A practical example of HANA optimization
In this blog, I’ll try to recap my journey through the land of SELECT’s and internal tables towards the realm of HANA. I wanted to take a slow-performing ABAP, not something from a tutorial or playground module from SAP, but something “real”, tangible, a proper “pain point”, and describe how it can be converted to using HANA view(s).
It became clear to me while writing up my experiences that this is not a task easily undertaken. First of all, ripping apart an existing ABAP – no matter how “simple” and “straightforward” it can appear at first sight, is not an easy job. Secondly, wrapping it all up in a comprehensible blog, complete with bullet-proof points and comprehensible screenshots, is aneven more difficult achievement. At times, I was not even sure if this was doable, or whether anyone could recap any learnings from what I had been through. Surely, there are more straightforward tutorial and examples out there?
But then again, why not? I finally decided that the only format I could give this document, is in the way of “key learnings” and general tips. I learned a lot, and hopefully I’ll be able to pass on some of the ideas I had along the way to others. I’m not going to discuss the original ABAP code in all its gory detail, this would possibly help you to understand the inner workings of the original program but would take ages. Instead, I hope to convey the general idea of the program to you, and show how you can translate it into HANA. I’ll explain the three view types, attribute, analytical, and calculation, and show you how I used them (and, more importantly, why I used them). I’ll also show some of the pitfalls and tips along the way, to help you avoid the same.
And, finally, this is my way of doing it. That’s a disclaimer, by the way, not a nod to Sinatra. There might be (and likely is!) better ways of solving the particular issue I had, and better ways of avoiding pitfalls. There might even be new issues arising from the way I implemented my solution – feel free to comment. I’m here to learn as much as you are.
How to find a suitable candidate
Some people use speed dating, others register to various internet-based… oh, sorry. We were talking about HANA-optimization, not marriage prospects.
I found a nice program in dire need of being optimized by executing ST03N in a production system. ST03N is the workload monitor, sort of the “health display” of your Netweaver system. It has lots of cool features; among others is an option to look at the statistics for the most resource-demanding applications (or processes) running in that particular systems. Boy, was I happy when one of the top offenders turned out to be a custom-made ABAP! I quickly decided to rip that piece of underachieving code to pieces, replacing every sub-performing line with calls to some sleek & cool HANA views. I envisioned my re-worked ABAP/HANA hybrid development as a sleek Ferrari, blasting past the sluggish old ABAP original. And, of course, it would be a fun, easy-piecy job to do.
I would bitterly regret that decision.
Wow, it’s so…. BIG!!
Looking back, I can honestly say I made a huge mistake selecting this particular development for HANA optimization. Don’t get me wrong; I learned a lot along the way, but for a first-time project, something smaller and more manageable would have been far more suitable. I’ll repeat this in the key learnings session towards the end, but here goes: pick a development that spans 5-10 tables, not more. It’s far better to scale a small ridge first, before setting your sight on Everest.
And the problem with large ABAP developments (particularly old, somewhat chaotic ABAP developments) is that they’re unwieldy, erratic, chaotic, and obfuscated. Of course, this provides more opportunity for optimization, but you might just end up completely overwhelmed, and give up. Or have a heart attack. Or attempt to stick your laptop down the shredder (it won’t fit. Not even the newer, sleeker models with solid state drives. Trust me). A look at Mt. Everest from a distance of roughly 2.5 million milliseconds:
