SAP has changed its architecture from "old" SAP R3 / ECC to "new" SAP HANA. The "new" HANA is not that new -> it was made few years ago, but when it first came out it was quite undercooked. In addition due to the nature of ERPs, it takes often few years to upgrade from one version to another.
SAP HANA basically sits on a custom server and (as far as I know) just uses a column database. The hardware is setup/customized by SAP. Perhaps they want to squeeze some margin here too. The conspiracy theory is that it is not for speed, but to spy on their clients. Or that they dont want to deal with customers mis-configuring their servers.
As for the table structure, then most stuff sits in just big tables. I dont know if the system shards them internally; from user point the tables are just one entity, not divided by anything. You can query them without having to glue anything. For example the table AUFK has order data for all orders, in all entities, in all years. (on a side note: many other ERP systems make separate tables for each year, what is a pain in the butt for reporting, but also allows to shard the data easier).
SAP also has this concept of "MANDANT" which is kind of abstract concept of access rights, that can be compared to entity, but it is not really a financial entity (there is a field for that too). Perhaps internally the databse somehow shards the data on MANDANT level, but from user perspective you just see one big table.
And boy, some of those tables are big that you only extract deltas to your business warehouse.
>Does every location or subsystem have its own db that get some compared in the software layer or is there usually a single source of truth db somewhere?
The idea of SAP ERP (and I'd argue the idea of most ERP systems) is the ERP system is your source of truth. All its subsystems sit on one database, that (if you stay within standard) shouldnt ever break.
Many companies have other systems that "feed" data to SAP -> e.g. you do some inventory planning in some super-nice stuff, but the actual "bookkeeping" (source of truth, orders, financials) is in SAP.
SAP HANA basically sits on a custom server and (as far as I know) just uses a column database. The hardware is setup/customized by SAP. Perhaps they want to squeeze some margin here too. The conspiracy theory is that it is not for speed, but to spy on their clients. Or that they dont want to deal with customers mis-configuring their servers.
As for the table structure, then most stuff sits in just big tables. I dont know if the system shards them internally; from user point the tables are just one entity, not divided by anything. You can query them without having to glue anything. For example the table AUFK has order data for all orders, in all entities, in all years. (on a side note: many other ERP systems make separate tables for each year, what is a pain in the butt for reporting, but also allows to shard the data easier). SAP also has this concept of "MANDANT" which is kind of abstract concept of access rights, that can be compared to entity, but it is not really a financial entity (there is a field for that too). Perhaps internally the databse somehow shards the data on MANDANT level, but from user perspective you just see one big table.
And boy, some of those tables are big that you only extract deltas to your business warehouse.
>Does every location or subsystem have its own db that get some compared in the software layer or is there usually a single source of truth db somewhere?
The idea of SAP ERP (and I'd argue the idea of most ERP systems) is the ERP system is your source of truth. All its subsystems sit on one database, that (if you stay within standard) shouldnt ever break.
Many companies have other systems that "feed" data to SAP -> e.g. you do some inventory planning in some super-nice stuff, but the actual "bookkeeping" (source of truth, orders, financials) is in SAP.