ODRC Mainframe Systems Migration Project OFFICE OF ADMINISTRATION Project Management Office.
Eranea's solution and technology for mainframe migration / transformation : detailled presentation
Transcript of Eranea's solution and technology for mainframe migration / transformation : detailled presentation
www.eranea.com
Migration mainframe x86 servers
automated transformation toward Web, Java, Linux
[email : [email protected]]
2 / 82www.eranea.com
agenda
● starting point● goals of migration● Eranea's proposition● methodology & technology● project steps● benefits / conclusions● complements :
– screenshots NeaControlCenter (NCC)– situation / market 2014
3 / 82www.eranea.com
starting point● a (very) large mission-critical application for the customer
● embedding his entire business expertise and driving key operational processes
● bullet-proof after 10 / 20+ years of continuous activity
● representing massive investments (10s, 100s, 1'000s men-years of development)
● doomed to technological obsolescence● highly expensive compared to 2014 standard systems (x86)
4 / 82www.eranea.com
goals ● tactical :
– massive cost savings → 80-90 % (capex + opex of M/F initial TCO)
● strategic :– technological transformation : web, java, linux, etc.– back to standards → alternative suppliers rather than monopoly, leverage on innovative « boiling» of x86 world
– functional improvements : ● rich and flexible UI + new uses (mobile, etc.)● web services for SOA, BPM : batch & transactional● (private) cloud computing
– adequation with 2014 competences : z/OS + Cobol getting (very) seldom → java, web and linux highly abundant
5 / 82www.eranea.com
business case / ROI example
10%
20%
30%
40%50%60%
70%
80%
90%100%
cpu
peripherals(disks, tape drives, printers, etc.)
3rd party software
system software(z/OS, Cics, DB2, etc..)
factor 10
4.25 MEuros / year
identical savings on numerous projects :minimum 80% and up to 90% !
project : media group (CH)
400 kEuros / yearvia
x86, Linux, Java(→ > 3.8 MEuros / year savings !)
6 / 82www.eranea.com
Eranea's proposition
● reduce IT costs while guaranteeing :– investment protection : application software and human business expertise
– deep technological transformation● via a unique migration path :
– allowing continued maintenance (no stop) during project execution
– fluid, incremental and secure progress : no big-bang (iso-functionnality, very small granularity)
– industrial and repeatable process (100% automated application transformation)
● with advanced and complete tooling for project execution and subsequent operations on new x86 system :– NeaControlCenter + associated components
8 / 82www.eranea.com
strict iso-functionality
● definition = – produce exact same results "to the last bit and pixel"
● some examples : – computation with 31 significant digits in Java (> precision of BigDecimal java library)
– Cobol semantics fully preserved for variables, groups with LEVEL, REDEFINES, management of POINTERs, implicit initialization according to PIC, etc.
– full EBCDIC support on Linux for sorting (« collating sequence »), for Cobol programming [ ex : IF A > B ; A and B → PIC X(N)]
● some advantages : – transparency for end-users, « invisible » migration for client systems, continued coherency in numbers on long-term historical reports, automated and quantified tests, etc.
9 / 82www.eranea.com
system transformation
bms3270 customUI
db2 vsam sequentialfiles
cics / ims other TPmonitor
cobol jcl utilities(sort,...)
batch scheduler
html/js (GWT)
relationaldata
(db2, oracle)
appl & transact services
java AS (JBoss, tomcat, etc.)
java xcl utilities(sort,...)
batch scheduler
standards
speed
security
sequentialfiles
browser
new application platform = Java 100%→ OS-agnostic
xms
10 / 82www.eranea.com
incremental migration (1)
DB2
TP / batch
z/OS
DB2 / ORCL
TP / batch
Linux
DB2
TP / batch
z/OS
TP / batch
Linux
time
migration :users & jobs
N monthsDB migration
mainframe stop
x86 setup install transco + NCC
initial tests
maintenance in Cobol
maintenance in Java
unique DBat all time
start end
11 / 82www.eranea.com
incremental migration (2)
part 1
part 2
part 3
part N
part1 1
part 2
part 3
part N
tests rollout
« cascading » migration
... ...
