DB2 Data Sharing Performance Beginners Guide

IBM System z Technical University – Vienna , Austria – May 2-6

zZS26 DB2 Data Sharing Performance For Beginners
Martin Packer

© 2011 IBM Corporation


Abstract

 This presentation provides an introductory-level view of how to
look at the DB2 Data Sharing performance numbers
– from both a z/OS / RMF and a DB2 perspective

 Performance topics include
– XCF
– Coupling Facility
– Data Sharing Structures
– The application's perspective

 Performance topics don’t include
– Other forms of Data Sharing eg VSAM RLS
– Overly detailed descriptions

2


Agenda

 What’s The Point Of Data Sharing?

 Introduction to Parallel Sysplex

 Introduction to DB2 Data Sharing

 Performance

 Summary

3


What’s The Point of Data Sharing?

 Higher Availability
– Reduction in single points of failure
• If configured properly
• If effectively planned
 Greater Scalability
– Additional z/OS LPARs providing additional resources
• Typically on different footprints
• Potentially “on the fly”
• Nearly linear
– Additional DB2 subsystems providing more virtual storage
 Software Licencing Considerations
– eg Parallel Sysplex Licence Charge (PSLC)

4


Hot Standby Scenarios

 Many installations configure redundant LPARs
– Actively processing their share of the load
– Spread 1 or more to 1 or more footprints
– Examples:
• 1 LPAR on each machine
• 2 LPARs on each of two machines
– When one Data Sharing member fails the rest service the work
– Issues:
• How do you avoid affinities?
• How do you ensure work gets routed to the right surviving LPAR?
• In general, what ARE your recovery policies?
• How do you avoid the cost of too many LPARs?
• Should we have hot standby DB2 Subsystems instead of LPARs?

5

Technical Introduction



Major Parallel Sysplex Components

Member 1 Member 2

Application XCF XCF Application

Coupling
Facility
Structures
Sysplex
Timer
7


Major Parallel Sysplex Components

 Coupling Facilities (CFs)
– With structures
– Usually more than one CF
– Run CFCC code rather than z/OS
 Members
– Such as LPARs
– Up to 32 members
 Links between Members and CFs
– Configured for bandwidth and availability
 XCF
– Provides a communications mechanism
 Applications
– Exploiting CF structures directly
– Using XCF services
 Timer
– Ensures all members are synchronised
– Traditionally Sysplex Timer
– Strategically Server Time Protocol network
 XES manages communications

8


XCF
 General signalling mechanism
– Introduced before the other Parallel Sysplex functions
 Traffic divided into transport classes
– These use either Coupling Facility structures or CTCs to pass messages
• Dynamically routed based on XCF observing performance
• Dedicated CTCs
– Originally were faster than CF structures
– Must define a pair of paths for each connection
– Definition can get quite complex
– Transport classes have a maximum message size
• Fitting messages to classes is a significant tuning item
– One message per buffer
> Buffer space wasted if message smaller than the buffer
> Additional processing if it’s bigger
 Applications use specific XCF group names
– Example: IXCLOxxx is XES lock resolution
– Application address spaces connect as members of the group

9


Coupling Facilities

 Usually on same footprint as z/OS, Linux or z/VM images
– Called an Integrated Coupling Facility (ICF)
– ICFs generally on (cheaper) PUs characterised as ICF PUs
– Can be stand-alone
– If on same footprint as z/OS images using the CF a footprint failure
would bring down both z/OS images and the ICF
 Speed of CF PUs relative to sharing z/OS image PUs important for
performance
– ICF PUs automatically matched to z/OS PUs on the same footprint
 CF PUs can be shared
– Generally reduces coupling performance
• Especially if Dynamic Dispatch turned on
– Only recommended for Development and Test Parallel Sysplexes
 CFCC code releases called Coupling Facility Levels (CFLEVELS)

10


Coupling Facility Structures
 Cache Structures
– Data is cached
• Requests await the return of data
– Example: Enhanced Catalog Sharing
• SYSIGGCAS_ECS
 Lock Structures
– Control granting of locks
• Requests don’t await the return of the lock
– Example:GRS Star
• ISGLOCK
 List Structures
– Groups of “lists”
• Which are more like arrays
– Example: XCF
• IXCPATH_CFP2
 Serialized List Structures
– Example Websphere MQ queues
• MQPGCSQ_ADMIN
 Different performance and availability characteristics for each exploiter

11


CFSIZER

 Website that enables you to size CF structures
– http://www.ibm.com/systems/z/cfsizer/
– Most IBM Product structures catered for
• Given a product name it tells you which structures you want
– And suggests a size
– Produces an “initial” configuration
• For example DB2 structures are likely to need tuning
– In fact CFSIZER probably suggests to you too few GBPs
– Has good Help
– Recently updated

