Differences Between SAP HANA and S4HANA
Differences Between SAP HANA and S4HANA
Differences Between SAP HANA and S4HANA
S/4HANA
Publicado el 20 de abril de 2018
Taryck Bensiali
SeguirSeguir a
Senior SAP Architect at Sopra Steria
SAP S/4 HANA is the successor of the R/3 (ECC) ERP on SAP HANA
In-Memory
Column store
Compression
Partitioning
No longer need of aggregate table
Insert Only
This database R&D has/will be financed by the 15% (average) of licence gather by SAP
that are redistributed to Database vendors (mainly Oracle).
This database features have been enhanced with other services to build the SAP HANA
Platform. Source.
Application Services
Database Services
Most of these services require a licence (not included within the runtime version).
Integration Services
Most of these services require also a licence (not included within the runtime version).
Runtime edition: You pay for a SAP product that require SAP
HANA, which is included (as a database mainly) in your product
licence costs. No additional fee.
Full edition: You pay for SAP HANA whatever you store within
this SAP HANA (either a SAP product or your own developments)
So, having a SAP Product installed on a non-runtime edition of SAP HANA, means in
some way to pay twice for SAP HANA (at least the database side)....
Zoom on S/4HANA
S/4 HANA is the successor of SAP R/3 (ECC) and is based on the possibilities provided by
SAP HANA, that why it's name contains both S/4 (R/3 related) and HANA (HANA
Database related). Please note that it is considered as a new product, which means that
upgrading from R/3 (or ECC) to S/4 leads to buy new licences. Around five times
your maintenance cost.
S/4 HANA runs only under SAP HANA where SAP R/3 (ECC) was database agnostic.
The main objectives of S/4 HANA is to have a simplified data model (versus SAP R/3 or
ECC) and to provide Analytics (OLAP) and Transactional (OLTP) on the same
system/data in real time.
S/4 HANA start with "Simple Finance" module, which provide a simplified data model
for R/3 (ECC) FI & CO modules. "Simple Finance" is considered now to be stable and
efficient.
The "Simple Logistics" module of S/4 HANA is almost all other SAP R/3 (ECC) Modules :
MM, SD, PP, PM, WM, ... At this moment, this module provide interesting new features,
but also some drawbacks, which prevent some customers to switch to "Simple Logistics".
Despite S/4 HANA and HANA looks like very similar regarding their names, it is
two distinct products that are connected, especialy in term of licence. Think twice
before chosing you licencing model.
SAP HANA data storage is organized to be optimized for read access. Each time
changed are made to the database they are stored in the "delta store". Periodicaly the
"delta store" is merge with the "main store" that is optimized for read access like this:
Using 3D Xpoint, which is 1 000 times faster than actual SSD, allow to have non-
volatile memory for main store, which provide the following advantages:
Scale OUT scenario is mainly used on analytical systems such as SAP BW, SAP
CAR, ...
So S/4 HANA size is limited by the size of one node/server which is around
12Tb. The next step would be probably to allow S/4 HANA to run on several nodes. Here
is, from my point of view, the requirements/solution:
Duplicate small tables on all nodes (to allow query with join
operations)
Spread (using partitions) big tables on several nodes
NUMA Issues
However, even on a single server you could have performance issue due to NUMA (Non-
uniform memory access). See SAP Note 2470289 - FAQ: SAP HANA Non-
Uniform Memory Access (NUMA) (require SAP S-User).
Roughly, your SAP HANA server/node is made by aggregating several blades which each
contains CPU and RAM. The access from a CPU to the RAM on the same blade is
fast because it use the local memory bus. While the access from a CPU to the RAM
on anotherblade is slow, because it needs to use the communication bus that connect
all the blades together, which is slower than the local memory bus.
This performance issue, is "like" the scale-out scenario (that is not supported) except that,
in the scale-out scenario the communication is made throught the network, while on the
NUMA scenario the communication is made throught the communication bus,
which is much much faster than the network, so NUMA issues ocurs only on
big/huge volumes.
Please feel free to comment, to provide additional information or to react on this article.
Feel free to contact me directly on Linked-In for any personal situation, which cannot be
exposed on the public comments.
Best regards,