Data centres & colocation
Data centre monitoring, down to the door.
A data centre is four operations problems in one building — the fabric, the compute, the environment, and who is standing in front of the cage — and most operators run four systems to watch them. PrahiX puts network, security, environmental and physical access telemetry on one platform, so an SLA event, a thermal excursion and an unbadged entry are one timeline instead of three consoles and a bridge call.
metrics ingested per second across the estate
video latency on cage and aisle cameras
timeline for network, power, environment and access
Why four systems is the default, and the problem.
Each of the four is a reasonable purchase on its own. Together they produce a facility where every layer is instrumented and no one can see an incident that crosses two of them — which, in a building this dense, is most of them.
The SLA is a contract, not a dashboard
Uptime is sold with penalties attached, so the evidence that you were available matters as much as being available. When a credit is disputed, an assertion loses to a record — and the record has to have been collecting all along.
Density outruns threshold alerting
Thousands of ports, nodes and sensors mean any static threshold produces a storm rather than a diagnosis. The information needed to act is present somewhere in the flood; that is exactly the problem.
Power and cooling belong to a different team
Facility telemetry sits outside IT monitoring, on its own system, watched by its own people. A thermal excursion and the node failures it caused arrive half an hour apart in two places, and get investigated as two events.
Physical access is the audit everyone fails
Cage entries, mantrap logs, escort records and camera coverage live in a fourth system that correlates with nothing. Answering "who was at rack 42 when that change was made" becomes an archaeology project.
What one operations platform gives an operator.
The layers stop being separate purchases and become one estate — which changes what can be diagnosed, what can be automated, and what you can hand a tenant without preparing it first.
SLA evidence, produced continuously
Per-tenant availability and performance records accumulate as the facility runs, so a service-credit conversation starts from a record rather than from two parties reconstructing a night from different logs.
Topology-aware diagnosis, not alert storms
Symptoms are correlated back to the device, link or row that actually failed and the downstream storm is suppressed. One diagnosed fault reaches the floor instead of four hundred notifications about its consequences.
Environment and power alongside the fabric
UPS, PDU, cooling units and environmental sensors are polled as part of the same estate as the switches and servers, so the temperature that preceded the failure is on the same timeline as the failure.
Multi-tenant separation, one console
Per-tenant scoping and reporting without a deployment per tenant. Your operations team sees the facility; each customer's report contains their racks, their ports and their uptime, and nothing else.
Cage, aisle and door on the same timeline
Access events, camera health and the clip itself sit alongside network and identity events, so a physical presence and a configuration change at the same rack are one investigation rather than two systems and a guess.
The record your auditors and tenants ask for
Availability, access, camera coverage and change history are retained as operating by-products. Whatever the audit — a certification body, an insurer, or a tenant's own security team — the underlying evidence already exists.
The building and the fabric, on one platform.
In most industries converged operations is an argument about efficiency. In a data centre it is an argument about causality — because here the network, the power, the security estate and the physical space are the same failure domain, and they fail into each other.
A thermal excursion and a node failure are one incident
Cooling telemetry, rack inlet temperature, device health and interface state correlate together, so the sequence is visible as a sequence. On separate systems it is two incidents with an unexplained coincidence between them.
An entry and a change at the same rack
Badge reads, camera coverage and configuration changes are on one timeline. Whether that pairing is routine maintenance or something to escalate is a question you can now actually answer, in the moment rather than in a review.
Cameras and controllers are network devices too
The access-control and surveillance estate is monitored for health like everything else, so the camera covering the cage is known to be working before the week somebody needs its footage.
Your tenants' questions, answered from your own record
"Was my rack affected?" and "who accessed my cage?" are the two questions colocation customers ask under pressure. Both are queries against one platform rather than requests routed to three teams.
The difference at 02:40 in a cold aisle
Four systems, three teams
- The facility system raises a temperature alarm in row 7.
- The NMS raises forty interface alarms nine minutes later.
- The SOC sees authentication activity on a management VLAN.
- Three tickets — and the tenant's call arrives before the correlation does.
PrahiX unified platform
- The excursion, the ports dropping and the management-VLAN activity arrive as one incident.
- The diagnosis names row 7, the cooling unit and the racks behind it.
- The access record shows who was in that aisle, with the clip attached.
- The tenant call is the one you make, not the one you take.
How it works in a facility.
Four stages. In a data centre the second one carries more weight than usual, because a rack has a normal and a row has a normal and they are not the same thing.
Discover the whole facility
Agentless discovery across the fabric, compute and storage, plus UPS, PDU, cooling units and environmental sensors, plus cameras and access controllers — with topology built from LLDP, CDP and routing so dependencies are known rather than assumed.
