, a commercial SaaS platform that places agents at each site and measures the paths between them. Both will tell you the link was bad at 12:40. How they get there — and what it costs you in time, money and maintenance — is very different.
At a glance
| SmokePing | Obkio | |
|---|---|---|
| Model | Self-hosted daemon + CGI web front end | SaaS dashboard + monitoring agents |
| Licence | GPL-2.0, no cost | Commercial subscription (Starter, Basic, Premium, Enterprise); 14-day trial |
| Latest version at time of writing | 2.9.0 | Service, continuously updated |
| Runs on | Unix/Linux with Perl, RRDtool, a web server; probes such as fping | Software agents on Windows, Linux, VMware, Hyper-V; hardware agent; vendor-hosted public agents |
| What it measures | RTT, RTT distribution (“smoke”), loss, from one host to many targets | Latency, jitter, loss between agents every 500 ms; VoIP MOS; speed tests; DSCP classes; SNMP device metrics |
| Direction | One-way view from the SmokePing host outward (round trip) | Agent-to-agent, so both ends of a path are covered |
| Probes | Plug-in probes: ICMP (fping), DNS, HTTP, TCP, SSH and others | Synthetic traffic between agents; traceroutes |
| Multi-site | Master/slave mode | Native — agents at each site |
| Traceroute | Not built-in (pair it with mtr or PingPlotter) | Visual traceroutes, live traceroutes |
| Alerts | Pattern-based alerts (“matchers”) via config | Built-in notifications (email; webhooks on higher tiers) |
| Data retention | As long as your RRD files keep it (fixed-size, consolidated over time) | Depends on plan (vendor lists 6–36 months) |
Where they genuinely differ
Deployment effort
SmokePing assumes you’re comfortable on a Linux box. It’s packaged in most distributions, but you’ll still edit its target hierarchy by hand, set up the CGI under a web server, and make sure fping has the privileges it needs. For someone who has done it before, it’s an afternoon; for someone who hasn’t, it can be a frustrating weekend with Perl module errors. After that it runs quietly for years.
Obkio’s appeal is that the first graph can appear within the hour: sign up, install an agent at two sites (or use one of the vendor-hosted public agents in AWS, Azure or Google Cloud as the far end), and the dashboard starts filling in. The trade-off is that your monitoring now depends on a third-party service and an agent on your network, which some security teams will want to review.
What “the network” means to each tool
SmokePing measures from where it runs to each target. That’s excellent for “how does our internet edge look to these ten destinations” and for watching your own infrastructure from a central point. It says less about the path between two branches unless you deploy slave instances at each branch.
Obkio measures between agents, in both directions, every 500 ms. For a multi-site business that’s the natural model: branch A to HQ, branch A to the cloud, and so on. The 500 ms synthetic traffic also catches short bursts that a 20-probe-every-5-minutes SmokePing configuration would smooth over — a meaningful difference for voice and video, where Obkio’s MOS score is a direct answer to “was call quality bad?”.
Reading the evidence
SmokePing’s graphs are dense but eloquent: the median line, the grey smoke showing spread, and coloured loss markers compress hours of behaviour into one image. Carrier engineers tend to accept a SmokePing graph without argument. It doesn’t tell you where on the path the problem was; you pair it with a trace tool.
Obkio adds the path: its traceroutes are stored alongside the performance data, so you can see whether the route changed when latency jumped. It also scores results against thresholds, which is friendlier for a manager or a service desk. What it doesn’t give you is a raw, portable file format you fully control; the evidence lives in the vendor’s portal, and you export reports from there.
Cost and scale
SmokePing’s cost is your time and a small VM. It scales to hundreds of targets on modest hardware because each probe is lightweight and RRD files are fixed-size.
Obkio is priced per plan, with limits on agents and monitored devices per tier; the pricing page advertises a starting price and a calculator, and annual billing is discounted. For a handful of sites it’s typically cheaper than an engineer’s time to build and maintain the equivalent; for a very large estate, the per-agent economics need a proper quote. Check the vendor’s current pricing page rather than relying on figures quoted here.
Maintenance and longevity
SmokePing has been around since the early 2000s and is still maintained (2.9.0 was released in February 2025). It changes slowly — a feature for monitoring that you want to forget about. Obkio evolves continuously as a service, which brings new capabilities and also means interface and plan changes you don’t control.
The verdict
Pick SmokePing if…
- you have Linux skills in-house and want monitoring you fully own,
- you mainly need “latency and loss from here to there” for a list of targets,
- budget is zero or near-zero and a few hours of work is acceptable,
- you want long-lived graphs that network engineers instantly recognise.
Pick Obkio if…
- you run several sites and need both directions of each path measured,
- voice and video quality (jitter, MOS) is the core complaint,
- you want traceroute history and alerting without building them,
- nobody on the team wants to maintain a Perl/RRDtool stack.
A common hybrid: SmokePing on a central Linux host watching key external targets, with an on-demand tool such as PingPlotter or mtr for localising a bad hop once SmokePing shows when it happened. The method for that second step is in finding where packet loss starts.
Full details are in the SmokePing review and the Obkio review; related tools are in Latency & Path Analysis. Get software only from the project or vendor site — see where to get tools safely.