Server Appliance Proposal - PowerPoint PPT Presentation

1 / 27
About This Presentation
Title:

Server Appliance Proposal

Description:

Jay Caviness. Sr. Systems Engineer. McKesson Provider Technologies. Topics ... Contact: jay.caviness_at_mckesson.com. Corporate: www.McKesson.com/clinicals ... – PowerPoint PPT presentation

Number of Views:106
Avg rating:3.0/5.0
Slides: 28
Provided by: taber
Category:

less

Transcript and Presenter's Notes

Title: Server Appliance Proposal


1
High Availability in the ERHow McKesson
Leverages Oracle Real Application Clusters and
ASM in Critical Hospital Environments
Jay Caviness Sr. Systems Engineer McKesson
Provider Technologies

2
Topics
  • McKesson / Oracle History
  • Why use RAC?
  • Why ASM?
  • RAC Management
  • Issues Questions

3
McKesson Provider Technologies
  • Largest Healthcare IS Company
  • McKesson Clinicals Applications
  • 17 Applications Suites
  • Meds Manager
  • ER
  • Billing/Revenue Management
  • Nursing
  • NICU/PICU

4
About Me
  • Oracle DBA for 19 years
  • Senior Engineering DBA for McKesson Provider
    Technologies
  • Administered Oracle from v5.1 11g
  • OPS on VAX/VMS to RAC on Oracle 10g
  • Oracle Certified Professional
  • Known as the Grumpy DBA

5
McKesson Horizon Clinicals and RAC
  • Over 800 hospital deploy McKesson applications
  • Over 400 hospitals use Horizon Clinicals suite of
    applications
  • RAC deployed starting in late 2005 (10gR1)?
  • As of today more than 75 hospitals run Oracle RAC
  • - over 250 clusters over 800 nodes
  • - 94 on Linux- remainder on HP/UX and AIX
    (largest supports 60 hospitals)?
  • - 98 of new installations are on RAC- 65 of
    existing legacy hardware customers intend to
    migrate to linux and RAC in the next three
    years.

6
McKesson Horizon Clinicals Client
  • 1-3 clusters
  • HCI Horizon Care Infrastructure
  • HCR Horizon Care Record data mart
  • HMM Horizon Medication Manager
  • 1-6 nodes each cluster
  • 3-5 databases per cluster
  • live/qa/training configuration databases
  • Total Disk space per cluster 400GB
  • depends on how often data is sent to data mart
    (HCR)?
  • Growth 2030 year over year
  • Retention requirements vary by HIPAA rules,
    typically 7-10 years

7
RAC Overview
  • Multiple instances running on separate servers
    (nodes)?
  • Single database on shared storage accessible to
    all nodes
  • Instances exchange information over an
    interconnect network

8
  • Why use RAC?

9
Why RAC?
  • Cost
  • Commodity hardware x86 - 64bit
  • Linux in most cases
  • Highly Available
  • Deployable Automated Scripts for
  • Installation
  • Add/Remove Nodes
  • Scalable
  • Load Balancing
  • Client/Server Side
  • Services

10
Deployment
  • Standard silent install process
  • Standard install difficult over limited bandwidth
  • Same script for all platforms
  • Pre-defined nodes, interconnect, vip, network
    storage from McKesson OS install process
  • Script determines RAC/non-RAC and prompts user
    for
  • RAC - full install (CRS, ASM DB),
    infrastructure (CRS ASM) or database
  • Non-RAC full install, ASM Only, Database Only
  • Installation includes software, patchsets and
    configuration

11
Typical Oracle RAC Installation
  • Red Hat Linux 4.6
  • Oracle 10.2.0.4
  • Dell 2950 or HP DL380 w/dual quad core processors
  • 16G RAM
  • Powerpath on EMC Clariion
  • 4 Nodes
  • UDP Interconnect via Private VLAN
  • No crossover cables for two node clusters
  • NIC bonding is supported
  • 40 Minute Installation Time

12
Failover
  • New and legacy applications have different
    connection types
  • JDBC
  • OCI
  • SQLNet
  • Application Failover via TAF for select
    statements
  • Applications designed to automatically reconnect

13
Load Balancing
  • Two of three types of load balancing handled by
    RAC
  • Server side load balancing via the cluster
    listeners
  • Client side load balancing via tnsnames.ora
    entries and jdbc connection strings
  • Application load balancing with hardware and
    virtual load balancers
  • Services are used to offload reporting and other
    processes to specific node(s) of a cluster

14
Application Changes
  • Few, but important changes were made to the
    applications to fully support RAC
  • Cached/No order sequences
  • JDBC connection strings
  • SQLNet connection strings
  • Reconnection process

15
Rolling Upgrades
  • With RAC different upgrade types can be done
    without downtime
  • Hardware upgrades
  • Operating system patches/upgrades
  • Oracle clusterware/ASM/database patches and some
    upgrades
  • Application upgrades

16
  • Automated Storage Management

17
Why ASM
  • ASM is required by McKesson for RAC and strongly
    suggested for non-RAC databases
  • We chose ASM for several reasons
  • Setup and Management
  • Simplicity
  • Scalability Performance
  • Robust and Stable

18
ASM - Setup
  • Installed in its own Oracle Home
  • Typically external redundancy
  • Some clients have two fail groups on physically
    separated SANs
  • 400G/per cluster to start
  • Easily expandable by adding 50G luns to ASM
  • Datafiles set to autoextend-
  • caveat 32g limit for 8k blocks unless using
    bigfile tablespaces

19
ASM - Simplicity
  • ASM allows DBAs to be less concerned about file
    management
  • db_create_file_dest
  • ex create tablespace foo datafile
  • db_create_online_log_dest_n
  • ex alter database add logfile thread 2 size
    200m
  • Oracle Managed Files

20
ASM Scalability Performance
  • ASM allows management without affecting running
    databases
  • Adding/removing ASM luns/disks
  • Add the lun (owned by oracle)
  • Mark the disk with the oracleasm service (linux
    only)
  • alter tablespace dgroup1 add disk vol009
  • Performance by striping
  • Datafiles striped in 1m chunks across all ASM
    disks
  • Creates a plaid on previously striped luns

21
ASM - Stability
  • Even at the first generation (10gR1) there where
    no show stopper bugs with ASM
  • All versions are very stable
  • Minor issues due to lun management, mainly at the
    OS level
  • Luns lose ownership on reboot HP/UX
  • Special files to denote ownership can be lost and
    must be recreated if the luns are re-imported
    AIX

22
  • RAC Management

23
RAC Management
  • Grid Control 10gR4
  • 10gR4 agents
  • Automated install process
  • One button install
  • Deploy agents
  • Standard alerts
  • Easy for DBAs and support personnel

24
Issues
  • The largest issues faced are generally due to the
    varying use of the applications.
  • Client use varies
  • Modifiable applications
  • Client queries
  • Client modifications
  • Reporting
  • Inconsistent setup
  • Hostnames
  • Network
  • Storage
  • Number/type of applications

25
Future
  • The future of Oracle/RAC is happening now!
  • OEL 5.2 retiring other operating systems for
    new clients
  • McKesson System Appliance
  • Consistency
  • Uniformity
  • Cost savings for clients in hardware, energy,
    support/services
  • Virtual Machines

26
  • Questions?

27
More information
  • Contact jay.caviness_at_mckesson.com
  • Corporate www.McKesson.com/clinicals
  • Personal www.grumpy-dba.com
Write a Comment
User Comments (0)
About PowerShow.com