Upload
elwin-hood
View
213
Download
0
Embed Size (px)
DESCRIPTION
Business Challenges Last year First Union and Wachovia initiated a merger, creating the nations Fourth largest financial services institution Corporate initiative to relocate remote data centers to a centralized location Meet and exceed OCC availability and disaster recovery objectives Increase stockholder value by reducing total cost of ownership, consolidating current environment, and leveraging new technologies
Citation preview
A Case Study
Julie J. SmithWachovia Corporation
Superdome Implementation at a Glance
Agenda Business Challenges Strategy Consolidation Superdome Executive Summary Project Scope and Timeline Migration Philosophy Migration Phases Current Status
Business Challenges Last year First Union and Wachovia initiated
a merger, creating the nations Fourth largest financial services institution
Corporate initiative to relocate remote data centers to a centralized location
Meet and exceed OCC availability and disaster recovery objectives
Increase stockholder value by reducing total cost of ownership, consolidating current environment, and leveraging new technologies
Strategy Identify projects which had the most to
gain by: consolidating to a new hardware platform and
reducing support maintenance costs providing high availability architecting a more robust disaster recovery
platform TCO Study was completed by HP and
helped to identify which projects would benefit the most from consolidation, upgrades, etc.
Why we chose Consolidate Reduce maintenance cost by eliminating
older or obsolete hardware Reduce number of servers to maintain
while actually increasing total TPM’s from 768,000 to 1,600,000
Provide robust hardware environment and increasing high availability components
Provide hardware foundation for disaster recovery initiatives
Why we chose Superdome Scalability High-Availability features inherent
to the hardware Performance data
Hardware Identified for Consolidation (6) V Class Servers (10) K Class Servers (4) N Class Servers (2) T Class Servers
Executive Summary Migrate 22 servers representing four
business units to four Superdome servers (2 production and 2 DEV/DR). Includes over 50 applications residing on 22 servers, approximately 15 Terabytes of storage in production, development, and test
Facilitate current data center move and enable applications to meet merger objectives.
Project Scope Install four Superdome 64000 servers in new data
centers Attach to Next-Generation network infrastructure
which is completely redundant Attach to all SAN storage and reconfigure I/O layout Implement new backup strategy using TSM Re-design new High-Availability Clusters and
packages Migrate existing applications to Superdome Perform full Disaster Recovery tests for each project Decommission obsolete hardware
TimelineMay October
Just Six Months….
Moratorium MergerComplete
Migration PhilosophyRely on what we know works HP LVM MC Service GuardMake improvements that must be made:
Network SAN Disk and LVM layout Backup/Recovery software MC Service Guard Basic DR implementation Maintenance strategy
Phases of Migration1. Discovery2. Design3. Implementation4. Migration5. Support and Maintenance6. Future
DiscoveryWhat can improve? performance availability disk and file
system layout backup windows network
infrastructure system standards
What works well?Humm….
