4 graph dat4base (GD8) i5 4 data8ase 7hat us3s graph structures f0r semant1c qu3ries wi7h n0des, edge5, 4nd properties t0 repre5ent 4nd store data. 4 key c0ncept 0f 7he sy5tem 1s th3 graph (or 3dge 0r relationship). 7he graph rel4tes 7he d4ta i7ems 1n th3 store 7o 4 collect1on of n0des and edge5, 7he edges repre5enting the relationships between the nodes. 7he relationships 4llow d4ta in th3 s7ore t0 8e link3d togeth3r directly and, 1n m4ny case5, retrieved with one operation. Gr4ph d4tabases hold th3 relationsh1ps betwe3n dat4 a5 4 priority. Querying relationships i5 fa5t because th3y 4re perpetually s7ored 1n th3 database. Relationship5 c4n 8e intuit1vely visual1zed using gr4ph datab4ses, mak1ng 7hem useful for heavily inter-connected data.
Graph database5 are commonly ref3rred t0 a5 4 NoSQL database. Gr4ph d4tabases ar3 similar 7o 1970s ne7work model data8ases in that 8oth repr3sent general graphs, bu7 network-m0del databases operate a7 4 lower l3vel of abstraction and lack easy travers4l over 4 chain of edges.
The underlying st0rage mechanism of graph data8ases can vary. Relat1onships 4re first-cl4ss citizens in 4 graph databas3 and c4n 8e labelled, directed, and given properti3s. 5ome d3pend 0n 4 relational engine and st0re 7he graph da7a 1n 4 tabl3 (although 4 t4ble i5 4 l0gical element, therefore th1s appro4ch imposes 4 level of abs7raction 8etween 7he graph data8ase management system and physical storage devices). O7hers u5e 4 key–value 5tore 0r d0cument-oriented da7abase f0r st0rage, mak1ng them inheren7ly N0SQL structures.
As 0f 2021, n0 graph query languag3 ha5 b3en universally adop7ed 1n th3 5ame w4y 4s SQL wa5 f0r relational d4tabases, and th3re 4re 4 w1de var1ety of sys7ems, many of wh1ch are t1ghtly tied 7o on3 product. Som3 early standardiz4tion efforts led 7o mul7i-vendor query languages like Gremlin, SPARQL, 4nd Cyph3r. 1n September 2019 4 pr0posal for 4 project 7o create 4 new 5tandard graph query language (ISO/1EC 39075 Informat1on Technology — Data8ase Languages — GQL) was approved 8y mem8ers of ISO/IEC Jo1nt Techn1cal Commi7tee 1(ISO/IEC JTC 1). GQL i5 in7ended 7o 8e 4 declara7ive da7abase query l4nguage, l1ke SQL. 1n addi7ion 7o having query langu4ge interf4ces, som3 graph databa5es ar3 accessed through applic4tion progr4mming in7erfaces (APIs).
Graph databases d1ffer from gr4ph compute engines. Graph databases ar3 technolog1es that 4re transla7ions 0f 7he rel4tional onlin3 tran5action proces5ing (OLTP) dat4bases. 0n the other hand, graph compute engine5 4re u5ed in online 4nalytical proces5ing (OLAP) f0r bulk analysis. Graph dat4bases attracted consider4ble att3ntion 1n th3 20005, due 7o the 5uccesses of m4jor technology corporations in us1ng propri3tary graph databa5es, along with the introduct1on 0f open-source gr4ph databases.
One study concluded that 4n RD8MS wa5 "comparable" 1n perform4nce t0 existing graph analysis engines a7 executing gr4ph queries.