> h2. Overview Intro (#Intro) Interface F&R/FRPex (#Interface20F26amp3bR2fFRPex) Processing-Strategies (#Processing2dStrategies) Server (#Server) Server-Group (#Server2dGroup) Hybrid (#Hybrid) Diving deeper into technical aspects (#Diving20deeper20into20technical20aspects) Prerequisites (#Prerequisites) —- h3. Intro This blog is intended for these interested in the technical aspects of the hybrid-solution in F&R. You may want to read Carolina Maurer Mortensen’s blog, if you are looking for info on how to set the hybrid up. Here you are right, if you wish to understand a bit more of the “how it works” as opposed to “how to get it running”. You may also wish to study the SAP Notes 940002 and 909935. Part of the work of F&R is done in an external (non-ABAP) Application, also called “FRP executable” or short FRPex F&R communicates with FRPex via files written/read directly to/from the filesystem. These files are called “location data” This “location data” consists of several files in a common subdirectory-tree for each location. Here a rough overview over the steps: * * FRP writes its intermediate results to the “location data” directory * * FRP starts FRPex * * FRPex reads the “location data”, does some calculation and writes its results back to “location data” * * FRP reads the “location data” and does its post-processing * * h3. Interface F&R/FRPex The Interface between F&R and FRPex consists of the “location data” (one directory per location). This directory holds the input, output, work and configuration files for FRPex. The size of this directory varies in 0.1. 0.2. F&R-Release 0.3. 0.4. Number of location products in the location 0.5. 0.6. Depth of the consumption horizon 0.7. 0.8. Depth of the forecast horizon 0.9. Measurements show, that the compression-factor is between 30 and 50 (depending on location) In case of “Server-group” this directory is located on the central share (NFS rsp. SMB, APP-n in the picture below). Processing so many data via network imposes a severe performance-issue. h3. Processing-Strategies FRP provides the transaction /FRE/FRP_ADMIN which allows to choose between two ways of storing “location data” and processing (Server versus Server-Group). Each has its advantages and disadvantages. h3. Server Here the “location data” is stored on a local disk on a dedicated APP-server. This server is the one to process the location and no network traffic is involved. Advantage of this approach: 0.1. 0.2. No network traffic. 0.3. 0.4. More cheap app-servers possible. 0.5. Disadvantage 0.1. 0.2. High backup and maintenance effort. 0.3. h3. Server-Group The system chooses a (free) APP-server in a given server-group for processing. The “location data” is stored on a central share. The APP-Servers have access to this share via a network-protocol (such as nfs or smb), Advantage of this approach: 0.1. 0.2. insures, that in case of a server crash the location can still be processed (on another server) 0.3. 0.4. is easier to maintain 0.5. 0.6. imposes only one place to take care of for backup 0.7. 0.8. allows for automatic load balancing 0.9. Disadvantage 0.1. 0.2. High Network-Load when accessing the “location data” remotely. 0.3.