Process for optimizing software components for an enterprise resource planning (ERP) application SAP on multiprocessor servers

• US 7,805,706 B1
• Filed: 06/21/2005
• Issued: 09/28/2010
• Est. Priority Date: 06/21/2005
• Status: Active Grant

• Current Assignee: Unisys Corporation
• Original Assignee: Unisys Corporation
• Inventors: Praditphollert, Kulwipa, Ly, Tania Trinh
• Primary Examiner(s): Zhen; Wei Y
• Assistant Examiner(s): Coyer; Ryan D
• Application Number:
  US11/157,395
• Time in Patent Office: 1,925 Days
• Field of Search: 717120-135, 717168-173, 709220-222
• US Class Current: 717/121
• CPC Class Codes: G06F 11/3428 Benchmarking
• CPC Class Codes: G06F 11/3471 Address tracing
• CPC Class Codes: G06F 2201/80 Database-specific techniques
• CPC Class Codes: G06F 2201/815 Virtual
• CPC Class Codes: G06F 2201/87 Monitoring of transactions
• CPC Class Codes: G06F 2201/885 Monitoring specific for caches
• CPC Class Codes: G06F 9/5083 Techniques for rebalancing the load in a distributed system
• CPC Class Codes: G06Q 10/06 Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling

Abstract

In a three-tier ERP implementation, multiple servers are interconnected through one or more network infrastructure. Users may observe poor performance due to the complexity and the number of interconnected components in the implementation. Herein is devised a process for tuning the software component by applying tuning techniques to the OS, SAP application and Database Management System software. For each component, the process identifies potential tuning opportunities of various subcomponents. The process is iterated numerous times through all software components while applying the tuning techniques to derive the most optimal performance for the ERP implementation.

Claims

1. In a three-tier client/server Enterprise Resource Planning (ERP) Sales and Distribution (SD) implementation wherein multiple servers are connected with a network infrastructure, and wherein a database server operates with a storage system to provide database management services (DBMS), a method for optimizing software components for performance improvement in ERP comprising:

(a) evaluating each software component in said network infrastructure;

(b) identifying tuning opportunities for each software component, the identifying comprising;

(b1) identifying tuning opportunities in an Operating System (OS), the identifying comprising;

(b1a) selecting an appropriate OS;

(b1b) selecting a proper device driver; and

,

(b1c) implementing Direct Input/Output (I/O) Bridge (DIB) to subpod mapping;

(b2) identifying tuning opportunities in said ERP application;

(b3) identifying tuning opportunities in said DBMS software; and

,

(c) applying tuning techniques to each software component to achieve best system performance and avoid poor performance and bottlenecks, the applying comprising;

(c1) applying the said tuning technique to said OS, the applying comprising;

(c1a) mapping DIBs to specific subpods to keep a process and its I/O interrupts together within the same subpod to reduce latency; and

(c1b) mapping every DIB to its subpod to keep a process and its corresponding I/O interrupts together onto the same subpod if every DIB is filled with at least one Host Bus Adapter (HBA) and if all processors in the database server are evenly utilized;

(c2) applying the said tuning technique to said ERP application; and

(c3) applying the said tuning technique to said database server, which runs DBMS software.

2. The method of claim 1 where step (a) includes the steps of:

(a1) evaluating the Operating System (OS);

(a2) evaluating the ERP application;

(a3) evaluating the Database Management System (DBMS) software.

3. The method of claim 1 wherein step (c1) includes the steps of:

(c1c) selecting a version of the OS that supports 32-bit or 64-bit technology;

(c1d) selecting an appropriate device driver for allowing the Operating System to manage hardware devices;

(c1e) managing memory in the 32-bit technology to address memory limitation and to allow said Operating System to extend memory space for user applications and/or to take advantage of memory larger than four gigabytes if physical memory is available.

4. The method of claim 3 wherein step (c1c) includes the step of:

(c1c-1) selecting an Operating System version that the hardware can support, given the type of processors, number of processors, and the amount of physical memory in each type of server utilized.

5. The method of claim 3 wherein step (c1e) includes the steps of:

(c1e-1) extending addressable memory beyond four gigabytes by utilizing Physical Address Extension (PAE) if said database server has more than four gigabytes of physical memory;

(c1e-2) utilizing a switch to allow the Operating System to increase the user mode addressable space from 2 gigabytes (GB) to 3 GB;

(c1e-3) eliminating the use a switch if a database server has more than sixteen gigabytes of physical memory.

6. The method of claim 1 wherein step (c2) includes the steps of:

(c2a) configuring ERP memory management;

(c2b) configuring ERP work process type and amount;

(c2c) setting ERP processor affinity;

(c2d) choosing an appropriate table buffering scheme;

(c2e) choosing an appropriate type of number range buffering techniques.

7. The method of claim 6 wherein step (c2a) further includes the steps of:

(c2a-1) allocating sufficient memory to each ERP process without starving other processes running on the same server;

(c2a-2) reducing the number of ERP instances or ERP work processes if memory is insufficient;

(c2a-3) increasing virtual memory by adding paging files if there is not sufficient memory to allow ERP instances to run;

(c2a-4) using Physical Address Extension (PAE) technology on 32-bit technology to address memory beyond 4 gigabytes (GB) if more than 4 GB of physical memory is available;

(c2a-5) allocating sufficient memory to ERP and the other applications when running on the same server.

