How DCENT monitors hundreds of servers and petabytes of infrastructure across Europe with Bleemeo
“We've been using Bleemeo for several years, monitoring anywhere from 100 to 400 servers at a time. What I've always liked most is its simplicity: install it, and very quickly you have beautiful insights into what your hardware is actually doing and alerts when something goes wrong. Our infrastructure has grown enormously since we started using Bleemeo — from individual servers to tens of petabytes of storage, large GPU clusters and our own European BGP backbone carrying hundreds of gigabits of traffic. Bleemeo has been a very useful tool throughout that growth.”
What DCENT monitors with Bleemeo
Server Monitoring
Every physical server — CPU, memory, network, NVIDIA GPUs, and each individual drive in the chassis.
Learn moreUptime Monitoring
An independent view of whether systems and services are still reachable, from outside the infrastructure.
Learn moreAlerting & Notifications
Telegram channels and the mobile app, so an alert reaches someone who is nowhere near a desk.
Learn moreAbout DCENT
DCENT is a European infrastructure provider specialising in large-scale storage, GPU compute and network infrastructure.
From its own datacentre in the Netherlands and deployments across major European datacentre locations, DCENT operates hundreds of physical servers and tens of petabytes of storage infrastructure, alongside high-density GPU compute clusters and dedicated bare-metal environments.
Behind all of this sits its own rapidly growing European network. Operating under AS216456, DCENT runs its own BGP backbone between major European datacentre locations, with multiple high-capacity and redundant routes.
Rather than simply purchasing Internet connectivity independently at each location, DCENT has built networking into a core part of its infrastructure. It operates multiple 100G links, directly controls its routing and peers with hundreds of networks across a growing number of Internet Exchanges. The network regularly handles hundreds of gigabits per second of traffic at peak.
Its peering footprint includes AMS-IX, Speed-IX, ERA-IX, Piter-IX, 1-IX, GNM-IX and GigaNET-IXN, with the backbone continuing to expand alongside its compute and storage infrastructure.
This combination of storage, compute and its own BGP network means DCENT controls a large part of the infrastructure stack itself — from disks, servers and GPUs all the way to how traffic is routed across Europe and reaches the Internet.
Visit dcent.ioThe Challenge
As DCENT grew from a relatively small hosting environment into a distributed European infrastructure provider, monitoring became increasingly important.
Today its infrastructure consists of hundreds of physical servers, tens of petabytes of storage, high-density GPU systems and its own BGP network spanning multiple European datacentre locations.
With thousands of individual hardware components and hundreds of gigabits of network traffic to keep healthy, manually checking systems simply is not an option anymore.
A problem can be anything from abnormal CPU or memory usage to a failed disk, an overloaded GPU, a network issue or an entire service becoming unreachable. DCENT therefore needs a centralised view of its infrastructure and, more importantly, reliable alerts when something starts behaving differently from normal.
The Solution
DCENT has been using Bleemeo for several years, with anywhere between 100 and 400 servers monitored through the platform at different points in time.
One of the main reasons they have continued using it is its simplicity. New servers can be added quickly and immediately provide useful insights into CPU, memory, storage, network and overall system health, without requiring significant engineering time to build and maintain the monitoring platform itself. Server monitoring covers every host in the fleet.
The same applies to their GPU compute infrastructure, where understanding utilisation, system load and hardware behaviour is particularly important when operating expensive, high-density hardware.
As their own network has expanded across Europe, availability monitoring has also become increasingly valuable, providing another independent view of whether systems and services remain reachable.
What DCENT changed in Bleemeo
That is DCENT's account. This part is ours, because some of what they use did not exist before they asked for it — their hardware is unusual enough that supporting it properly meant changing the product.
Servers with more than a hundred drives
This took three separate fixes. A chassis with 100+ disks drew a chart that was one continuous rainbow band — technically correct and completely unreadable — so we made the displayed metrics selectable. Glouton needed several optimisations to keep emitting a point every 10 seconds on a host that dense. And S.M.A.R.T. disk health is read by external programs whose execution time did not survive being multiplied by a hundred, so that collection logic was reworked.
NVIDIA GPU metrics
DCENT was the first Bleemeo customer to need them, so that is where GPU support started: utilisation, memory, temperature and power draw, read through nvidia-smi. It is part of server monitoring for any GPU host now.
