Multiple Indexes / Multitenancy

[kevinsproles]

Multiple Indexes / Multitenancy

[cigolpl]

Hey @kevinsproles,

I've added multi-tenancy feature (#188)

It's possible now to run thousands of indices on one single machine. Each index can have a different product catalog and configuration (aggregations, sorting fields, etc)

More info in docs https://docs.itemsapi.com/guide/movies.html#multi-tenancy

[cigolpl]

It depends how many products you have and also workload of search and indexing. This use case is great. ItemsAPI works well with flexible requirement per index and was designed for such a cases

Personally right now I'd:

• deploy it by ansible (or terraform) to 100 x 10$ digitalocean instances. I'd set SWAP from SSD for about 24 GB RAM. I'd run 100 ItemsAPI servers / indexes on each DO instance

OR

• create all indexes on one ItemsAPI instance on high quality metal bare server (64 GB RAM, SSD or NVME). I think one such a server could handle 10,000+ sites. ItemsAPI works very well with tens of millions of records on small machine (except slowier indexing). If millions of records comes from 10K sites it could be even better because all indexes are smaller so searching and indexing is faster then one big index. The bottleneck might be a traffic to Node.JS API and not core c++ search engine itself.

The second solution is IMO best but I am not supporting multi-indexes from one ItemsAPI instance yet. I am going to implement it in the future. If I also add a RAFT (i.e. https://github.com/baidu/braft) on top of it then you could replicate it and make SLA on 99.99%