Baseline per device, per rack, per row
Each device learns its own normal, and so does each rack and hall — inlet temperature, power draw, port utilisation. Density means the useful signal is usually a deviation from local normal rather than a breach of a facility-wide threshold.
Correlate across all four layers
Environmental, network, security and access telemetry are correlated together against one incident object, so a sequence that crosses layers is recognised as a sequence instead of arriving as unrelated alarms in three consoles.
Act, then evidence it
Known faults trigger their remediation playbook and novel ones escalate with the diagnosis attached. Everything — availability, actions, access and changes — accumulates into per-tenant and per-facility reporting that is exported rather than assembled.
What one platform replaces.
The saving that shows up is not the fourth licence. It is the correlation nobody was doing, and the hours spent every month producing reports out of systems that were never designed to produce them together.
| Capability | NMS + facility system + SOC + VMS | One vendor, several products | PrahiX unified platform |
|---|---|---|---|
| Estate coverage | Complete per layer, blind between them | Broad, still stitched at the seams | One estate: fabric, power, security, access |
| Cross-layer correlation | Not possible — four data models | Via integrations, on forwarded fields | Native, on one incident object |
| Alert storms | One per console | Reduced, not diagnosed | Suppressed to the causing device or row |
| Per-tenant reporting | Assembled by hand each month | Per product, then reconciled | Scoped and exported per tenant |
| Physical access | A separate system, correlated with nothing | Rarely in scope | Same timeline as network and identity |
| SLA evidence | Reconstructed when disputed | Partial, per product | Continuous, per tenant |
| Integrations to maintain | Several, and they break on upgrades | Fewer, still versioned | None between the layers |
| Cost shape | Four stacks plus the integration effort | One vendor, several subscriptions | One operating subscription |
The evidence an operator is asked for
A data centre is audited from three directions at once — by certification bodies, by insurers, and by tenants' own security teams, who increasingly audit their providers directly. All three ask for the same underlying material: what was available, who had access, what changed and when. Here it accumulates as the facility runs.
- Per-tenant availability and performance records
- Access, escort and camera-coverage history
- Configuration-drift and change history
- ISO 27001:2022-aligned controls · DPDP Act-aligned handling · Made in India
Trusted by operations teams across India
Have Questions? We've Got Answers.
DCIM is facility-centric: capacity planning, floor and rack layout, asset placement, power budgeting. This is operations-centric: continuously watching the fabric, the compute, the environmental and power telemetry, the security estate and physical access, and correlating across all of them when something goes wrong. PrahiX is not a DCIM replacement and does not do capacity planning or floor layout — where you run one, we monitor alongside it and bring its telemetry into the same timeline as everything else.
UPS, PDUs, cooling units and environmental sensors are polled as part of the estate over the same open standards as everything else, so inlet temperature, power draw and unit health sit on the timeline next to device and interface state. What we do with that is correlation and alerting, not thermal modelling or capacity forecasting.
Yes. Scoping is per tenant, so a customer's view and their reporting contain their racks, ports and services and nothing else, while your operations team sees the whole facility. It is one deployment — you are not standing up an instance per customer, which is what makes it practical at colocation scale.
It can, and for most operators that is the cleanest outcome — network monitoring is the layer PrahiX is strongest on and it is usually the oldest tool in the room. Where a monitoring platform has years left on its licence, we can ingest from it instead and consolidate later. Worth raising early, because it changes the sequencing considerably.
Access controllers and cameras are monitored as network devices for health, and their events join the same incident timeline as network and identity activity. That is what makes "who was in that aisle when this changed" answerable as a query rather than as a request to three teams — and it is the part almost no monitoring platform does.
Yes. Sites are monitored identically and rolled up, so a facility with a small local team is as visible as your flagship. Reporting can be scoped per site, per tenant, or both, which matters when a single customer has racks in more than one of your buildings.
It supplies the operating evidence those audits examine — availability, access, camera coverage, change history — retained continuously rather than assembled beforehand. It does not certify anything, and no platform can: the certification is awarded against your facility, your processes and your design, by a body that assesses all three. What we remove is the fortnight of evidence-gathering each time.
Yes, and we would suggest one hall rather than one layer — the whole argument is cross-layer correlation, and it cannot be demonstrated on the fabric alone. A hall with its own cooling, its own access control and a mixed vendor estate is the honest test.
Keep reading
Show us your messiest hall.
Mixed vendors, its own cooling, its own access control, and documentation that stopped being accurate two refreshes ago. That is where cross-layer correlation either proves itself or does not — so that is where we would scope the POC.