start end
12 / 82www.eranea.com
incremental migration (3)
part 1
part 2
part 3
part N
part 1
part 2
part 3
part N
tests + rollout
overlapping and combined migration
...
start end
recommandation
● reduced duration● early hands-on● rapid feedback● fast savings
13 / 82www.eranea.com
multi-level transformation
ImportSources
Transcode ImportJava
CreateProjects
CompileJava
Packagewar
Deployn targets
automated controlled
tests
automated
PROD
INTEG
DEV
full transformation process ateach level = guaranteed QA
15 / 82www.eranea.com
technological core
“Cobol” support
SQL support
CICS Emulation
Display support
Tracing / logging
InternalObject
implementation
JavaProgram
(incl SQL)
XMLScreen
DBMS
LexicalAnalysis
Syntax Analysis
Semantics Analysis
NeaTranscoder
NeaRuntime
SOA/ BPM
Online
Batch
2 distinctive features : 100% automation + strict iso-fonctionnality
remainsafter
project
XCL
Code Generation
Cobolcopy
Cobolpgm
BMSdesc
JCL
disappearsafter
project
wsfront-end
(soap, rest)
16 / 82www.eranea.com
NeaRuntime (incl. NeaWebTerm)
● runtime framework– 100 % Java
● its services :– advanced class loading (bytecode optimization, old version boycott, etc.)
– substitute / emulation of APIs (CICS / IMS API, batch, SQL API, BMS → xms, etc.)
– Cobol iso-functionnality– optimized management of JVM resources (sql connections, tcp/ip client sessions, pools for memory structures, gc minimization, etc.)
– etc.
17 / 82www.eranea.com
incremental migration
Yweeks
DRDA
time
activity
Cobolon mainframe
Javaon x86
100%
Xmonths
(X defined by customer)
• 100% of data on z/OS• Cobol remains reference
Java=
reference
dat
a m
igra
tio
nto
lin
ux
db
ms
stopmainframe
CICS DB2
progressivemigration
big bang avoidance → critical success factor
instantaneousway-back
0%
tomcat
costreduction (if MLC)!
JDBC
18 / 82www.eranea.com
architecture (1)
production BATCH
sourcesrepository
DB ERIT
NeaControlCenter(+ standard console)
migration factory
CIengine
application code
DB2
production TP
analyze,transcode,compile,package
run
tools installed on-site→ “continuous migration”
multi-layer processing cluster :(unification, distribution, processing)
JDBC /DRDA
feedback(jmx, snmp)
private cloud (CloudStack,OpenStack,etc.)
« engineered x86 system » : Cisco UCS, IBM Flex, Oracle Exa, EMC VCE, HP CloudSystem, etc.
deploy(war, zip)
synchronousdata
mirroring(real time)
backupdatacenter
RACF /TopSecret
LDAP
19 / 82www.eranea.com
architecture (2)
DB2
scheduleragent
DFSserver
mailserver
HFS
docsserver
printserver
SAN
accèsDB
shared filesaccess
common servicesaccess
storageaccess
logical conditionsmanagement
+ job launching
remotesynchronous
mirroring
backupSAN
z/OS scheduler (ex : TWS, Control/MAutosys, CA7,etc.)
