Paper

Technical basis: SAP HANA
T-Systems' Dynamic Services customers ben-
efit from SAP HANA platform performance as
part of Application Operations for SAP Solutions.
The core of SAP HANA is the SAP in-memory
computing engine which stores the data directly
to the RAM of the server and not to external
hard drives or flash memory. Data is therefore
accessed more quickly since it does not have
to be reloaded or rewritten. In addition, a com-
puter's RAM has a considerably faster access
time than the hard drives in classic databases
which allows in-memory solutions to work
considerably faster.
When starting up the database, SAP HANA
loads the entire data pool from the hard drive
into the memory (not only in cache) so that
no more data has to be uploaded once the
system is up and running. This simplifies the
access algorithms and increases the applica-
tion's performance. To guarantee that data
stored to volatile RAMs is unaffected even
when malfunctions or disruptions occur, the
solution makes use of technology such as
snapshots, transaction logs and replications
which regularly store the data to conventional
hard drive systems. Moreover, SAP HANA re-
duces memory consumption by implementing
specific compression algorithms and data
structures.
The second cornerstone to SAP HANA, besides
in-memory computing, is the type of data stor-
age used. SAP uses a hybrid model in HANA
which allows data to be stored in rows and
columns. Classic relational data bases are row-
oriented where the information for customer
records is entered consecutively: Name, place,
country. This is then followed by the next cus-
tomer record. Column-oriented data bases
follow different principle. They first enter all
of the names consecutively before entering
other items of information such as place and
country.
HANA thus combines the best of both worlds.
Row-oriented data storage enables faster writing
access. However, it takes longer for the data to
be read since the system has to spring through
data entries when searching for a particular
name. As a result, this type of data base is suited
primarily to OLTP systems (online transactional
processing) such as ERP from which a high
volume of information is continually fed into
the data base.
Column-oriented data bases are more advan-
tageous for read access. When searching for a
name, the data base simply selects the relevant
column in sequence without jumping between
entries. The process is extremely fast. This is
why column-oriented data bases are mostly
integrated into OLAP systems (online analyti-
cal processing), such as data warehouses and
other systems that have been specialized to
conduct assessments and analyses.
Up to now, T-Systems has relied on two manu-
facturers – both of whom are SAP-certified –
to produce its AOfSAP and SAP HANA appli-
ances. The company's aim in cooperating with
Dell on this project, however, was to optimize
the overall costs for SAP HANA in the private
cloud environment and so to provide its cus-
tomers with an even more attractive offering.
Increased standardization, faster allocation
and a more efficient operation all needed to
feature in this offer. To do this the appliance
must be integrated deeply into T-Systems'
Dynamic Services infrastructure.
The solution: Dell appliance with
Intel Xeon processors E7 v2
As a result, T-Systems launched a market
analysis in August 2013 and conversed with
various providers. After three months the de-
cision was made to go with an SAP HANA scale-
out appliance based on the Dell PowerEdge
R920 platform with Intel Xeon processors E7
v2. Next, T-Systems simulated test scenarios
for its Dynamic Services on a reference archi-
tecture as part of a proof-of-concept (PoC)
phase. The Dell appliance is now available in
T-Systems' SAP HANA shopping basket. For
example, the company’s private cloud environ-
ment currently hosts 15 SAP BW systems via
SAP HANA appliances for a major client with a
total of around 55 TB SAPA HANA memories.
In practice, T-Systems integrates Dell appliances
in a scale-out configuration using four Intel
Xeon processors E7 v2 4880/90 and 1 TB RAM
per server. The solution is scalable for memo-
ries of at least 2 and up to 16 TB. Compellent
array is also used to store a copy of data on
the server memory. All of these components
are installed for servers that have been devel-
oped specifically for HANA which in turn are
based on fast Intel Xeon processors E7 v2. The
Intel Xeon E7 v2 is the only certified CPU plat-
form for live operation with SAP HANA at pres-
ent (January 2015).
Optimize operating costs through standardization
and increased efficiency