12


Coupling Facility Links
 Different types:
– Internal Coupling (IC)
• Extremely fast, within one footprint
– Integrated Cluster Bus ( ICB-4 and ICB-3)
• Fast, very short-distance links between footprints
– Inter-Systems Coupling (ISC-3)
• Slower, longer distance
– Speed decreases with distance
• Needed for eg cross-town sysplexes
– Infiniband Statement of Direction for System z10 EC

 Redundancy and bandwidth are both important
 Each generation of each of these provides greater capability
– Typically speed
– Generations tend to go with processor families
 Typical configuration:
– 2 footprints
• Each has 1 or 2 member z/OS images
• Each has 1 Internal Coupling Facility (ICF) image
• IC links between members and ICF image on the same footprint
– ISC links between a member and its remote ICF
> ICB links would imply footprints physically very close

13


Synchronous and Asynchronous CF Requests
 Synchronous (Sync)
– z/OS engine waits for completion
• Each microsecond of request service time is a microsecond of lost engine capacity
– e.g GRS Star
 Asynchronous (Async)
– z/OS engine does not wait for completion
– Response times usually longer than for Sync requests
– e.g XCF signalling
 Automatic Sync to Async Conversion
– Algorithm introduced by z/OS Release 2
– Requests converted wholesale
– With conversion an occasional request is tried as Sync
• Governs whether conversion is the right thing to do
• Factors
– Larger data transfer
– Longer / slower links
– Processor speed
– Duplexing
– Thresholds recently refined

14


Structure Duplexing
 2 copies of the same structure in different CFs
– Maintained in sync
– Higher resilience to component failures
• Loss of z/OS images and ICF on same footprint less likely to cause an outage
• Faster than structure rebuild
 Bidirectional links required between the two CFs
– Preferably more than one
 User-Managed
– “User” is DB2
– DB2 writes data to both structures
• Async write to primary
• Then Sync write to secondary
• Completion when both writes succeed
• Reads only from the Primary
• In event of failure reads from Secondary
 System-Managed
– XES writes to primary and secondary
• Both CFs communicate through a separate link to ensure status is shared
• Request only completes when both structures have been accessed

15


CFLEVELS
 Coupling Facility Code Releases
– Loosely related to processor family
– New functions and algorithm enhancements introduced this way
• Eg CFLEVEL=11 is required for structure duplexing
• Exploitation may require corequisite functions in eg z/OS or DB2
– Eg z/OS Release 2 + PTFs for System-Managed Duplexing
 Footprints can have different CFLEVELS for each LPAR
– An LPAR can only run one level
– This facility to ease migration
 Structures occasionally need to increase in size when upgrading
 Useful information in
http://www-03.ibm.com/servers/eserver/zseries/pso/cftable.html
– Brief summary of CFLEVEL features
– Matrix of processor support for each CFLEVEL

 Latest CFLEVEL is 15
– More concurrent CFCC tasks
– Base for reduced synchronisation traffic for Structure Duplexing
– Structure-level CPU recording

16


Intelligent Resource Director
 Manages resources within a “LPAR Cluster”
– The members of a parallel sysplex on ONE machine
• Physically impossible to move resources BETWEEN machines
 Varies on and offline Logical CPs
 Manages LPAR weights between members
– The total for the cluster remains constant
 Manages CHPIDs between members
 Memory NOT managed
 Decisions based on WLM Goal attainment and PR/SM’s view of resource utilisation
 Helps “takeover” scenario
– Eg LPAR weights move to the surviving member(s) on the machine

17


Workload Distribution Mechanism Examples

 WLM
– Batch Initiators – JES2 and JES3
– VTAM Generic Resources
– Sysplex Distributor for TCP/IP
 DB2 Data Sharing Group Attach

 CPSM
– Dynamic CICS workload management
– Plus many other functions for managing CICS regions

18


Major DB2 Data Sharing Components
Shared Data

z/OS Image 1 z/OS Image 2

DB2 1 IRLM 1 XCF XCF IRLM 2 DB2 2

XCF Structures

LOCK1 Coupling
GBPs
SCA Facility

Sysplex
Timer
19


Major DB2 Data Sharing Components

 Locking
– DB2 Subsystems in a group in one z/OS image share an IRLM address space
– IRLMs communicate through their LOCK1 structure
• groupname_LOCK1
– IRLMs also communicate via XCF
• DXRnnn groups
• XES locking services also use XCF
– IXCLOnnn groups
 Group Buffer pools
– 1 CF structure per GBP
• groupname_GBPn
• Members connect directly to GBP structures
 Shared Communications Area (SCA)
– Status sharing
• groupname_SCA
• Much lower activity than LOCK1 and GBPs
– Not usually considered for tuning
• Members connect directly to SCA structure

