Computing the daily average of a tag by scanning a year of raw points, every day, for hundreds of tags, is the kind of work that brings a historian to its knees. A rollup avoids it by doing the arithmetic once and storing the answer. It is a pre-computed periodic summary, kept alongside the raw data, so that long-range queries and routine reports read a small table of finished numbers instead of grinding through the full archive.
Data Rollup in one line: A data rollup is a pre-computed periodic summary of a tag, such as its hourly and daily average, minimum, maximum, and total, calculated once and stored alongside the raw history. Long-range queries and production reports read these small summary rows instead of recomputing over millions of raw points, so they return quickly. Rollups trade a little extra storage and some upfront computation for dramatically faster aggregate retrieval over wide time ranges.
A rollup takes the expensive aggregate calculations a report keeps asking for and performs them ahead of time on a fixed schedule. As each hour closes, the historian computes that hour's average, minimum, maximum, count, and total for a tag and writes them as a single summary row. As each day closes, it does the same for the day, often by rolling up the twenty-four hourly summaries rather than rescanning the raw data. The result is a compact ladder of summaries at increasing time granularity sitting next to the full-resolution history.
The payoff is at query time. Asking for the daily average of a tag over a year, against raw data, means reading and time-weighting every stored point in that year on the spot. Against a daily rollup it means reading three hundred and sixty-five finished rows. The heavy arithmetic was done once when each day closed, so the query that used to churn through the whole archive now touches a tiny fraction of the data and returns almost immediately.
Rollups are stored, not derived on the fly, which is exactly what distinguishes them from computing an aggregate live. An on-the-fly average is recomputed from raw data every time it is requested; a rollup is the saved answer, read back like any other stored value. This is why rollups are described as a storage-architecture feature rather than a calculation feature: the historian is spending disk space to buy query speed, deliberately and permanently.
The summaries a rollup keeps have to be the right ones for the questions that will be asked, and averages in particular must be time-weighted to be meaningful over compressed data. A rollup that stored a plain mean of stored points would bake the event-weighting error permanently into every report that reads it. Well-built rollups therefore integrate over time within each period, and they keep enough companion values, such as the count and the total duration, that higher-level rollups can be assembled from lower-level ones without returning to the raw data.
Not every aggregate composes the same way when rolling hours into days. A daily maximum is simply the largest of the twenty-four hourly maxima, and a daily total is the sum of the hourly totals, so both roll up cleanly. A daily average, however, cannot be found by averaging the twenty-four hourly averages unless each hour covered equal time and equal weight, so it is computed by combining the hourly areas and durations. Getting these composition rules right is what keeps a daily rollup consistent with the raw data it summarizes.
Rollups also underpin tiered retention. An operation may keep full-resolution raw data for a limited window and keep hourly and daily rollups for years, because the summaries are small enough to retain long after the raw points are aged out. This lets a historian answer detailed questions about recent history and coarse questions about distant history at very different storage costs, with rollups serving as the durable, low-cost long-term record.
Production reporting is where rollups earn their keep. A daily production report, a monthly summary, or a year-over-year comparison is fundamentally a request for periodic aggregates across many tags, which is precisely what rollups have pre-computed. Reading finished hourly and daily rows lets these reports generate in seconds rather than straining the historian, and it lets many users pull the same summaries at once without contention over the raw archive.
For distributed oil and gas operations, rollups also make dashboards over large fleets practical. Showing yesterday's average pressure or total volume for every site at once would be punishing if each figure were recomputed from raw data on demand. With daily rollups, the dashboard reads one small row per site, so a supervisor can scan a whole area's daily numbers instantly and drill into the raw trend only for the site that looks off.
Merobix maintains periodic rollups so long-range trends, daily and monthly reports, and fleet-wide summaries load quickly from pre-aggregated data while the full-resolution history stays available for investigation. An operator can open a month of a site's production at a glance because the monthly figure comes from stored rollups, then zoom straight into the raw, minute-level history when a number needs explaining, getting both the speed of summaries and the detail of the archive in one place.
An on-the-fly aggregate is recomputed from raw data every time it is requested, while a rollup is the pre-computed answer stored once and read back like any other value. Rollups spend some storage and upfront computation to make repeated long-range queries fast. On-the-fly calculation stays exact and current but re-does the heavy work on every request.
Not by simply averaging the twenty-four hourly averages, because that ignores how much time and weight each hour carried, especially with compressed or uneven data. A correct daily average combines the hourly areas and durations, which is why good rollups store companion values like counts and totals. Maxima and totals do roll up cleanly, but averages need this careful composition.
No, they sit alongside it. Rollups speed up aggregate and long-range queries, but the raw full-resolution data remains for detailed investigation, at least within its retention window. Many operations keep small rollups for years while aging out raw points sooner, so distant history is available as summaries and recent history is available in full detail.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.