M365Con 2027 – free online conference · 11–22 Jan 2027 · 120 sessions m365con.net

2026-11-21

SAP HANA Database – Simply Explained

Follow the showNew episodes on Microsoft 365, Copilot and more, free in your podcast app.Apple PodcastsSpotify
Cover: SAP HANA Database – Simply Explained
SAP rebuilt its entire flagship product line around one database, built on a simple bet about where data should live. That bet made SAP systems much faster and much more expensive to run. In this episode of Simply Explained you will learn: - What SAP HANA really is and why keeping data in memory…

Show notes

SAP rebuilt its entire flagship product line around one database, built on a simple bet about where data should live. That bet made SAP systems much faster and much more expensive to run. In this episode of Simply Explained you will learn:

  • What SAP HANA really is and why keeping data in memory changes everything
  • How column storage, compression and the delta merge make queries faster
  • What multi-model means: tables, text, spatial, graph, JSON and vector search
  • Where HANA runs: on-premises, SAP HANA Cloud and Microsoft Azure
  • How HANA connects to Power BI, Microsoft Fabric and Azure Backup, and where the limits are
  • The real bill and the common mistakes: sizing, licensing, old habits and reporting on production

Simply Explained by M365.FM breaks down Microsoft 365, Azure and modern IT topics in clear, jargon-free language. Hosted by Mirko Peters. Become a supporter of this podcast: https://www.spreaker.com/podcast/m365-fm-microsoft-365-azure-cloud-power-platform-ai--6704921/support. This episode includes AI-generated content.

Host: Mirko Peters

Guests

Guests in this episode:

Related articles

Articles on the topics of this episode:

Transcript

Transcript (TXT)