20


Locking - LOCK1 Structure

 Locks must be known and respected between members
– Data Sharing uses global locks to achieve this
 But not all locks need to be propagated to achieve this
– Only the most restrictive state needs be propagated
 Locking is propagated from IRLM, via XES to the LOCK1 CF structure
– IRLM knows about locking states that XES doesn’t
• XES only knows about “shared” and “exclusive” locks
• DB2 had many more states, even before Data Sharing

21



– LOCK1 Structure has two parts
• Lock table aka “Hash table”
– Each entry has:
> Resource name
> First system to have exclusive interest
> Flags for each system that has shared interest
– Entry size controls maximum number of connecting members
> 2 bytes up to 6 members
> 4 bytes up to 22 members
– Generally fewer entries than there are resources to lock
> Real resources hash to resource names
• Modified resource list
– Used for recovery purposes
– Less interesting – from a tuning perspective

22



 Contention types:
– Real Contention
• Different members really do need to use the same resource at the
same time
• Real application delay inherent while the holder retains the lock
– XES Contention
• When XES believes there is contention but IRLM knows there isn’t –
because of its more comprehensive view of locking
– IRLMs have to talk via XCF to resolve this - DXRnnn
– False Contention
• When the hashing algorithm for the lock table provides the same
hash value for two different resources
– XESs have to talk via XCF to resolve this - IXCLOnnn

23


Group Buffer Pools & Structures – How They Work - 1
 Inter-DB2 Read / Write interest:
– When one member has a write interest and at least one has a read
interest
– Tracked via Page Set Physical locks
• Always propagated to the LOCK1 structure, even when only one
member
– Some cost for a single member data sharing group
– If there is Inter-DB2 Read / Write interest in an object:
• The buffer pools in the members cooperate via the corresponding
Group Buffer Pool
– “GBP Dependency”
– Objects can go into and out of GBP Dependency
• “Read Only Switching” (RO Switching)
• Detected by Data Sharing Group
– PCLOSET and PCLOSEN parameters affect how often to check
• Affects GBP dependency
– Trade off between GBP Dependency time and RO Switching rate

24


 Updates cause Cross Invalidation
– DB2 members check their copy of a page is valid before using it
• Via a bitmap that each system’s XES maintains in memory
– To use an invalidated page the member retrieves it from the GBP
– Updates to the GBP generally happen at an application’s Commit point
• Synchronously forcing changed pages to the GBP
– “Force At Commit”
• Sometimes we get writes to the GBP when locking at row level
– To allow updated page to be retrieved by another member
 Local pool misses search the GBP first
– So the GBP can act as an extra layer of buffering
– Retrieval would generally be quicker than from disk / cache
 At intervals the Castout process purges some pages from the GBP
– Written back out through the local DB2 subsystems
– Threshold driven

25


 Installation can control regimes
– At Group Buffer Pool Level
– By object
 GBPCACHE(CHANGED)
• Only the writes are cached
 GBPCACHE(ALL)
– Cached as they are read
• Even if no Inter-DB2 R/W Interest
– Writes also cached
 GBPCACHE(NONE)
– No cacheing done
• Just used as a Cross-Invalidation mechanism
 GBPCACHE(SYSTEM)
– For a special kind of tablespace called a LOB
– Only certain types of pages are cached

26

Performance



Tuning Objectives

 Response Times
– Additional time could be caused by CF activities
• Also additional locking problems
 Throughput
– For a given capacity throughput might be less
 Minimised Cost
– Minimise the configuration needed for data sharing
• So long as that doesn’t conflict with other objectives
 Robustness
– Ensure performance doesn’t degrade with load

28


Parallel Sysplex Instrumentation
 XCF:
– SMF 74 Subtype 2
• RMF XCF Activity Report
– Applications
– Groups
– Paths
> CTCs are treated like real devices so SMF 74-1, 73 and 78-3 can be useful
– Members
> Job name in z/OS R.9
– DISPLAY XCF operator command
 Coupling Facility
– SMF 74 Subtype 4
• RMF Coupling Facility Activity Report
– Usage Summary Section – Structure sizes and CPU usage
– Structure Activity Section
– Subchannel Activity Section – Path / Subchannel information
– CF to CF Section – Duplexing traffic at the CF level

29


Data Sharing Instrumentation
 Accounting Trace
– Generally provides a time breakdown for each application
• Plan, Correlation ID and Package level
• Excellent tuning instrumentation for applications
– Includes:
• Global Lock Wait
• Time to retrieve pages from GBP
– Subsumed within Sync DB Wait and Async Read
• Time for commits
– Can involve GBP traffic
– Activities
• Group Buffer Pool
• Global Locking

 Statistics Trace
– Activities
• Group Buffer Pool
– Note: MXG change 25.075 required to support incompatibly-changed DB2 Version 8 GBP
statistics
• Global Locking
30


Capacity and “White Space”

