Ho lokolloa ha DBMS Nebula Graph 3.2 e bulehileng e hatisitsoe, e etselitsoeng ho boloka ka katleho lisebelisoa tse kholo tsa data tse hokahaneng tse etsang graph e ka balang libilione tsa li-node le libilione tsa likhokahano. Morero o ngotsoe ka C ++ mme o ajoa tlasa laesense ea Apache 2.0. Lilaebrari tsa bareki bakeng sa ho fihlella DBMS li lokiselitsoe lipuo tsa Go, Python le Java.
DBMS e sebelisa mohaho o abuoang ntle le ho arolelana lisebelisoa (ho arolelanoa-ho se letho), e leng se bolelang ho qalisoa ha mekhoa e ikemetseng le e ikemetseng ea ho sebetsana le lipotso tsa graphd le mekhoa ea polokelo e bolokiloeng. Ts'ebeletso ea meta e hlophisa motsamao oa data mme e fana ka leseli la meta mabapi le graph. Ho netefatsa ho tsitsa ha data, ho sebelisoa protocol e thehiloeng ho algorithm ea RAFT.
Lintlha tse ka sehloohong tsa Nebula Graph:
- Ho netefatsa ts'ireletso ka ho thibela ho fihlella feela ho basebelisi ba netefalitsoeng bao litumello tsa bona li behiloeng ka mokhoa oa taolo ea phihlello ea karolo (RBAC).
- Bokhoni ba ho hokahanya mefuta e fapaneng ea lienjineri tsa polokelo. Ts'ehetso ea ho holisa puo ea tlhahiso ea lipotso ka li-algorithms tse ncha.
- Ho netefatsa latency e nyane ha o bala kapa o ngola data le ho boloka ts'ebetso e phahameng. Ha e lekoa ka har'a sehlopha sa node e le 'ngoe ea graphd le li-node tse tharo tse bolokiloeng, polokelo ea polokelo ea 632 GB, ho kenyeletsoa kerafo ea li-vertices tse limilione tse likete tse 1.2 le likarolo tse limilione tse likete tse 8.4, latency e ne e le li-milliseconds tse' maloa, 'me tlhahiso e ne e fihla ho likopo tse likete tse 140 motsotsoana.
- Linear scalability.
- Puo ea potso e kang SQL e matla le e utloisisehang habonolo. Ts'ebetso e tšehelitsoeng e kenyelletsa GO (bilirectional traversal of graph vertices), GROUP BY, ORDER BY, LIMIT, UNION, UNION DISTINCT, INTERSECT, MINUS, PIPE (ho sebelisa sephetho sa potso e fetileng). Li-index le mefuta-futa e hlalosoang ke basebelisi lia tšehetsoa.
- Ho netefatsa boteng bo phahameng le ho mamella ho hloleha.
- Ts'ehetso ea ho theha linepe ka selae sa boemo ba database ho nolofatsa tlhahiso ea likopi tsa backup.
- E loketse ts'ebeliso ea indasteri (e seng e sebelisoa meahong ea JD, Meituan le Xiaohongshu).
- Bokhoni ba ho fetola sekema sa polokelo ea data le ho se ntlafatsa ntle le ho emisa kapa ho ama ts'ebetso e tsoelang pele.
- Tšehetso ea TTL ho fokotsa nako ea bophelo ba data.
- Litaelo tsa ho laola litlhophiso le li-host host.
- Lisebelisoa tsa ho laola mosebetsi le ho hlophisa ho qalisoa ha mosebetsi (tsa mesebetsi e tšehetsoeng hona joale ke COMPACT le FLUSH).
- Ts'ebetso ea ho fumana tsela e felletseng le tsela e khuts'oane lipakeng tsa li-vertices tse fanoeng.
- OLAP segokanyimmediamentsi sa sebolokigolo bakeng sa ho kopantswe le batho ba boraro analytics platforms.
- Lisebelisoa tsa ho kenya data ho tsoa ho lifaele tsa CSV kapa ho tsoa ho Spark.
- Romella metrics bakeng sa tlhahlobo o sebelisa Prometheus le Grafana.
- Nebula Graph Studio web interface bakeng sa ho bona ts'ebetso ea kerafo, ho tsamaisa kerafo, ho rala polokelo ea data le merero ea ho kenya.
Tokollong e ncha:
- Ts'ehetso e ekelitsoeng bakeng sa ts'ebetso ea extract() ho ntša khoele e nyane e ts'oanang le polelo e fanoeng.
- Li-setting tse ntlafalitsoeng faeleng ea tlhophiso.
- E kenyellelitsoe melaoana ea ntlafatso ho tlosa opareitara e se nang thuso ea AppendVertices le ho tima ts'ebeliso ea lifilthara tsa bohale le tsa vertex.
- Palo ea data e kopilitsoeng bakeng sa ts'ebetso ea JOIN, hammoho le bakeng sa basebelisi ba Traverse le AppendVertices, e fokotsoe.
- Ts'ebetso e ntlafalitsoeng ea SHORTEST PATH le SUBGRAPH
- Kabo e ntlafalitsoeng ea memori (Arena Allocator e lumelletsoe).
Source: opennet.ru