AN-362: log attribute ingestion

[am-toledo]

No description provided.

[am-toledo]

Cost analysis

Taking the case of only 1,000 segments:

Per 1,000 segments, the DLR product itself comes out to ~$2/month. The problem is the
metrics we compute for anomaly detection, each asset + attribute combination is a
separate billed series.

For those same 1,000 segments (~15 attributes across the whole DLR product) that's
~15,000 series, which would be ~$750/month at list price.

Our DataDog plan (Pro) includes 100 custom metrics per host. With 191 hosts that's
19,100 metrics included, and we're currently using 1,240 — so ~17,860 free.

So one line of 1,000 assets does fit, but it eats ~80% of the headroom. A second line
of 1,000 assets already pushes us past the allowance: 2,000 assets ≈ $600/month in
overage, 10,000 assets ≈ $6,600/month. (And Exelon alone is already close to 8,000
assets.)

So, this isn't a long-term fix.
Cardinality scales linearly with asset count. With a client as Exelon we'd cross the metric limit, and start paying a significant amount for information whose value isn't clear to me.