 “White Space” is capacity which needs to be kept free for
oncoming work from other Coupling Facilities
– Memory, CPU and links
– Duplexing reduces the need for this

 Coupling Facility Control Code requires CPU utilisation below
about 50%
– Above that response times begin to degrade
• With impact on coupling cost to z/OS images and on response
times

31


Link Performance

 Configure the fastest suitable links
– Type: IC vs ICB vs ISC vs CIB vs CIB LR
– Generation: eg ISC-3 vs ISC-2
 Configure enough of them
 Use Peer Mode
 Monitor for
– Signalling response times
– Path Busy Conditions
– Subchannel Busy Conditions
– Request Failures

32


XCF Tuning
 Aim to reduce transfer times
– “Mean Transfer Time” MXFER TIME in RMF
 Aim to minimise traffic
– Rates at all levels in RMF
– Eg Minimise Locking False Contention
– Eg Set up GRS Star in a way that minimises ENQs
 Optimise the use of links
– More modern CF-based links tend to outperform CTCs
• But CTCs still better for small messages
• CTCs drive SAP utilisation
– RMF counts the number of times each path was chosen
• Understand why signals use the paths in the ratio they do
 Transport Class buffer sizes
– Buffers that are too big waste memory
– Buffers that are too small have to be expanded
• Sometimes with cost
– RMF has counts of “Fit”, “Small”, “Big” and “Big With Overhead” messages
– RMF lists transport classes and their maximum buffer size values

33


Group Buffer Pool Tuning
 GBP tuning has similarities to local pool tuning
– But some twists
 Important to minimise traffic
– Application GETPAGEs in general
– Traffic to GBPs
 Also important to minimise response times
– Which is mainly a matter of tuning the underlying CF access
 Minimising the amount of data actually shared may be practical
– For many designs it isn’t
 Important to avoid invalidations due to too few directory entries
– GBP space divided into Directory entries and Data elements
• Directory entry reclaims if too few entries
– Causes invalidations of local buffers
• Installation can alter the balance
• Installation can increase the size of the group buffer pool

34


Group Buffer Pool Tuning - Traffic
 Reads:
– Cross invalidation reads
• Data returned i.e. is in GBP
• Data not returned i.e. is known to be down level but page not in the GBP
– Requires disk read
– Buffer pool miss reads
• Data returned i.e not in the local pool but is in the GBP
• Data not returned i.e is in neither the local pool nor the GBP
– A bigger GBP ought to provide more hits and fewer misses
• But I rarely see high GBP hit ratios
 Writes:
– Avoid GBPCACHE(NONE) as writes are SYNCHRONOUS TO DISK at Commit time
• Harmful to the Committing unit of work’s performance
– Writes can also be caused by the LOCAL pool’s Deferred Write thresholds being hit
• In this case Commits aren’t waited for

 Castouts:
– Dribbling out a good idea
• Just like for local pools

35


Locking Tuning

 It’s important to reduce locking traffic at all levels
– Application
– DB2 subsystem
 It’s also important to reduce False Contention
– Usually by increasing the Lock Table portion of the LOCK1 structure
• Number of entries will be a power of 2
– 4-byte lock table entry means fewer entries for same size than 2-byte
 It’s nice that in DB2 Version 8 there’s a remapping of IRLM lock states to
XES ones
– May reduce XES lock contention
 CF Request response times also important

36


Special Coupling Facility Commands

 A number of special commands have been introduced to make CF
requests more efficient
– Generally have CFLEVEL prerequisites
– READ_COCLASS
• DB2 Version 6
– WARM
• DB2 Version 8
– RFCOM
• DB2 Version 8
 DB2 Version 8 Statistics Trace instruments WARM and RFCOM

37


DB2 Version 9

 Restart performance improved
 Command to remove GBP Dependency at object level
– ACCESS DB MODE(NGBPDEP)
– Typical usage would be before batch run
• Issue on member where you plan to run the job
 Improved performance for GBP writes
 DB2 overall health taken into account for WLM routing
 Balance Group Attach connections across members on same LPAR
– Usermod to Versions 7 and 8
 etc…

38


Summary
 Parallel Sysplex has many benefits

– More fully realised with Data Sharing
 Need to manage carefully performance and cost

 Configuration choices make an enormous difference

 Avoid shared coupling facilities for Production

 Good monitoring tools for both z/OS / Hardware, and DB2

 Tune not only DB2 structure and XCF Performance

– But also other structures and users of XCF

39

DB2 Data Sharing Performance Beginners Guide

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Viewers also liked

Viewers also liked (17)

Similar to DB2 Data Sharing Performance Beginners Guide

Similar to DB2 Data Sharing Performance Beginners Guide (20)

More from Martin Packer

More from Martin Packer (20)

Recently uploaded

Recently uploaded (20)

DB2 Data Sharing Performance Beginners Guide