Read the full transcript

Speaker 1: SAP rebuilt its entire flagship product line around one database. That database is SAP HANA, built on a simple bet about where data should live. The bet made SAP systems much faster, and it also made them more expensive to run. So let's go through what HANA actually is, how it works, where it runs, including on Microsoft Azure, and where it costs you. We'll keep the jargon out and stick to the plain version. Let's start with what a database really is because the word gets thrown around a lot. A database is an organized place to keep information and ask questions of it. You put data in and later you ask things like how many orders did we ship last month or which customers are behind on payment. That's the whole job. Most databases keep that information on a disk or a solid state drive and they pull what they need into memory only when a question arrives. The trip to the disk is the slow part and on a busy system that happens constantly. SAP HANA flips that around, it keeps the working data in the computer's memory, the same fast memory your laptop uses to run programs, and it asks the disk for help only when it has to. When a sales manager asks for this quarter's numbers, the answer already sits in the fastest place the machine has. SAP describes HANA as a multi-model database that stores data in memory instead of keeping it on a disk, and that one sentence is the heart of it. The idea came from SAP co-founder Hasso Plattner, who spent the mid-2000s pushing for a database that could handle transactions and analysis at the same time and answer business questions in real time. HANA shipped in 2010, and the timing mattered. Memory had gotten cheaper and processors had gotten wider, which made keeping everything in memory practical instead of exotic. That difference is easy to feel in practice. In a classic setup, a report that scans a year of transactions can run for minutes while the disk feeds data to the processor. In memory, the same scan works on data the processor can already reach, so the weight shrinks to something a person barely notices. That's the bet SAP made, and it's why the company rebuilt so many of its products around this database. Now we get to the honest part, the one that shows up later in your budget. Memory costs much more than disk and it forgets everything the moment the power goes out. So HANA still writes your data to disk for safety and it moves older, colder data out of memory to keep the cost down. SAP calls this multi-tier storage and it's how the system trades speed against price. So the first big idea is a simple keep the data where the processor can reach it fastest. There's a second idea behind it and it's the one most people never hear about. Picture a table of orders. Each order has a customer, a date, a region, and an amount. It's a simple grid, and the interesting question is how you lay that grid out in memory. That layout is the second idea I mentioned, and it turns out to matter as much as keeping the data close. A row store keeps every order together, so one order's customer, date, region, and amount sit right next to each other. A column store does the opposite. It keeps all the amounts together, all the regions together, all the dates together, That difference changes what a query costs. Say you want the total of every amount. In a column store, the machine reads one clean list of numbers and adds them up. In a row store, it has to walk past the customer, the date, and the region in every single record just to reach the amount. On a table with millions of rows, that gap is enormous, and SAP describes the column design as a way to avoid costly full table scans. Columns help a second way too, through compression. Values sitting next to each other in a column tend to look alike, and similar values squeeze down far better than a mixed row of names, dates, and numbers. A smaller data footprint means less memory to buy, which loops straight back to that cost question. The compression isn't a small detail either. A column full of repeated region codes or dates can shrink to a fraction of its original size, and every byte you don't store is a byte you don't have to pay for in memory. That's how the column layout saves money as well as time. There's one more piece you'll hear about and it's called the Delta Merge. New records can't be jammed straight into a heavily compressed column because that would be slow. Sororna writes new data into a fast write-friendly area called the Delta Store, then merges it into that compressed main store in the background. SAP documents it plainly, writes land in Delta Storage, and the merge moves them into the read-optimized form. That merge is real work, and on a busy system, it never really stops. Column store isn't magic for every job either. If you're looking up one specific record by its key, a row store can be the better fit, which is why HANA keeps both and lets you choose per table. So the two big ideas stack up, keep data in memory, and lay it out in columns. Together, those are the engine. Once you have that engine, you can do more with it than store plain tables. SAP calls HANA a multi-model database, which is a fancy way of saying it handles several kinds of data in the same place. The ordinary kind is relational data, which is just tables with rows and columns, the same thing business systems have used for decades. Then there's text, so you can search inside documents and notes with normal, equal, there's spatial data for anything with a location on a map. There's graph data, built for relationships, like which supplier feeds which plant, and there's JSON for the looser semi-structured data that turns up in modern applications. That matters because it means one question can pull from all of them at once. Find the nearest warehouse to Frankfurt that belongs to a certified supplier, has low carbon tax rates, and isn't flagged for customs delays. That example comes from SAP's own explanation of multimodal queries, and the point is that location, relationships, and regular business data get answered together instead of across three separate systems. SAP