See Everything.
Store Everything.
Explain Everything.
Every Forge.io deployment ships with a fully integrated observability stack: metrics, logs, traces, and zero-code instrumentation. No third-party contracts. No data leaving your sovereign boundary. No gaps.
Live Demo:Coming Soon
METRICS SERIES: 2.4M active
LOG INGESTION: 1.2TB/day
TRACES: 847K spans/min
BEYLA COVERAGE: 100%
ForgeObserve / Production / PACS Gateway Overview
DICOM C-STORE Latency (p95)47ms
Error Rate
0.01%
Ingestion
1.2GB/s
Live Logs (Loki)
10:42:12 [INFO] DICOM Association Accepted (AE: MODALITY_1)
10:42:13 [INFO] C-STORE-RQ received SOP=1.2.840...
10:42:15 [INFO] Storage commit successful
The Problem We Solved
Healthcare infrastructure is too critical for observability gaps.
Most healthcare platforms run on a patchwork of monitoring tools: one for infrastructure, another for applications, a third for compliance logging, none of them talking to each other. When a PACS query times out at 2am, you're correlating timestamps across four dashboards and three vendors.
We built ForgeObserve because observability shouldn't be a project. It should be infrastructure.
Traditional Stack
ForgeObserve
Metrics + Logs + APM vendors
Single integrated stack
Agents per service, per language
Zero-code eBPF
30-day retention (budget)
13-month compliance default
Data shipped to 3rd party SaaS
Data stays in sovereign region
Correlation is your problem
Auto-linked Signals
Platform Architecture
Modular by Design. Sovereign by Default.
Every layer is purpose-built and independently scalable. Like precision-engineered blocks, each component snaps together to form a complete sovereign infrastructure.
YOUR WORKLOADS
APIs
Workers
AI Models
Databases
IsoCells
Firecracker MicroVMs
IsolationHardware (KVM)
Boot Time<125ms
MemoryDedicated
c4.microc4.standardc4.dedicated
ForgeMesh
Zero-Trust Network
EncryptionmTLS Always
PolicyeBPF Enforced
Latency<1ms
L3/L4L7 Proxy
Sovereign Store
Object Storage
EncryptionAES-256
ResidencyLocked
ForgeObserve
LGTM Stack
MimirMetrics
LokiLogs
TempoTraces
GrafanaVisualize
Forge Control Plane
Orchestration & Compliance
SchedulerPlacement
PolicyCompliance
SecretsVault
RegionsRouting
Sovereign Regions
Hardware-Enforced Data Residency
SYD-SOV-1Sydney
SIN-SOV-1Singapore
FRA-SOV-1Frankfurt
LON-SOV-1London
TYO-SOV-1Tokyo
IAD-SOV-1N. Virginia
Every layer is independently auditable, scalable, and replaceable
The LGTM Stack
Four systems. One unified view.
Your Sovereign Workloads
PACS Gateway
Worklist Svc
AI Model
Portal App
Grafana Beyla (eBPF)
Zero-code instrumentation at the Linux kernel level
Grafana Alloy
Unified Collector: Routing, Filtering, Sampling, Enrichment
Mimir
Metrics Storage
Loki
Log Aggregation
Tempo
Distributed Tracing
Grafana
Visualize. Alert. Correlate. Query.
Grafana: The Single Pane
One dashboard. Every signal.
Grafana is the query and visualisation layer for everything in your sovereign environment. Metrics, logs, traces: correlated by time, linked by trace ID, explorable in a single interface.
- Unified exploration: Jump from a spike in latency → logs → trace.
- Pre-built dashboards: DICOM ingestion, HL7/FHIR throughput.
- Alerting: Threshold, anomaly, and composite alerts routed to PagerDuty/Slack.
