IBM Linux Unix to Linux Migration - IBM
1 / OS 2 3 / 4 5 IBM 2
Kernel : BSD, SYSV system calls, POSIX.1 Shell : System Utility : Application : gcc, g++, javac, vi, emacs, make, gdb/dbx, perl, etc 3
? No.1 OS in the future 4
Linux Kernel 2.6 Kernel IPSec O(1) Scheduler OProfile Feature Native Posix Thread Library (NPTL) Asynchronous I/O (AIO) Linux 2.6 kernel Provides High performance POSIX compliant multithreading IPSec layer available for use by kernel modules Improved application performance Highly scalable SMP scheduler CPU-hardware-based performance monitoring kksymoops Reverse Map Virtual Memory (rmap VM) HugeTLBFS Remap_file_pages 2.6 Network stack features (IGMPv3, Ipv6, etc.) Improved kernel bug reporting Performance improvement in memory constrained systems Performance improvement for large virtual memory applications (e.g. Databases) Kernel memory optimization for shared memory applications Improved network performance & messaging 5
Linux Kernel 2.6 IPvs Feature Linux 2.6 kernel Provides Network load balancing Access Control Lists (ACLs) Improved file system security management 4GB-4GB memory split No Greatly increased x86 physical memory support and larger application address space Scheduler support for hyperthreaded CPUs No Improved hyperthreaded CPU performance. (2.6 implementation not yet comparable) Block I/O (BIO) block layer Major rewrite of the I/O subsystem (stabilization and driver support in progress) Support for > 2TB file system New I/O elevators Interactive scheduler response tuning Support for very large volumes. Red Hat Enterprise Linux 3 supports up to 1 TB. Fine tuning for I/O subsystem performance (stabilization in progress) Scheduler improvements for interactive tasks (stabilization in progress) 6
1 / OS 2 3 / 4 5 IBM 7
Approach to Migration Pilot Project Phase Rollout Phase Develop horizontal Linux strategy in advance of development Identify selected environments for vertical Linux proof-of-concept projects Determine status of required ISV software dependencies Perform initial customized install of Linux software and hardware Begin determination of IT components which can be migrated quickly and easily Port and test initial application Test Linux as part of existing infrastructure Begin documentation on overall Linux deployment methodology Develop overall migration cost model Finalize strategy for mass deployment of Linux software and hardware Begin rollout of infrastructure replacement components Begin widespread application porting Perform final deployment QA Train system administrators and developers Begin all across rollout 8
1 / OS 2 3 / 4 5 IBM 9
??,,? 10
GNU C/C++ Performance 11
1. 2. 3. ThirdParty Tool,,, 4. 64 64 5. Edian format SUN Big Endian Endian 12
Step 1 : GNU C/C++ 1. GNU make GNU make Sun make GCC Sun C, Sun C++ 2. make GNU make 3. makefile gmake, C 4. 5. 13
Step 2 : 1. SUN 2. GNU 3. 4. 14
Step 3 : 1. 2. makefile 3., 4. 15
Step 3 : GCC 77, 90/95 (call) C, ACL, API 16
Step 3 : C++ Math X Xlib Xt X11R6X Motif Motif : CDE GNOME/KDE /LWP(LightWeightProcess) : /proc 17
Step 4 : Performance Performance Performance Performance Inspector OProfile Performance Inspector Performance analysis tool set Performance Performance OProfile RHEL4 SLES9, Translation Look-aside Buffer,, 18
1 / OS 2 3 / 4 5 IBM 19
FLOW From inventory of runtime From inventory of source P3. Test Reports and Captured Output From porting P3. Test Reports and Captured Output On no nonjustifiable defects To acceptance test T1. Testing Team Knowledge Transfer T2. Configure Baseline Testing Environment & Tools T4. Execute Baseline Test T5. Establish Defect Management Procedures T6. Execute Baseline Test in Target Environment P5. Test Reports and Captured Output T7. Track Defects and Progress T3. Assemble Baseline Test Scripts & Data P2. Acceptance Criteria and Test Suite P3. Test Reports and Captured Output To porting(breake/fix) On any non justifiable differences with baseline test P6. Defect Management DB. Testing Setup and Baseline End of port testing continues until no more defects 20
T1. Knowledge Transfer T2. Configure Baseline Testing Environment and Tools T3. Assemble Baseline Test Scripts and Data Application knowledge is required Prepare portfolio scope review Provide the environment for baseline testing All baseline testing will be done on customer premises, on customer provided hardware, software and network facilities Verify that the environment is ready to support baseline test efforts Definition of acceptance test criteria by the customer s application staff/users Testing methodology requires the comparison of results from the baseline test, using pre-ported code Test cases and test scripts for both on-line and batch functions must be developed/assembled and documented 21
T4. Execute Baseline Tests T5. Establish Defect Management Procedures T6. Execute Baseline Test in Target Environment T7. Track, Manage Defects and Progress Perform to ensure that the pre-ported code is executable and the load libraries produce deliverables Test results provide the basis for future comparisons to the test results The output from this task should be the recorded and verified results of running the test On larger projects, sets up the automated test management and defect tracking and resolution tools for use during post-port testing On smaller projects, standard procedures and a simple spreadsheet can be used In the same order as the baseline environment test, loading the same data and executing the same scripts The output from this activity should be the test results Compare the results with the baseline results Execute test scripts until all non-justifiable differences are eliminated 22
1 / OS 2 3 / 4 5 IBM 23
IBM Migration Factory Extensive Re-platforming Capabilities IBM Migration Factory 24
IBM Migration Factory Services for Solaris to Linux on IBM IBM Migration Factory Linux Assessments for Qualifying Customers IBM e x,, i, p, z IBM IBM IBM Migration Factory for Linux Solaris C/ C++ to IBM Linux Porting Oracle Database Migration Services Custom code applications, Databases, SAP/PeopleSoft/Oracle E-Business Suite 25
Thank you... 26