SAP HANA Architecture & Persistence

You're choosing the replication mode for HANA system replication to a secondary site. What does synchronous buy you, what does it cost, and when would you deliberately choose asynchronous?

A HANA system has hit its memory ceiling on a single host. You can scale up or scale out — what's the difference, and which one is the prerequisite for HA via standby nodes?

Which parameters define where HANA data and log volumes live, and in which file and section?

Your log volume is filling up even though full data backups run nightly. Given how HANA triggers log backups, why?

Why must you never relocate, rename, or delete HANA datafiles with OS commands, even with the database stopped, and what marker warns you?

What hard sizing requirement does an in-memory database (IMDB) like HANA impose that a traditional disk-based database does not?

You're modeling a table that is written constantly but almost never read analytically (transient, write-once staging data). Why might row store be the right choice, and how does HANA otherwise soften the write cost of column tables?

You fail over to a HANA system-replication secondary that has table preload disabled. The takeover succeeds in seconds, but users complain reports crawl for the next hour. What happened?

HANA redo logs are written as sequentially numbered log segments. Why does forgetting to schedule log backups eventually halt the database?

What is the SAP HANA service port schema for instance number xx?

Stakeholders treat each HANA tenant as a fully independent system. In which respects are tenants genuinely isolated, and where does that isolation stop?

What is the SAP HANA administrator OS user, and what are its groups?

You need backups to land in an enterprise backup tool with minimal performance impact on HANA. Of the three destinations (file system, storage snapshot, Backint), which two are candidates and how do they differ?

An ABAP-trained admin wants to change a HANA database parameter and goes looking in the instance profile directory. Why is that the wrong place, and where does the parameter actually live?

HANA's main compression is dictionary-based encoding. Given how it works, which kind of column compresses well and which barely compresses?

On a HANA host with three tenant databases, how many index servers (hdbindexserver) do you expect, and how does hdbdaemon's job differ from the index server's?

A security team asks you to enable high isolation mode on a multitenant HANA so one tenant's admin cannot touch another tenant's files. What does high isolation actually change, and what does it NOT give you?

If HANA writes savepoints only every five minutes by default, doesn't that mean you can lose up to five minutes of committed transactions on a crash? Explain.

A reviewer says 'HANA is columnar, so delete all the indexes.' Where are indexes on HANA column tables still mandatory, and when do performance indexes still earn their keep?

RAM is volatile, yet HANA survives a crash without losing committed data. Which two on-disk structures make that possible, and what happens to in-flight transactions?

A change manager wants a maintenance window to 'take the database down for a clean backup.' Why is that unnecessary for HANA, and what two other things are always true of a HANA backup?

When HANA reads a column table, why must it also consult delta storage, and why is the resulting inefficiency tolerable?

Why is row storage good for OLTP but poor for OLAP, and what does HANA's column store let you stop maintaining?

Why can't you run your business application in SYSTEMDB, and which administrative tasks force you to connect to SYSTEMDB rather than a tenant?

A HANA service process crashes and is back within seconds. Which component restarted it (sapstartsrv or hdbdaemon), and what is the actual startup chain?

During a HANA recovery, in what order do you apply full, differential, incremental, and log backups, and what's the easy-to-get-backwards step?

On a HANA host one process shows the largest memory footprint and highest CPU. Is that the nameserver or the indexserver, and what does the other one actually do?

You take a full backup Sunday, then deltas daily. With differential vs incremental, what exactly does Wednesday's backup contain, and which choice makes restores simpler vs backups smaller?

Why run multitier system replication with a near secondary AND a remote third system, instead of one remote replica that 'covers everything'?

Since HANA 2.0 SPS 01 you can only install multitenant, with each SAP system in its own tenant. What's the maintenance consequence that should shape which systems you co-locate?

An admin assumes /usr/sap/<HSID> holds the HANA installation, as it largely does on an application server. Why is that assumption misleading for HANA?

You're about to use application function libraries (AFL) on a fresh HANA and they fail to load. Which of the compile/preprocessor/script servers did you probably forget, and how do these three differ?

You deleted a large volume of data from HANA expecting the filesystem to free up, but datavolume_000.dat is the same size. Why, and what do you have to do?

After a crash HANA rebuilds from the last savepoint plus log replay. Given that, what does the savepoint interval actually trade off, and what is it NOT protecting?

Today SAP HANA picks among several SQL execution engines to plan and run a statement. Where is SAP heading architecturally, and what shouldn't you assume yet?

After a HANA restart, the reload (ramp-up) phase can be counted as business downtime while column-store tables reload into memory. What two optional techniques speed startup and shorten that phase?

With the SAP HANA Local Secure Store (LSS), your encryption keys ride along inside the regular database backup. What must you set so a restore actually succeeds?

SAP HANA XS Classic is still auto-installed, yet you're told both to migrate off it and not to remove it. How do you reconcile that?

Beyond the Row Store and Column Store, how does SAP HANA natively store and query JSON documents?

You want to place the SAP HANA data and log volumes on the modern btrfs filesystem. What stops you, and what is supported instead?

Which large S/4HANA tables does SAP already pre-configure as good Native Storage Extension (NSE) candidates, and since when?

A colleague plans to tier rarely-accessed data in a new S/4HANA system using Data Aging. Why does SAP now steer you toward Native Storage Extension instead?

When is a scale-out SAP HANA deployment actually permitted for S/4HANA, rather than just running everything on one bigger host?

A colleague assumes Native Storage Extension has to be switched on manually in S/4HANA. Is that correct, and how many tables come pre-classified for it?

On a fresh S/4HANA system you check which technical database user the ABAP work processes log on with and see a name different from what older documentation taught. What is the current default, and why does the name keep changing?