Making the Machines Talk Faster

Hi, Paul here again, CEO of CreeperHost.

A little while ago, I wrote about the systems behind CreeperHost, the bits of software and automation that keep everything moving behind the scenes.

One of those systems was Hermes, and its companion HermesMux.

Hermes is one of those things most customers will never know exists, which is exactly how it should be. You press a button in CreeperPanel, something happens on your server, and it should feel quick, boring, and reliable.

That simplicity hides a lot of machinery.

Every restart, status check, install, config change, log read, and background task involves our control systems talking to customer servers safely and quickly. Years ago, that meant a lot more waiting around for slow connection setup and repeated handshakes. Hermes was built to cut that down.

Recently, we went back and made it smarter again.

The Tiny Waits Add Up

Performance work is not always about one giant rewrite or one dramatic bottleneck.

Sometimes it is about all the tiny waits. The half-second here, the extra handshake there, the repeated setup work that happens thousands of times a day because a system is technically doing the right thing, just not in the smartest possible way.

That was the case here.

HermesMux already helped by keeping communication between CreeperPanel’s backend systems and customer servers fast. But as our platform has grown, the control systems themselves have become more distributed. We run multiple copies of important services so that no single process has to carry everything on its own.

That is good for reliability and scaling, but it creates an interesting problem.

If one part of the system has already done the work to establish a fast path to a server, another part of the system may not know that. It might end up doing the same setup work again somewhere else.

Nothing was broken. It was just wasteful.

And at our scale, wasteful eventually becomes noticeable.

Keeping The Fast Path Warm

The first part of the work was improving how HermesMux reuses existing communication paths.

Instead of treating every little management action as something that needs a fresh start, HermesMux is now better at keeping useful paths available for a short time and checking that they are still healthy before using them again.

That sounds simple, but it matters.

A lot of server management is made up of small actions. One request checks status. Another reads something. Another starts a process. Another updates state in CreeperPanel. If each of those actions has to pay the full setup cost every time, the panel feels slower than it needs to.

By keeping more of that work ready, the whole system feels more immediate.

Not because we changed what customers do, but because we removed some of the waiting between each step.

Teaching HermesMux To Share

The bigger improvement was making HermesMux better at working as a group.

Previously, if a request landed on one running copy of HermesMux, that copy handled it itself. That is simple and reliable, but it means different copies can end up building their own separate paths to the same place.

Now, the running copies of HermesMux can coordinate internally.

If one copy is the best place to handle a particular server, the others can send that work to it across our private control network. That means requests are more likely to use an already-warm path instead of starting from cold.

From the outside, nothing changes.

CreeperPanel still asks for something to happen. Aries still coordinates the work. Hermes still talks to the server. The customer still sees the result.

The difference is that the machinery in the middle is making better decisions.

It is a bit like having a team of people at desks. Before, whichever person picked up the request handled it from scratch. Now, if someone else already has the right file open, the request can go to them instead.

That is not a flashy feature. It does not get a new button. But it makes the whole thing smoother.

Measuring The Invisible

The other important part of this work was visibility.

Whenever we change something deep inside the platform, we want to know what it is doing. Not vibes. Not “seems faster on my machine”. Actual numbers.

So we added more internal health checks and dashboards around HermesMux.

We can now see whether requests are using warm paths, how often work is being shared internally, whether anything is queueing, whether fallback paths are being used, and whether the system is getting healthier or just hiding a new problem somewhere else.

That last part matters.

A system can look faster because it is doing less work, or it can look faster because it is failing earlier. Those are very different things.

The goal was not just to reduce a graph. The goal was to make sure the system was faster, calmer, and still behaving correctly.

So far, that is exactly what we are seeing.

Why Customers Should Care

Most customers will never see HermesMux. You will not find it in CreeperPanel, and there is no setting for it.

But you should feel it.

You should feel it when server controls respond without that odd little pause. You should feel it when staff can help you more quickly with a support request. You should feel it when the panel feels more consistent, especially at busy times.

That is the kind of work we spend a lot of time on.

Not every improvement is a new page, a new theme, or a new feature announcement. Sometimes the best work is making something you already use feel less annoying.

Less waiting. Less duplicated work. Less weirdness.

More of the panel doing what you asked, when you asked.

Built, Measured, Improved

This is also why we keep building so much ourselves.

If we were just reselling someone else’s panel or stitching together off-the-shelf tools, this kind of improvement would be difficult or impossible. We would be waiting for someone else to care about our exact problems.

But this is our platform.

We built Aries. We built Hermes. We built HermesMux. We run the control systems ourselves, on our own infrastructure, with our own monitoring, for our own customers.

That means when we see something that can be better, we can go and make it better.

Sometimes that means new features. Sometimes it means replacing old systems. Sometimes it means spending a long evening staring at graphs until the lines finally move the way you hoped they would.

This one was the graph kind.

And honestly, those are some of my favourite wins. Quiet, technical, invisible to most people, but real.

The Work Never Really Stops

Hosting is full of these little layers.

Hardware matters. Networks matter. The panel matters. Support matters. But between all of those are thousands of small decisions, tiny pieces of software, and background systems that decide whether everything feels smooth or frustrating.

HermesMux is one of those pieces.

It was already doing an important job. Now it does that job with less waste and better coordination.

That is the kind of improvement we will keep making.

Because good hosting is not just about having powerful machines. It is about making all the machines behind the machines work together properly.

- Paul (CEO, Founder, CreeperHost)