has pushed HANA into AI work as well. SAP HANA Cloud includes a vector engine, which finds things that mean the same thing rather than matching exact words, and a knowledge graph engine for relationships. SAP says those support common patterns like VectorRack and GraphRack, which are ways of giving an AI model real company data to reason over. There's also a built-in application server for small services and web apps, plus machine learning functions you can call from inside the database. One honest note is worth making here, multi-model means you can put a lot of different data in one database, it doesn't mean you should pile everything in there. Every kind of data you add is another thing to model, secure and maintain. So the question to ask is which of these models your business actually needs, not how many the engine supports. You understand the engine now, so let's look at where you put it. There are three main answers and the third one is where Microsoft comes in. The first is your own servers on-premises. You buy the hardware, you install HANA and your team runs it. That gives you the most control and it also gives you the most work. The second is SAP HANA Cloud, which SAP describes as a managed database service. You don't install and patch the engine the way you would on your own team because SAP runs it and you consume it. SAP announced HANA Cloud in 2019 and launched it in 2020 as its next-generation database platform as a service. The third is a public cloud, and this is where Microsoft enters the picture. SAP HANA can run on Microsoft Azure. Azure's M-series virtual machines are certified for SAP HANA, and Azure also offers HANA large instances, which are dedicated bare-metal servers rather than shared virtual machines. That word certified carries real weight. SAP keeps a hardware directory that lists exactly what's allowed for production, so running HANA on Azure isn't a guess or a workaround, it's a documented, supported path. Teams still have to pick a size from the certified list, and that choice ties straight to cost and performance. On paper, this sounds simple. In a real enterprise environment, there are a few more things to consider. The network between the SAP application layer and the database layer needs to be direct because edit hops slow down a system that lives on speed. Storage has to be chosen carefully for the data and log volumes. Clock accuracy matters more than you'd expect to because SAP HANA depends on reliable timing to operate correctly. There's one more angle worth naming because it changes the conversation for a lot of teams. If your company already runs Microsoft infrastructure, HANA on Azure means one less platform to learn. Your existing network, identity, and backup tools can reach the database automatically. which shortens the gap between the SAP team and the platform team. The limit to keep in mind is this. Putting HANA on Azure is real, but it's still a serious infrastructure project. The network design, the storage layout, and the size you pick all shape what you get, and getting them wrong is expensive to undo. And when you connect this database to the Microsoft tools your team already uses, there's one limitation that catches people out, which is exactly where we're headed next. If HANA can sit on Azure, The natural follow-up is how it works with the tools your teams already use. Two everyday bridges do most of the work, and both are well-documented. Start with Analytics, Power BI, and Microsoft Fabric both have a native SAP HANA database connector, so you can pull HANA data into reports and pipelines without building a custom integration. In practice, you point the connector at the HANA server and its port you authenticate, and then you choose how to bring the data across. Import copies the data in and works on it locally. DirectQuery asks HANA at the moment someone opens the report. The second bridge is operational. Azure Backup supports SAP HANA database backup, which sounds dull until you remember that a database you can't restore isn't really production ready. This is the part I hinted at earlier because the connector has real limits. Power Query Online currently supports basic authentication only. The connector needs an on-premises data gateway, even when the data lives in the cloud. And it doesn't support connecting to a cloud database through a proxy. So if your network forces traffic through one, You'd fall back to the generic ODBC connector instead, a connector that talks to almost any database. Version details matter too. Microsoft documents are support for specific HANA versions, so check the current support list before you promise anyone a connection. Tools and connectors change, and a compatibility node from two years ago may already be stale. One more caution, and this one causes trouble in real projects. Pointing a reporting tool at a live production HANA database is supported, but that database is also running the business. A heavier report can compete with the transactions your company depends on, so teams often connect to a replica or an analytics layer instead of the live system. So let's see what all of this gets used for. The biggest answer is SAP's own flagship product, SAPS-for-HANA. That's the modern ERP system, the software that runs a company's core operations, and it's built to run on HANA. If you're an SAP consultant or developer, that single fact is why this database shows up on your projects whether you asked for it or not. Our whole family sits around it. SAP BW slash for HANA is the data warehouse built for HANA. SAP Analytics Cloud and Data Warehouse Cloud run on HANA Cloud for reporting and planning. The database isn't just one product's engine, it's the foundation under a big part of SAP's portfolio. Take a plain scenario that shows why people care. A sales order lands in the system, and because the transactional data and the analytical data live in the same place, a report on that order doesn't have to wait for a nightly batch job to finish. The number is fresh when you look at it. That's the promise, and it's a clear shift from a classic warehouse that catches up overnight. The