TrueNAS Core and TrueNAS Scale
Both are on the supported list because DCENT runs storage appliances on them — the appliances were part of the fleet, so they had to be part of the monitoring. They sit in the integrations catalogue with the other supported operating systems.
“DCENT handed us a server with more drives in it than most of our customers have servers — and then asked for GPU metrics and TrueNAS while they were at it. Our collection loop, our S.M.A.R.T. probes and our charts all quietly assumed a machine with a handful of disks, and all three had to change. We did not treat any of it as an edge case, because it is not one: that density is where storage is going. Everybody's dashboards are better for it, and it took a customer with unusual hardware to show us.”Lionel Porcheron — CEO & co-founder, Bleemeo
The Results
For DCENT, Bleemeo's biggest value has always been its simplicity: it makes a huge amount of infrastructure surprisingly easy to understand.
Over the years, Bleemeo has helped them monitor hundreds of servers, detect downtime quickly and better understand how their hardware is actually being used. That visibility has helped them improve hardware utilisation, troubleshoot problems faster and get much better insight into their infrastructure as they have scaled.
Their infrastructure has grown enormously since they first started using Bleemeo. Some of their largest environments are gradually moving towards larger self-hosted monitoring platforms, simply because of the scale they have reached — and they remain, in their own words, a big fan of Bleemeo for its simplicity, utility and how quickly it gives you the information that actually matters.
Hundreds of servers understood through one platform
Downtime detected quickly
Hardware utilisation improved with real usage data
Key Benefits
Simplicity at scale
A new server reports CPU, memory, storage and network health within minutes of the agent landing.
Density handled
Per-disk metrics every 10 seconds on machines carrying more than a hundred drives.
The systems they actually run
NVIDIA GPU metrics, and TrueNAS Core and Scale, next to the ordinary Linux fleet.
European end to end
A French company, metrics stored in the EU in the Paris region — consistent with what DCENT builds.
Frequently Asked Questions
Can Bleemeo monitor a server with more than 100 drives?
Yes, and DCENT is the reason it can. Their storage hosts carry over a hundred drives each, which took three changes on our side: selectable metrics, so a chart with that many series stays readable instead of collapsing into one coloured band; agent optimisations to keep producing a point every 10 seconds on a host that dense; and a reworked S.M.A.R.T. collection path, because the external programs that read disk health were too slow to run a hundred times per cycle. Dense storage hosts are a supported case, not an exception.
How quickly is a new server monitored?
Minutes. The agent installs with one command and discovers what runs on the host, so a new machine reports CPU, memory, storage, network and overall system health without anyone configuring dashboards first. At DCENT's scale that matters more than any single feature: adding a host is a line in the provisioning script rather than a monitoring project.
Does the Bleemeo agent run on TrueNAS?
Yes, on both TrueNAS Core and TrueNAS Scale. Support was added because DCENT runs storage appliances on them, and both are listed in the integrations catalogue alongside the other supported operating systems.
Does Bleemeo collect NVIDIA GPU metrics?
Yes — utilisation, memory used and total, temperature, fan speed, power draw and encoder/decoder utilisation, read through the nvidia-smi interface. DCENT was the first customer to ask for it, which is where the feature came from; it is now part of server monitoring for any GPU host.
How does Bleemeo collect application metrics?
Two ways, both in use at DCENT. If the application exposes a Prometheus endpoint, the agent scrapes it. If it does not, a script can produce the metrics and the agent picks them up. Either way they arrive beside system metrics, on the same dashboards and under the same alerting rules.
Where do Bleemeo alerts go?
DCENT uses Telegram channels and the Bleemeo mobile app, which delivers push notifications to whoever is on call. Email, Slack, webhooks, PagerDuty, Opsgenie and ilert are available too — 15+ notification channels in total.
Is Bleemeo a fit for hosting and bare-metal providers?
It is built for the case. Bleemeo is billed per monitored host rather than per feature, so a fleet of several hundred machines costs what the fleet is. And the monitoring runs outside the infrastructure it watches, which is what you need when that infrastructure is what you sell. Bleemeo is a French company and stores metrics in the EU, in the Paris region.
Monitoring that does not depend on your own infrastructure
Install the agent on one host — even one with a hundred drives in it — and see what it finds. Server monitoring starts in minutes.
No credit card required15-day trialEU data residency