SAP Mobility Apps on High Availability
I just prepared a initial cut on the information that user should know before they plan to place the SAP Mobility System in HA mode.
please post back your comment. Planning to incorporate in then in the subsequent version.
Need for SAP Mobility in HA architecture
1. Purpose of this document
Based on our previous experience with some of the mobility implementations, we feel that if a HA needs to be introduced in Mobility environment, then it needs to done in the very beginning, as late entry would prove very costly. In this article we will touch up on some of the parameters that can help you to assess the need for HA along with other entities that are either directly or indirectly associated with HA. The target audience for the document is System / Solution architect, Project Manager & Senior management who are interested in implementing SAP Mobility platform.
2. Need for HA
The ray of mobility applications have changed course, though it all started with small and interactive games, it has taken-up a shape where it can manage a complete industry. Every industrial vertical has been disrupted by the innovation of smartphones. B2B and B2C applications are flocking the market. With such development and dependencies it becomes variable mandate to keep the mobile infrastructure up and running. In many organizations mobile applications have started supporting various business activities likes Dashboard, Equipment maintenance, Support arm for Sales and training related activities and so on. With these functions and activities that are getting processed/ activated through mobile, outage of mobile application will impact business to a greater extent.
3. Demystifying Role of HA
High Availability is one of the several quality attributes that is important in a given mobility landscape. But many a time people get confused with the terms HA, Scalability & Load Balancing though they are often used in mobility environment.
HA: High Availability of a system at a given site is typically achieved through architecture configurations termed “active-active” or “active-passive”. In this paper we will refer “active-passive” for database. HA can be achieved through Cluster (“Group of servers”). Currently, there are few third party vendors who can provide software for clustering service, but this paper talks about HA for mobility applications in general context keeping technical implementation apart.
LB: Load Balancer, where in it is used to balance the load to ensure the availability of Service. LB along with HA is considered to be just a complement. LB doesn’t require any Specific HW spec. unless end user is more specific. Load balancer is especially required in the case where the distribution of Mobile traffic is not clear
Scalability: Represent a ways to handle increase in data traffic. The system should be designed in a scalable manner such that it allows the infrastructure team to scale up the system both vertically and horizontally. Vertical scaling is the way where we increase the RAM size and processor so that more floating operations can be performed. Horizontal scaling is where new system will be introduced. This is typically used when the existing servers are full and there is no scope of expansion.
4. Parameters for HA identification
Based on our experience the below table (Table1) has been prepared (this is just for reference).
We find that in many circumstances a non-conformance to HA will result in any of the following.
- Service outage.
- Reduced Operational efficiency.
- No real time access to Enterprise data.
- Reduced visibility on SMP events like DCN.
Table1: key parameters to identify the need for HA
Based on the financial comfort HA can be extended for QA landscape. This will really help if any production related HA issues needs to be reproduced.
Note: Mobility Application will still work well even in absence of High availability but with some exception.
5. Typical Mobility deployment architecture
In the below figure (figure2) we have detailed out how a system can be placed if user wants to achieve HA. Since we have defined 2 boxes it can be even referred for scalability.
Reverse Proxy and Relay Server
Reverse proxy would be used in case of backend system being SMP 3.0 without any MBO runtime. Relay server would be used for AFARIA & SMP MBO Runtime system. These boxes are under LB (Microsoft Load Balancer) so that load request can be equally shared.
SMP 3.0 Boxes
From SMP 3.0 onwards the concept of Farm is no required, if MBO runtime is not installed. Instead SAP recommends to have LB in front of the SMP server so that both scalability and HA is guaranteed. In case of any contingencies at least one of the server will be up and running.
NetWeaver Gateway can be front ended by LB (web dispatcher) which helps the system to both scale as well to handle the load in a Round Robin fashion. In Some cases if organization wants two level of defense, then Gateway can be placed the same in DMZ.
Any database that we have defined in the below diagram will always runs behind the Microsoft Active Cluster or Red hat Enterprise Server Cluster Service. Pls. note not all Operating System supports Cluster in their Standard release.
This is a Physical box where data are stored and retrieved. There are Logical Units (LUN) where data will reside. The best practice is to have separate LUN for each Database & it’s log.
Figure2: Typical Mobility platform deployment.
6. Caveats on HA system
The following points are to be noted if there is a plan to have HA system on mobility architecture:
Impact on installation timelines: From SMP 3.0 onwards mobile server runtime has several choices for database (Please refer PAM list of detail information). If either of the above listed database is there on HA mode then it is quite simpler else there should be separate provision of effort for setting up the DB on HA mode, for which additional effort and resource may be required.
Impact on development timelines: Developer should handle all sort of failure cases in their code invariable of HA solution in their landscape.
Impact on testing timeline: Separate test plan and test cases should be made available to check HA conformance.
Impact on system maintenance: HA system will require special attention in terms of maintenance like SAN/NAS availability. Regular log verification has to be mandated, since system failure may result in cascading failures.
7. About Author
Arun Santhanam [Project Manager SAP Mobility at Tech Mahindra Chennai ]
Arun has over 10 yrs. of experience in IT and has extensively worked on SAP mobility. He has worked with various customers starting from automotive industry, discrete manufacturing &Healthcare. One of the projects that he spearheads has been awarded by SAP under “ACE award for automotive industry” for the year 2013. He has done mobility solution & implementation in various geographies across US and Asia-Pacific region. In my subsequent paper would like to describe in depth technically on the way to achieve HA and planning to cover even Fiori under the umbrella of SAP mobility Apps.