cancel
Showing results for 
Search instead for 
Did you mean: 

Index Server and Scale Out

Former Member
0 Kudos

Hi Gurus

I have never done an Abap on Hana install.

I am having doubts on how is the data accessed in scaleout scenarios.

e.g. If I have a 3 server landscape for the hana server (One Master, One Slave and one Standby) and I want to use this server for a  ERP install.

Question 1. How is the communication routing done from the abap server to the hana in a scaleout scenario?  When I install the abap server I have to provide ONLY one hana hostname, right? If so, is the master server hostname?

Question 2.  What happen if the master hana server crashes, how the ERP will know where is the standby server that will assume the load?

Question 3. I was told that the hana webdispatcher only provides loadbalancing for XS engines. Is this true? I also noticed that in a multihost scenario the xsengine and the webdispatcher are not running in the slave server. Why is that?

If you can clear this for me... I will appreciate it very much...

Regards

Henry

Accepted Solutions (1)

Accepted Solutions (1)

Former Member
0 Kudos

Hi Henry,

First, ERP on HANA is not certified for scale-out scenarios and will only be considered in very special cases where scale-up is not feasible. For your answers below, I will assume BW on HANA, which is NW on HANA and available for scale-out.

Question 1. How is the communication routing done from the abap server to the hana in a scaleout scenario?  When I install the abap server I have to provide ONLY one hana hostname, right? If so, is the master server hostname?


     Yes, for initial install you can use the name of the master server. In the DEFAULT hdbuserstore on the ASCS, you can specify multiple hostnames to ensure connectivity to each host if one fails.

Question 2.  What happen if the master hana server crashes, how the ERP will know where is the standby server that will assume the load?


     As of SPS09, each node will have a Name Server service that can take over in the event of a failure. All the name servers have the topology information for the cluster. The ASCS will attempt to use the hdbuserstore to reestablish connectivity and will essentially use a round robin approach to connect to whatever node is available.

Question 3. I was told that the hana webdispatcher only provides loadbalancing for XS engines. Is this true? I also noticed that in a multihost scenario the xsengine and the webdispatcher are not running in the slave server. Why is that?


     The internal web dispatcher is for load balancing the internal XS engines in an MDC deployment, but an external web dispatcher can be deployed as normal. The XSengine service will run on multiple hosts in a scale out. See section 5.1.4.3 of the admin guide. http://help.sap.com/hana/sap_hana_administration_guide_en.pdf



I tried to be brief, but hope it helps.

-Jake


Former Member
0 Kudos

Great explanation!!! Thanks!

Answers (0)