vs. Grafana Cloud

Core Assumptions for this Comparison¶

Workload Definition (based on Micromegas Example Deployment):
- Total Events (over 90-day retention):
  - Logs: 9 billion log entries
  - Metrics: 275 billion metric data points (equivalent to 1,000,000 active series for pricing)
  - Traces: 165 billion trace events
- Monthly Ingestion Rate (for pricing comparison):
  - Logs: 3 billion log entries / month (9 billion / 3 months)
  - Metrics: ~92 billion metric data points / month (275 billion / 3 months)
  - Traces: 55 billion trace events / month (165 billion / 3 months)
- Retention: 90 days (3 months)
- Users: 5 active users.
Grafana Cloud Pricing Assumption:
- We will use the publicly available pricing for the Grafana Cloud Pro plan as of mid-2025.
- Pricing is calculated based on logs ingested (GB), metric series, traces ingested (GB), and users.
- Assumption on Grafana Cloud Data Size: To enable a dollar-for-dollar comparison based on events, we must estimate the storage consumed by these events in Grafana Cloud. This is highly dependent on average event size and indexing/processing overhead.
  - Average log entry size for Loki (after processing): 500 bytes
  - Average trace event size for Tempo (after processing): 1 KB
Micromegas TCO Assumption:
- The Total Cost of Ownership (TCO) for a self-hosted solution must include both direct infrastructure spend and the cost of personnel to manage the system.
- Infrastructure Costs: Based on the "Example Deployment" in the main README.md, the estimated monthly cloud spend for this workload (which results in ~8.5 TB of storage for Micromegas) is ~$1,000 / month.
- Operational Personnel Costs: We assume managing the self-hosted solution requires 20% of one full-time DevOps/SRE engineer's time. At a fully-loaded annual salary of $150,000, this equates to ~$2,500 / month.

The Challenge of Traces: Why Direct Comparison is Impractical¶

Micromegas is designed to ingest and store a very high volume of raw trace events (165 billion total, or 55 billion per month in our example) and process them on-demand. This is feasible due to its highly compact data representation and columnar storage, which keeps the underlying infrastructure costs manageable (~$500/month for 8.5 TB of total data, including traces).

Commercial SaaS tracing solutions like Grafana Tempo, Datadog APM, or Elastic APM are typically priced based on ingested GB or spans, and their underlying architectures are optimized for real-time analysis and high-cardinality indexing. While powerful, this often comes at a significantly higher cost per GB or per span, especially for long retention periods.

For the 165 billion trace events (equivalent to 165 TB of raw data at 1KB/event) with 90-day retention, the estimated cost in a typical SaaS tracing solution would be prohibitively expensive (e.g., hundreds of thousands of dollars per month, as seen in previous calculations). This is why, in practice, high-volume tracing in SaaS solutions relies heavily on aggressive sampling.

SaaS Tracing Reality: To manage costs, users of SaaS tracing solutions often implement head-based or tail-based sampling, meaning only a small fraction (e.g., 1-10%) of traces are actually ingested and retained. This sacrifices data completeness for cost control.
Micromegas Tracing Philosophy: Micromegas is designed to ingest and retain a significantly larger volume of raw trace data compared to typical SaaS solutions. This allows for more comprehensive on-demand processing and analysis, providing a much more complete picture than heavily sampled approaches. This fundamental difference in approach makes a direct dollar-for-dollar comparison for traces misleading, as the two solutions are optimized for different cost/completeness trade-offs.

Analysis for Hypothetical Workload¶

1. Estimated Monthly Cost: Micromegas (Self-Hosted)¶

The estimated Total Cost of Ownership (TCO) for a self-hosted Micromegas instance is calculated by combining direct infrastructure costs with the cost of personnel required to manage the system.

Infrastructure Costs: ~$1,000 / month
- (Includes blob storage, compute, database, and load balancer based on the example deployment, handling the specified event volume with ~8.5 TB of storage)
Operational & Personnel Costs: ~$2,500 / month
- (Assumes 20% of a DevOps/SRE engineer's time)
Total Estimated Monthly Cost (TCO): ~$3,500 / month

2. Estimated Monthly Cost: Grafana Cloud Pro (Logs & Metrics Only)¶

The Grafana Cloud cost for logs and metrics is calculated by summing the costs for its individual components based on the defined workload and assumed data sizes, including the cost of extended retention.

Logs (Loki):
- Ingestion Volume: 3 billion log entries/month * 500 bytes/entry = 1,500 GB/month
- Ingestion Cost: 1,500 GB/month * ~$0.50/GB = ~$750 / month
- Retention Cost (90 days): Storing 1,500 GB for an additional 60 days is estimated to cost ~$0.30/GB. 1,500 GB * $0.30/GB * 2 months = ~$900 / month
- Subtotal (Logs): ~$1,650 / month
Metrics (Mimir):
- 1,000,000 active series (retention is typically longer for metrics and included)
- Subtotal (Metrics): ~$1,000 / month
Users:
- 5 users
- Subtotal (Users): ~$100 / month
Subtotal (Platform): ~$2,750 / month
Operational & Personnel Costs:
- Grafana Cloud is a managed service, but it still requires internal expertise to build dashboards, run searches, and manage data onboarding. This cost is considered part of the subscription's value.
Total Estimated Monthly Cost (Logs & Metrics): ~$2,750 / month

Therefore, for this comparison, we will focus on the costs of logs and metrics, acknowledging that the trace handling philosophies and associated costs are fundamentally different and not directly comparable on a per-event basis without considering sampling strategies.

Dollar-for-Dollar Comparison Summary¶

Category	Micromegas (Self-Hosted)	Grafana Cloud (SaaS)
Infrastructure Cost	~$1,000 / month	(Included in subscription)
Personnel / Ops Cost	~$2,500 / month	(Included in subscription)
Licensing / Subscription	$0	~$2,750 / month
Total Estimated Cost	~$3,500 / month	~$2,750 / month