PACS PERFORMANCE OVERVIEW
Open in Grafana C-STORE Rate
42/min
P95 Retrieval
145ms
S3 Write IOPS
880
Active Assocs
12
DICOM Ingestion Volume (MB/s)Last 1 hour
10:0010:1510:3010:4511:00
Active Modalities
CT_SCANNER_01
45 MB/sMR_HEAD_03
12 MB/sUS_CARDIO_02
-CR_CHEST_01
-Storage Pools
Hot (NVMe)42%
Warm (S3)81%
dashboard.json
{
"title": "PACS Performance Overview",
"panels": [
{
"type": "timeseries",
"title": "Study Retrieval Latency (p95)",
"targets": [
{
"expr": "histogram_quantile(0.95, sum(rate(pacs_study...))",
"legendFormat": "p95 latency"
}
]
}
]
}
Mimir: Metrics at Scale
Prometheus, without the scaling headaches.
Mimir is a Prometheus-compatible time-series database that scales horizontally to billions of active series. It's what runs under Grafana Cloud: now running in your sovereign region.
13mo
Default Retention
1B+
Active Series
MIMIR QUERY PERFORMANCE
Open in Grafana Instant QPS
845req/s
P99 Latency
142ms
Cache Hit Ratio
99.4%
Query Latency Heatmap (Last 1h)
bucket size: 10ms-60m-45m-30m-15mNow
Slow Query Log (>500ms)
14:22:01SUM(rate(dicom_ingest_bytes...))542ms
14:21:58HISTOGRAM_QUANTILE(0.99, ...))210ms
# Average DICOM C-STORE latency by modality, last 24h
avg by (modality) (
rate(dicom_cstore_duration_seconds_sum[24h])
/
rate(dicom_cstore_duration_seconds_count[24h])
)
Loki: Logs Without the Bill Shock
Index the labels. Store the logs.
Loki takes a radically different approach to log aggregation: instead of indexing every word in every log line, it indexes only the metadata (labels) and stores log content in cheap object storage. The result? 10x lower cost than traditional log platforms.
Compliance Ready: Store 13 months of logs without budget panic. Logs are immutable and PHI-aware.
Compliance Ready: Store 13 months of logs without budget panic. Logs are immutable and PHI-aware.
LOKI LOG EXPLORER
Open in Grafana {service="pacs-gateway"}|="error"Last 5m
20:21:42.105
info
service=pacs-gatewayDICOM Association negotiation started (AE: CT_SCANNER_04)traceID=7f8a9b
20:21:42.145
info
service=pacs-gatewayAssociation accepted. Transfer Syntax: 1.2.840.10008.1.2.4.90traceID=7f8a9b
20:21:42.220
info
service=pacs-gatewayReceived C-STORE-RQ (SOP: 1.2.840...4892)traceID=7f8a9b
20:21:42.450
info
service=store-serviceWriting object to S3 bucket: forge-pacs-sysov1-imgtraceID=7f8a9b
20:21:42.890
info
service=index-serviceMetadata extracted successfully. PatientID: 998231traceID=7f8a9b
# C-STORE errors for specific modality
{service="pacs-gateway", modality="CT"}
|= "error"
| json
| processing_time > 500ms
Tempo: Traces Without Sampling
Every request. Every span. No sampling.
Tempo stores 100% of your traces: no head-based sampling, no tail-based sampling. When a request fails, the trace exists.
- Full fidelity: That one-in-a-million failure? Traceable.
- Trace-to-logs: Click a span, see the logs from that exact moment.
TEMPO TRACE WATERFALL
Open in Grafana traceID=3a1b2c...245ms
pacs-gatewayHTTP POST /dicom/instances (245ms)
auth-serviceverify_token (25ms)
pacs-coredicom.validate_sop (35ms)
storage-servicepersist_object (135ms)
s3-gatewayS3.PutObject (115ms)
index-servicedb.insert_metadata (15ms)
Gateway
Storage
Auth
Total Spans: 6
# Find traces where PACS query exceeded 2s
{ span.service = "pacs-gateway" && span.duration > 2s }
Zero-Code Instrumentation with Beyla
Your code doesn't need to change. Beyla sees everything.
Grafana Beyla uses eBPF, a technology that allows safe programs to run inside the Linux kernel, to capture telemetry without any application modification. No SDKs. No code changes. No vendor lock-in.