8. The method of claim 6 wherein step (c2b) further includes the steps:

(c2b-1) configuring ERP work processes to serve the user workload;

(c2b-2) configuring ERP work processes to perform tasks based on an amount of workload and available system resources.

9. The method of claim 6 wherein step (c2c) further includes the steps of:

(c2c-1) grouping ERP work processes of the same instance together, when more than one instance is running, so that these processes can access and share the same memory area;

(c2c-2) assigning ERP work processes of an ERP instance to a processor in the same sub-module (subpod), when running on a 32-bit system, to allow work processes from the same instance to access the same shared cache on each subpod to minimize memory latency;

(c2c-3) assigning ERP work processes of an ERP instance to processors in the same processor/memory module, when running on a 64-bit system, to minimize memory latency;

(c2c-4) balancing the ERP work processes so as to spread the workload evenly across all processors.

10. The method of claim 6 wherein step (c2d) further includes the steps of:

(c2d-1) enabling table buffering for frequently read tables with a small amount of updates in order to improve the database access time by storing and accessing data from local memory of an application server instead of going to the database.

11. The method of claim 6 wherein step (c2e) further includes the steps of:

(c2e-1) configuring the number range buffering techniques in an ERP system to improve performance by avoiding lock wait at database level during parallel processing;

(c2e-2) choosing “

no buffering”

technique for number ranges when it is essential to have no gaps in number allocation of documents and where very low system throughput is needed;

(c2e-3) choosing “

main memory”

buffering technique for number ranges when it is desirable to have high performance but allowing it to have some gaps in number allocation and the number can be assigned in a non-consecutive sequence;

(c2e-4) choosing “

local”

, or on instance level buffering only technique, to temporarily store number range on a database table NRIV_LOKAL and have almost gap-free in number allocation, but allowing the drawback that locks still occur within an instance;

(c2e-5) choosing “

local”

, or on instance level and work process level buffering technique, to temporarily store number range on a database table NRIV_LOKAL and have it almost gap-free in number allocation, as well as eliminating a lock/wait problem.

12. The method of claim 1 wherein step (c3) includes the steps of:

(c3a) configuring structured query language (SQL) Server memory;

(c3b) setting SQL Server threads affinity to run on specific processors;

(c3c) enabling lightweight threads or fibers;

(c3d) configuring Windows threads;

(c3e) configuring scheduling priority;

(c3f) configuring parallel queries where appropriate;

(c3g) setting network connection affinity;

(c3h) disabling SQL Server statistic creation and update;

(c3i) disabling SQL Server statistic collection and selecting the appropriate database lock granularity.

13. The method of claim 12 wherein step (c3a) further includes the steps of:

(c3a-1) configuring SQL Server memory specifically for the 32-bit technology to take advantage of the addressable memory space beyond 4 GB;

(c3a-2) using Address Windowing Extension (AWE) to manipulate memory beyond 4 GB;

(c3a-3) setting PAE at the OS level prior to using AWE;

(c3a-4) making a determination whether to use static or dynamic memory based on the implementation;

(c3a-5) setting the set working set size to reserve physical memory space for said SQL Server.

14. The method of claim 12 wherein step (c3b) further includes the step of:

(c3b-1) setting an affinity mask to specify the processors on which SQL Server threads can run in a multiprocessor environment.

15. The method of claim 12 wherein step (c3c) further includes the step of:

(c3c-1) setting lightweight pooling to reduce context switches by enabling said SQL Server, rather than the OS scheduler, to handle scheduling.

16. The method of claim 12 wherein step (c3d) further includes the step of:

(c3d-1) configuring the maximum number of Windows threads that said SQL Server can use.

17. The method of claim 12 wherein step (c3e) further includes the step of:

(c3e-1) setting priority boost option to specify that said SQL Server runs at a higher OS scheduling priority than it otherwise would.

18. The method of claim 12 wherein step (c3f) further includes the step of:

(c3f-1) setting parallel query to allow queries to be executed in parallel for performance improvement.

19. The method of claim 12 wherein step (c3g) further includes the steps of:

(c3g-1) setting network connection affinity to improve performance through workload balancing and data locality;

(c3g-2) configuring the Virtual Interface Architecture (VIA) network protocol;

(c3g-3) configuring the network threads to run on a specific set of processors.

20. The method of claim 12 wherein step (c3h) further includes the steps of:

(c3h-1) reducing intrusive database activities that interfere with overall system performance;

(c3h-2) disabling SQL Server automatic statistics creation and update.

21. The method of claim 12 wherein step (c3i) further includes the steps of:

(c3i-1) reducing intrusive database activities that interfere with overall system performance;

(c3i-2) disabling the SQL Server statistical maintenance function;

(c3i-3) using less intrusive row-level or table-level lock as opposed to page-level lock.