Architecture / structure : ● scheduling logic and organization
fully preserved ● dynamic (optimized) worker
sheduling via agents
● centralized supervision / reporting on NeaControlCenter
● file sharing with z/OS (via HFS) as needed
● ad hoc (= out of plan) job launching via NeaControlCenter
scheduleragent
scheduleragent
scheduleragent
20 / 82www.eranea.com
architecture (3)● high availability :
– multiple stages , redundant instances● stringent security :
– inter-entity isolation● top performances :
– Intel Xeon : last generation extremely powerful!– flash arrays for million IOPs
● total manageability : thousands of JMX or SNMP probes / samplers for all kinds of values / actions
21 / 82www.eranea.com
tests (1) - transactional
Tomcat
CICS DB2
XMLscreen / kb
data
3270
XMLscreen /kb
data
HTML
XMLscreen / kb
data
bug fix :runtime / application
(1)
(2)
(3)
(4) if (1) & (3) different
DRDA to DB2(via JDBC)
capturereplay
(batch or TP)
code coveragemeasurementsto quantify tests
exhaustivity
100s / 1'000scaptures
JAVA
example: 35'000 screensto cover 10M lines
at 90% level
application programssometimes still have
bugs ...
XMLmodifieddb lines
triggers
25 / 82www.eranea.com
NeaControlCenter (1)
● inventory tool for application assets – analysis of cross-references– full-text search engine
● console to define and monitor tests :– organization of capture / replay campaigns– automated validation of iso-functionality (screens / DB / files identical)
– supervision of code coverage progress for application software
26 / 82www.eranea.com
NeaControlCenter (2)
● monitoring / management console for transcoding factory :– status and progress for transcoding / compilations / packaging
– status and progress for tests and rollouts● monitoring / management console for opérations
– exhaustive monitoring for transactional processing (sessions, TD, TS, SQL, MQ, ...)
– launch + monitoring for status / progress of batch jobs
● 200+ screens and functions
32 / 82www.eranea.com
project : standard steps
initialpresentation(s)
technicalworkshop(s)
in-depthanalysis
Proofof Concept
proposal/ contract
solution/ installation
part 1/ part N
client OK
globalvalidation
detailedconcept
+ business case / ROI
N transactions +batch jobs
fully runningby customer
33 / 82www.eranea.com
conclusion● x86 systems are simultaneously very strong (availability, performances) and most economically efficient in 2014 to support any large-scale mission-critical application
● Eranea proposes an extremely efficient and risk-free transformation solution :– massive costs savings (allowing project auto-financing)– deep technological transformation– based on unique migration methodology
● 2 keys for success :– savings and transformation (application + infrastructure) done in a single shot (<> pure « rehosting » like type Cobol-IT, MicroFocus limited to savings only without stategic benefits -> dead-end ?)
– fluid and progressive migration → total absence of risks● no disturbance / productivity loss for end-users● rapid adaptation + adherence for developers
34 / 82www.eranea.com
Thanks !
● questions ?
Eranea | Lausanne (Switzerland)
51 / 82www.eranea.com
neaControlCentertests : captures <> replays
(examples for TP)
80 / 82www.eranea.com
market trends : why x86Source : Phreshing
performances / functions / availability : x86 has come to par with RISC ! (at least...)
since 2004, x86 > RISC & mainframe
2 new generations since then...
Source : Intel
81 / 82www.eranea.com
market trends : h/w → x86● x86 is becoming standard
● sold per year:
– 4'000 mainframes
– 220'000 Risc servers : x 50
– 9'950'000 x86 servers : x 2'000● performances I/Os & CPUs :
– x86 = Risc = M/F (→ Moore's Law...)
– when architecture / setup is appropriate
– much more cost efficient ! ● where innovation happens :
– heavy R&D :● Open Compute Project by Facebook
– in-memory Dbs, flash arrays, ...
– green IT
– etc.
Windows = lotsof
office automationservers
source : Intel
82 / 82www.eranea.com
market trends : s/w → linux
worldwide server market(roll-outs, subscriptions, shipments)
Windows = lotsof
office automationservers
● linux (on x86) is growing fast
● proprietary systems disappear
● linux references (finance): London Stock Exchange, NYSE, Euronext, etc.
● for the “3S” : Speed, Stability, Security
● cost ≈ free
– based on OSS
– compared to z/OS
NB : Eranea's solution is OS-agnostic and works on any platformsupporting Java : Solaris, AIX, MS-Windows, z/OS, zLinux, etc.