The Healthcare Reality:
Your PACS vendor ships a binary. No source code. No instrumentation hooks. With Beyla, you get full observability (latency percentiles, error rates, traces) without asking the vendor for anything.eBPF PROBE
HTTP/S TrafficRED Metrics
SyscallsSpans
SQL QueriesLatency
beyla.yaml
open_port: 8080
service_name: pacs-gateway
prometheus_export:
port: 9090
otel_traces_export:
endpoint: http://alloy:4317
Unified Collection with Alloy
One collector. Every signal. Every protocol.
Grafana Alloy is the unified telemetry collector that routes metrics, logs, traces, and profiles from your applications to the LGTM stack. It speaks OpenTelemetry, Prometheus, Loki, and more.
- Protocol translation: Receive OTLP → Emit Prometheus.
- Pipeline processing: Filter, transform, enrich, and route telemetry.
- Fleet management: Centralised configuration for all collectors.
Grafana Alloy
OpenTelemetry Collector Distribution
config.alloy
// Receive OpenTelemetry traces
otelcol.receiver.otlp "default" {
grpc { endpoint = "0.0.0.0:4317" }
}
// Process and batch
otelcol.processor.batch "default" {
timeout = "5s"
send_batch_size = 1000
}
// Export to Tempo
otelcol.exporter.otlp "tempo" {
client { endpoint = "tempo:4317" }
}
otelcol.processor.batch "default" {
output { traces = [otelcol.exporter.otlp.tempo.input] }
}
Healthcare-Specific Observability
Dashboards built for healthcare,
not retrofitted to it.
ForgeObserve ships with pre-built dashboards and alerting rules designed for healthcare workloads. Not generic APM. Actual healthcare observability.
DICOM GATEWAY
C-STORE Performance
Association Failures0
C-FIND Latency12ms
Latency Trend (1h)
HL7 INTEGRATION
Message Throughput
ADT Ingestion84/sec
ACK Latency4ms
Volume (1m)
STUDY LIFECYCLE
Archive Stats
Retrieval p95240ms
Storage Growth+4TB/day
Ingress Rate (24h)
Query Power
Three query languages.
One learning curve.
The LGTM stack uses purpose-built query languages that share common patterns. Learn one, understand all three.
The Power Move: Correlation
Click a metric spike → See the logs from that window → Jump to the trace that caused it → Identify the slow span → See the SQL query. All without leaving Grafana.
Visualisation that earns its screen space
Dashboards that answer questions, not just display numbers.
Grafana's visualisation library is the deepest in the industry: 40+ panel types, from simple stat counters to full geospatial maps.
- Stat: Current value with sparkline (e.g., "p95: 47ms")
- Heatmap: Latency distribution over time
- Node graph: Service topology, dependency mapping
Version control your observability. Promote dashboards through environments.
ISOCELL FLEET HEALTH
Active IsoCells
48/ 48
Avg Boot Time
130ms
Fleet CPU
38%
Alerting & Incident Response
Alerts that reach the right person, with the right context.
Grafana Alerting evaluates rules against metrics, logs, or traces, then routes notifications through escalation chains.
Slope-based alerts
Multi-signal rules
PagerDuty
Critical: DICOM Gateway Down
alerts.yaml
groups:
- name: pacs-alerts
rules:
- alert: HighStudyRetrievalLatency
expr: histogram_quantile(0.95, sum(rate(pacs_study...))) > 3
for: 5m
labels: { severity: warning }
annotations:
summary: "Study retrieval latency is elevated"
description: "p95 is {{ $value }}s (threshold: 3s)"
Compliance & Retention
Observability data is compliance data.
In healthcare, your metrics and logs aren't just operational telemetry: they're audit evidence. ForgeObserve treats observability data with the same compliance rigour as your application data.
13-Month Retention
Default for metrics and logs. Covers HIPAA audit windows without custom pricing.
Immutable Storage
Logs and traces cannot be modified post-ingestion. WORM compliant.
Audit Export
One-click export to PDF or JSON for auditors. Chain of custody included.
See what your infrastructure
is actually doing.
Every Forge.io deployment includes ForgeObserve: no add-ons, no extra contracts, no data leaving your sovereign region.