Design Four Superdomes in identical
configuration 2 Superdomes in Production
Cluster at one location 2 Superdomes in Dev/DR Cluster at
alternate location 4 partitions each 4 MC Service Guard Clusters
Fibre Channel100BaseT
MC/SG Heartbeat for Partition 1 cluster
SD #1
Users LAN Backbone - primary *
SMSPC Console
AdvanceStack Hub-8TXEPower
1 3 5 10
15+
1 2 3 4 5 6 7 8
1 2 3 4 5 6 7 81 515
30
65+
1X 2X 3X 4X 5X 6X 7X8MDI
-X8MDI
100Base-TXClass II
Repeater
HPJ3235
A
Mgmt. LANConsoleLAN
Hub
5 5
8 8
SD #2
5 5
88
Users LAN Backbone - stand-by *
Storage Area Network
EMCSYMMETRIX
3830
2EMCSYMMETRIX
3830
2EMCSYMMETRIX
3830
2EMCSYMMETRIX
3830
2
MC/SG Heartbeat for Partition 2 cluster
MC/SG Heartbeat for Partition 3 cluster
MC/SG Heartbeat for Partition 4 cluster
AdvanceStack Hub-8TXEPower
1 3 5 10
15+
1 2 3 4 5 6 7 8
1 2 3 4 5 6 7 81 515
30
65+
1X 2X 3X 4X 5X 6X 7X8MDI
-X 8MDI
100Base-TXClass II
Repeater
HPJ3235
A
( Peripheral cabinet for each SuperDome is not shown)
SD #2SD 64-way with:
4 hardware partitions16 cells
44 active CPUs20 iCOD CPU128 GB RAM
Each partition has:2 x I/O chassis
2 x Core I/O card5 x FC
3 x Ultra2 LVD SCSI2 x 4-port 100 BaseT card
(Partition 4 (HR) has 4 x 4-port 100BaseT card)EMC Storage Subsystems(will be specified by customer)
SD #1SD 64-way with:
4 hardware partitions16 cells
44 active CPUs20 iCOD CPU128 GB RAM
Each partition has:2 x I/O chassis
2 x Core IO card5 x FC
3 x Ultra2 LVD SCSI2 x 4-port 100 BaseT card
(Partition 4 (HR) has 4 x 4-port 100BaseT card)
Production SuperDome Configuration(Test/DR SuperDome has the exactly same configuraion as the Production SuperDome)
Server Mapping to Superdome PartitionsPartition based
Failover (Four Partitions per eachSuperDom
e)
Commercial
V2200/10-CPU
Commercial
V2200/10-CPU
Partition 1 (4 Cells, 8 Active CPUs)
Partition 3 (4 Cells, 12 Active CPUs)
Partition 2 (4 Cells, 16 Active CPUs)
Partition 4 (4 Cells, 8 Active CPUs)
Partition 3 (4 Cells, 12 Active CPUs)
Partition 2 (4 Cells, 16 Active CPUs)
Partition 4 (4 Cells, 8 Active CPUs)
Partition 1 (4Cells, 8 Active CPUs)
SD #164-Way
16 Cells, 44 CPUs, 128 GB Memory
SD #264-Way
16 Cells, 44 CPUs, 128 GB Memory
EIM
V2600/10-CPU
K410/4-CPU
Finance
K570/6-CPU
K580/4-CPU
Finance
N4000/4-CPU
K410/4-CPU
K100/1-CPU
EIM
V2250/12-CPU
V2200/10-CPU
K410/4-CPU
HR
V2200/10-CPU
V2200/6-CPU
HR
V2200/10-CPU
Storage Design Current layout
Some direct attached arrays
Some attachment via SAN Outdated EMC Timefinder
version used for database replication
Logical volumes are not striped and possibly on single spindle
Disk array is mirrored and not striped
2 fibre controllers on each host
Target layout All SAN attached storage Install current EMC
Timefinder version and rewrite production scripts to replicate database and to backup file systems offline
Stripe logical volumes utilizing LVM Distributed options to optimize across multiple spindles
Continue with mirrored array configuration
4 Fibre controllers on each partition to attach to storage
Volume Group & Logical Volume Layout Scripts are used to build each
environment Run Inq and create device list Pvcreate Vgcreate with alternate links and
physical volume groups (PVG) Lvcreate (lvcreate –D y –s g –r N) Newfs & fsck Create vg map files & ftp to alt system
Naming standards Vgpkg10 /pkg
Disaster Recovery Strategy Currently all tape recovery from
TSM master at remote location Future design is still being
determined, but is based on Business Unit’s priority and budget.
MetroCluster and Oracle Replication are being evaluated.
Current Status To date, there are 6 production
applications that are live There are over 40 more to go
between now and mid-October DR testing to be complete by October Performance tests are still being run,
but tests show very positive results…. We’re SMOKIN’!
Questions or DiscussionThanks for attending