17
余談
Why Mon-aaS 2.0 will adopt Cortex
Point of View Cortex Thanos
architecture complex simple
multi-tenant support native support new cluster for each tenant
architecuture with multi-tenant complex x 1 simple x N
management only need to manage single cluster need to manage all clusters
scale out
easy to scale-out by adding nodes or
replicas
can't simply scale-out compactor,
ruler and receiver
query reliability
only need to access ingesters and
NoSQL
need to access all receivers and store
gateway
query speed fast (have cache) slow
retention policy global only per tenant
storage efficiency store single data store N replicas
• Query-Speed
• Scalability
• Efficiency (results in reducing cost and ease-to-manage)
Deciding factors for that are:
下記は、当時(約2年前)のCortexとThanosを比較したもの。現在、CortexとThanosは双方の良さを補完しあっ
ており、最終的には一つになりそうな予感です。
例えば:
• クエリキャッシュ
• Cortex → Thanos
• リテンション
• Thanos → Cortex