Full-text search vs label-based filtering

Purl vs Grafana Loki

Loki indexes only labels and requires Grafana plus object storage to function. Purl indexes every log line in ClickHouse and ships with its own dashboard, alerting, and pattern detection — no external dependencies, no extra components.

FeaturePurlGrafana Loki
Search
Full-text search + SQL
Label-based filtering, LogQL
Storage backend
ClickHouse (built-in)
Object storage (S3/GCS required)
Setup complexity
Single binary, 5 min
Multiple components, object storage config
Dashboard
Built-in web UI
Requires Grafana
Indexing
Full content indexed
Only labels indexed (grep for content)
Alerting
Built-in (Telegram, Slack, Webhook)
Via Grafana alerting rules
Resource usage
256MB min
Varies (object storage + multiple components)
Query performance
Sub-second full-text
Fast on labels, slow on content grep
Self-contained
Yes (all-in-one)
No (needs Grafana, object storage)

Recommended when

When to choose Purl

You need full-text search

Purl indexes every character of every log line. Search for any string, regex, or SQL expression across millions of logs in milliseconds — no label pre-planning required.

Simpler operations matter

One Docker Compose file. No object storage buckets to configure, no Grafana to maintain, no Loki compactor or ruler components. Purl is a single binary backed by ClickHouse.

You want an all-in-one solution

Purl ships with a built-in web dashboard, alerting (Telegram, Slack, Webhook), pattern detection, and live tail — everything in one place with zero external dependencies.

No external dependencies

Loki requires object storage (S3, GCS, or Azure Blob) and Grafana to be useful. Purl needs nothing beyond Docker — perfect for self-hosted environments with strict network policies.

Consider the alternative when

When to choose Loki

You have an existing Grafana stack

If your team already runs Grafana, Prometheus, and Tempo, Loki integrates natively. Adding Purl would mean maintaining a separate UI.

Label-centric, structured workloads

If your logs are highly structured with consistent labels (Kubernetes pod, namespace, service) and you rarely search raw content, Loki's LogQL model fits well.

Massive scale with cheap object storage

At petabyte scale, Loki's object storage backend (S3) can be cheaper than running large ClickHouse clusters — if you can accept slower full-text search.

Try Purl Free

No credit card required. Self-hosted, full-text search, built-in dashboard — up and running in 5 minutes.