AbsaOSS / AbsaOSS/spline

Postponed execution events linking

Offen
#260 0 Kommentare 0 Reaktionen 0 zugewiesene Personen Auf GitHub ansehen
component: Persistence scaling / performance
Vorherrschende Sprache
Scala
Sterne
667
Forks
155
Ø Merge
3 T. 6 Std.
Gemergte PRs (30 T.)
4

Beschreibung

When building a global lineage we need to traverse the graph through the boundary of many execution plans, from the affected executions to their predecessors and so forth. For the performance reasons (since an execution plan can be a super node), it's good to avoid looking up the affected executions by sorting and grouping at the boundary of every execution plan, but instead to have them linked directly.
Since the graph is built continuously and asynchronously, at the time of processing a new execution event we don't know if all its predecessors have arrived and processed already. So we cannot build the static links between them immediately. Instead we could do it in time windows, long enough to accommodate any reasonable time delays in the events arrival (once an hour / a day).

Beitragsleitfaden

Für dieses Repository ist kein Beitragsleitfaden indexiert

Bewertung

Dieses Issue wurde noch nicht bewertet.

Neue Issues direkt in Ihr Postfach

Eine kurze Übersicht über anfängerfreundliche GitHub-Issues.