Skip to main content

Event Metrics

An embedded, event-time metric collection library built for serving system

Metric systems like prometheus aggregate metric at “processing time”, whenever the scraper is able to scrape the metric. event_metrics capture and record metrics at event time, when the event happens.

Features

Comparing to other metric system, the event_metrics library:

  • Write data to sqlite3 database to keep low-memory footprint

  • Minimal dependency (only requires numpy for numeric aggregation)

  • Aggregate with the full data by default (no reservior sampling)

  • Allow select from past duration with timedelta windowing

  • Timestamp all observation by default

  • Small API footprint: observe and query is all you need to know

  • Compute raw timeseries with different aggregation strategies:

    • Scalers: last, min, max, mean, count, sum

    • Buckets for histogram

    • Percentiles for summary

    • Array and timestamps for native python wrangling

  • Metrics can be labeled with arbitrary key value pair and querying supports multidimensional label matching.

Install

  • Install from source: pip install -e .

  • PyPI package is work in progress

Usage

from event_metrics import MetricConnection

conn = MetricConnection("/tmp/event_metrics_demo")

conn.observe("latency", 1.2)
conn.increment("counter", -1)

# labeling
conn.observe("latency", 2.0, labels={"service": "myapp", "other": "label"})

# querying
(conn.query("latency", labels={"service":"myapp"})
      # select from past duration using one of the following
     .from_beginning()
     .from_timestamp(...)
     .from_timedelta(...)

      # perform aggregation using one of the following
     .to_scaler(agg="last/min/max/mean/count/sum") # counter, guage
     .to_buckets(buckets=[1, 5, 10], cumulative=False) # histogram
     .to_percentiles(percnetiles=[50, 90, 99]) # summary
     .to_array() # -> value array
     .to_timestamps() # -> timestamp array
     .to_timestamps_array() # -> 2 array, (timestamp, value array)
)

Speed

The library is fast enough. It can ingest about 34,000 data point per seconds:

You can run pytest tests -k bench to generate benchmark on local hardware:

-------------------------------------------------- benchmark: 1 tests --------------------------------------------------
Name (time in us)            Min       Max     Mean  StdDev   Median     IQR  Outliers  OPS (Kops/s)  Rounds  Iterations
------------------------------------------------------------------------------------------------------------------------
test_bench_ingestion     25.3340  297.9320  28.8541  9.1582  26.8090  0.8650   521;814       34.6571    6496           1
------------------------------------------------------------------------------------------------------------------------

Metadata

Release files for event-metrics 0.2.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for event-metrics 0.2.0
File Size Uploaded Provenance Actions
event_metrics-0.2.0.tar.gz 7.9 kB Without provenance Details for event_metrics-0.2.0.tar.gz

Release files / event_metrics-0.2.0.tar.gz

Download URL event_metrics-0.2.0.tar.gz
Size 7.9 kB
Tags Source
SHA-256 checksum
How to use checksums
92293d1afd45a5e26c1edbef785530434a52773e57116eb59c463805de9dd0fe
BLAKE2b-256 checksum
How to use checksums
6c0a131512008a0ef45899008e72c3a22264d680dc7d1b7414240c58fb225fe6
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/3.8.0 pkginfo/1.8.2 readme-renderer/32.0 requests/2.26.0 requests-toolbelt/0.9.1 urllib3/1.26.6 tqdm/4.62.2 importlib-metadata/4.2.0 keyring/23.5.0 rfc3986/1.5.0 colorama/0.4.4 CPython/3.7.11

Security for event-metrics 0.2.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

0.2.0 This release

1 release file No provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page