Skip to content
Raw data
Logos / Order
MAKORLogos
Semantic layer
One graph across all sources
Own graph database · Petabytes

Data becomes statements.

Makor Logos builds a semantic layer over your data sources and integrates them into one shared, queryable graph. Databases, files and streams then speak the same terms, and every answer brings its provenance with it. The data stays in its sources or sits in the built-in graph database where speed matters.

RoleSemantic layer
IntegrationOne graph across all sources
Data storageSource or own graph
ProvenancePer result
Inference22 M facts/s
Designed forPetabytes
Principle

Three steps from the source to a verified statement.

01 / Map

Map sources to terms

Mappings describe how tables, files and streams point to classes and relations of the ontology. This turns many sources into one shared graph. The data stays in its systems or is taken over into the built-in graph database.

Data storage virtual or native
02 / Infer

Derive what holds implicitly

Logos adds what follows from the ontology: class membership, part-whole relationships, temporal sequence. The dataset itself remains unchanged.

Semantics formally defined
03 / Prove

Every answer with provenance

For every result, Logos supplies the chain: which source, which mapping, which rule. The audit trail starts at the first row, not at the report.

Provenance per result
Integration

All sources. One graph.

Logos connects databases, files and streams and makes them queryable as one graph. Whatever has to be available instantly sits in the built-in graph database, everything else stays in the source systems.

Connected sources remain the authoritative version of their data. Whatever sits in the graph database keeps its proof of provenance.

For you, it stays one query, no matter how many sources are behind it and how large they are.

What Logos connectsOverview
SourcesDatabases, files and streams, connected without moving them
Virtual
Graph databaseBuilt-in storage for data that has to be available instantly
Native
OntologyShared terms for all connected sources
OWL
ReasoningAdds what follows from the terms
Inference
QueryOne graph, one query, no matter how many sources are involved
SPARQL · GQL
Provenance comes with every resultMakor Logos
Many sourcesOne shared graph
Performance

Ultra-fast. Faster than every system in our benchmarks.

Ten graph, RDF and relational systems ran on the same machine with the same data. Every answer was checked against the expected result.

22 MInferred facts per secondNew knowledge from loaded data, peak value
75 sFor 50 M triplesLoading real data: read, encode, store
46 ms1,000 sourcesMerged into one answer
Friends of friends10 M edges · seconds
Makor Logos0.74 s
RDF triplestore A4.94 s
Leading graph database10.3 s
RDF triplestore B14.1 s

Two-hop graph join, same data, same machine. Shorter is better: 14 times faster than the leading graph database.

Measured with Makor Logos v0.5.0 on one node, 8 performance cores, 64 GB RAMBest of 3 to 10 runs · every answer checked
Big data

Made for semantic analysis of large data volumes.

Semantic tools often reach their limits with large datasets, while big data tools know nothing about terms. Makor Logos combines both: ontology, reasoning and provenance, distributed across a cluster and designed for data volumes in the petabyte range.

Big data
Semantic analysis
1015bytes with meaning
Runs in a cluster
One graph, all datasets

Runs in a cluster

Logos runs distributed across several machines. As the dataset grows, more nodes are added instead of a single server reaching its limits.

The whole dataset, not a sample

Analyses run across all the data, not across an extract that happens to be manageable.

Reasoning at scale

What follows from the ontology is available even across very large graphs, not just in the prototype.

Provenance at any size

Even in large datasets, every statement can be traced back to its source.

Source size does not matterCompute time per 10 rows transferred
10,000 rows0.29 ms
1 M rows0.35 ms
100 M rows0.37 ms
1 bn rows0.41 ms

Data in the source, not in Makor Logos. An answer costs as much as the answer is large, not as much as the source.

Projection

One petabyte in a cluster.

Extrapolated from measurements on one node, with a cautious scaling efficiency of 50 to 70 %. Not measured in a cluster.

NodesLoad 1 PBInfer over 1 PB
1001.7–2.5 days2.1–2.9 h
1,0004.2–5.9 h12–17 min
10,00025–35 min1.2–1.7 min
Provenance

Every answer, backed up.

ExampleResult with provenance

Vehicle F-112 belongs to unit V-4.

Inferred
SourceTable fahrzeuge, fleet management
MappingM-07 maps einheit_id to teilVon
FactSub-unit E-9 is part of unit V-4, from table einheiten
RuleteilVon is transitive
Queried14:42 UTC
Why it matters

Explainable, because the chain comes with it.

A result without provenance is a claim. Logos records which source, which mapping and which rule led to a statement. That keeps a decision explainable, even when nobody asks any more.

Stance

What Logos does not ask of you.

No migration

No data lake as a prerequisite, no moving of existing data. Logos reads through mappings from the systems you run today. The built-in graph database is an option, not a condition.

No proprietary schema

Ontologies in RDFS and OWL 2, validation rules in SHACL, queries in SPARQL 1.1 and GQL according to ISO/IEC 39075. Of the 981 tests in the W3C test suite for SPARQL, 899 pass and none fail.

No black box

What Logos infers follows from defined terms and rules. Every inference can be traced back to its source.

No third-country hosting

Developed, operated and hosted in Germany. No routing through third countries, no externally controlled services in the data path.

Name

Why Logos.

Logos means word, sense and order all at once, and it is in every “-logy”, including ontology. That is exactly what the semantic layer does: it gives scattered data an order and a sense.

Raw dataLogos / Order

Lo·gos

[ˈlɔɡɔs]noun
Origin
From Ancient Greek λόγος (lógos), word, speech, reckoning; from λέγειν, to gather, to pick up, to say.
1 / Philosophy
In Heraclitus, the principle that gives all events order and coherence.
2 / Language
Word and sense in one. It is in every “-logy”, including ontology, the study of being.
3 / Data
Makor Logos, the semantic layer of Makor: integrates data sources into one shared graph and gives them meaning. See also Makor Äther and Makor Atlas.
ReferencesHeraclitus, fragments B 1 and B 50
AI & agents · Role: Memory

Makor Logos and AI

For AI, Logos is the memory. Language models and agents draw on proven facts and defined terms instead of guessing. Every answer brings its provenance with it: which source, which mapping, which rule.

Talk to us about your data situation.

Bring a question that today fails because of scattered sources. We will show you how Logos makes it answerable.

Contact info@bonn-systems.com