scale is real too. SAP says more than 31,000 direct customers run on SAP HANA, so this isn't a small corner of the market. Now step back for a moment, because this is where the Microsoft and SAP worlds meet. In a mixed estate, the picture looks like this. SAP holds the core business processes and the system of record. Microsoft tools handle reporting, collaboration, and much of the cloud work around it. HANA is where the trusted numbers live, and the Microsoft side is often where people actually look at them. There's a practical consequence to that split when the trusted numbers live in HANA and the reports live in Microsoft tools. Someone has to own the connection between them. That usually means agreeing on which data moves, how fresh it needs to be, and who is allowed to see it. And that agreement is often more work than the technical connector itself. The limit is worth stating plainly. The real-time story holds when the data does sit in HANA. If half your data lives in other systems, you still need integration to pull it together, and that's exactly where complexity creeps back in. The database is fast, but it can't be fast with data it doesn't have. Every good story has a bill, and this is where HANAS comes to you. The cost driver is memory, because that's where your working data lives. Memory costs far more than disk, so the size you choose is the single biggest lever on what you spend. That makes sizing a discipline, not an afterthought. SAP treats memory sizing as a required step, and it warns that you need to plan for the data that sits still and the data that moves around during processing, plus disk space for the persistent copy. Get it too small and the system struggles or can't fit the workload. Get it too large and you've paid for memory that just sits there. Then there's licensing. HANA is licensed software and the licensing model has its own rules and its own conversations. I won't quote numbers here because published licensing figures change and a stale number would only mislead you. Check the latest terms with SAP or your partner before you budget. The delta merge adds an ongoing cost that's easy to forget. Every time HANA folds new data into the compressed main store, it spends processing power and a workload with a constant stream of small writes keeps that engine busy. It's the price of the compression you get, and on a heavy system, it's something you have to plan for. Lifecycle is another one. HANA versions have maintenance windows, and you're expected to move between them. SAP HANA 2.0 SPS 08 reached general availability on November 20, 2024. with four years of maintenance, which means upgrade planning is part of running the system, not something you do once and forget. The data itself carries a cost that people forget. Every year of history you keep in the fast tier is memory you keep paying for, and most companies keep far more than they ever query. Cleaning up old and duplicate records before you move them isn't glamorous work, but SAP recommends it, and it can shrink the footprint enough to change which size you have to buy. The cheapest memory is the memory you never needed in the first place. And then there's the human cost, which is the sneaky one. Moving to HANA means learning new sizing, new operations, and sometimes rewriting logic that assumed a disk-based database underneath. That's people time, and people time's often larger than the hardware line on the invoice. So the bill isn't just memory. It's memory, licensing, ongoing maintenance, and the skills to run it well. None of that makes HANA a bad choice. It just means the speed you're buying has to be worth all of it. Most of those costs show up in the same handful of mistakes and they are worth naming so you can dodge them. Start with the biggest one, which is skipping sizing and buying by feel. It's the most expensive mistake and the easiest to prevent because SAP publishes the guidance and your partner can run the numbers before anyone signs a purchase order. Next comes treating HANA like the old database. Copy old row-based thinking, old code patterns and old overnight batch jobs straight onto a new engine. and you spend the money while keeping the old performance. The point of moving is to do things differently. Then there's the habit of keeping everything forever. Pouring years of stale data into expensive memory is a choice, and it's the wrong one when multi-tier storage exists for exactly this reason. Move the cold data down and keep the hot data fast. Another one is forgetting high availability and disaster recovery. HANA supports system replication, backups, and failover, but none of that switches itself on. You configure it, you test it, and you find out whether it works before you need it. The last one is assuming a reporting tool can hammer production. Connecting Power BI or Fabric straight to a busy HANA system is allowed, but it competes with the business processes running there, so point the heavy reporting somewhere else. None of these are reasons to avoid HANA. They are the checklist that separates a rollout people are happy with from one that turns into a long, expensive argument. Most of them come down to the same habit, which is planning the size and the workload before you spend the money. Let's pull the thread together. SAP HANA keeps your working data in memory rather than on a disk, and it lays that data out in columns. So scanning and adding up millions of records costs far less. It handles more than plain tables. and it runs on your own hardware in SAP's cloud or on Microsoft Azure, where it's certified to run on Azure virtual machines. The speed is real, and so is the bill, because memory is the expensive part, and sizing it wrong is the mistake that hurts most. Here's my question for you. What's the one piece of SAP HANA you'd like unpacked next? The in-memory design, the column store, cloud versus on-premises, or the Microsoft side? Tell me in the comments, and I'll build the next episode around the answer. If this made the topic feel less intimidating, subscribe to M365FM to master the tools and the mindset shaping the future of work.