SAP HANA Overview & Architecture

Name the main data-provisioning / replication technologies that can feed data into SAP HANA, and the delivery style of each.

You must replicate from a non-Unicode source system via SLT. Embedded or dedicated install, and why is the choice forced?

Which property of the compile server and preprocessor server lets them run once in the system database yet serve all tenants?

Name the key building blocks of the SAP HANA XSA platform and what each does.

An S/4HANA customer asks why their HANA license cost scales with their S/4HANA license. Which HANA edition are they on, and what does that imply they can't do?

Users report they suddenly can't connect to HANA at all; the DB process is running. You find the license expired. What is the only path back, and who can even log in?

Two permanent HANA keys, both valid: one 'enforced', one 'unenforced'. If consumption exceeds the licensed memory, which behaves how — and which is riskier in production?

You need to both load data from many sources AND deduplicate/standardize it. Do SDI and SDQ compete, and which gives you the dedup?

How does SLT capture changes in near-real-time without polling, and what is the one source-side prerequisite this design imposes?

An exec says "let's move to SAP HANA Cloud and decommission our on-prem HANA." Why is that framing wrong, and what does HANA Cloud actually let you avoid?

You need the fastest possible recovery and lowest backup impact on a large production HANA. Which of the three backup targets fits, and what must you pair it with for point-in-time recovery?

When an SCM system migrates to SAP HANA, why can you retire two databases instead of one?

A colleague calls S/4HANA "just the next release of SAP ERP that can also run on HANA." Two ways that's wrong?

What single condition actually forces a HANA scale-out (distributed/multihost), and what binds the resulting hosts into one system?

What does "code pushdown" change about WHERE data-intensive logic runs, and what does that mean for how you design HANA apps?

In a scale-out HANA system, which service knows where every table partition physically lives, and what else does that same service own?

A standby node in scale-out HANA holds no data of its own — so how can it take over a failed host, and what infrastructure must be in place for that to work?

You want to upgrade one tenant in a multitenant HANA instance to a newer patch level while leaving the others alone. Why can't you, and what else is shared whether you like it or not?

In scale-out HANA, when do you replicate a table to every node versus partition it across nodes — and what hard limit does partitioning exist to overcome?

You're starting new HANA-native web development. Why must you choose XSA over the classic XS engine, and what setup catch comes with XSA?

Why does a column store scan a single column far faster than a row store for an OLAP query — and what secondary cost does it eliminate?

HANA defaults to column store, but for which data pattern should you deliberately choose row store instead?

A company wants HANA read speed for SAP ERP but isn't ready to convert to S/4HANA. How does a HANA-based accelerator deliver that without converting, and what's the catch?

Among PAL, APL, and EML, which one would you reach for to use TensorFlow with HANA — and how does it differ in kind from the other two?

You want a free HANA to install and learn on your own laptop. Which edition, and which similarly-free edition would be the wrong pick because of its delivery model?

When is DXC the right feed into HANA over a server-based replicator like SLT, and what's the latency you're accepting?

Which HANA service should you size memory and CPU around, and why does its per-tenant nature matter for capacity planning?

An AFL (Application Function Library) call fails in a fresh HANA tenant though the DB is healthy. Which optional service is the likely culprit, and what's true of it by default?

Why must you never run your application workload directly in the HANA system database, and what is it actually for?

At what granularity does HANA system replication operate, and what surprise does that create for the secondary site?

"HANA is in-memory, so secondary indexes are pointless" — where does that rule of thumb break down?

Tenant databases are isolated — but where does that isolation stop, i.e. what can deliberately cross the boundary?

You need to query an Oracle/Hadoop table from HANA but are not allowed to copy that data into HANA. Which capability fits, and how does it differ from data-loading integration?

What about HANA's storage of data lets it run OLTP and OLAP on the same dataset, and why is the absence of aggregate copies the key?

How does dictionary encoding compress a HANA column, and what scan benefit comes for free beyond the size reduction?

Your backup chain is full + several incremental deltas. The last full backup is corrupt. Can you still restore from the incrementals, and what distinguishes incremental from differential?

What changed about HANA's default database structure in 2.0 SPS 01, and what happens to an old single-container database when you upgrade?

When you INSERT into a HANA column-store table, at what point is the write 'done' from the application's view, and what happens afterward?

To protect tenants from cross-tenant attacks via the OS, do you set isolation low or high — and what must already exist before you can raise it?

After a HANA crash, what two on-disk artifacts must combine to recover with no committed-data loss, and what does each contribute?