Expanding the Horizons of Mainframe I/O
Today, analysts who are responsible for z/OS I/O subsystems face a changing environment and a variety of new technologies as traditional mainframe environments attempt to exploit new technologies such as FICON Express16S, High Performance FICON for z, EC12 and z13 processors, flash drives, Extended Address Volumes, encryption, multi-site remote copy schemas, and storage tiering.
To respond to these challenges and to exploit these new technologies, it is necessary to understand the underlying purpose and critical features of z/Architecture hardware and software. This 4 day seminar provides a unique opportunity to learn about these significant developments in z/OS I/O Subsystems.
Syllabus updated 15 January 2015.
Audience
This seminar is intended for systems programmers, performance analysts, capacity planners, and other systems personnel who are responsible for selecting, configuring, optimizing, or managing z/OS subsystems.
Syllabus
z/Architecture
What are the primary building blocks of the z/Architecture? What measurement
facilities are available?
Comparing OS/390 and z/Architecture
z/Architecture ORB and CMB
zSeries subchannel measurements
Intelligent Resource Director
Prefix CCW
MIDAWs
Extended Address Volumes
Encryption manager
zSeries Servers
What are the architectural characteristics of the zSeries servers. How does
the EC12 processors expand on the z900 thru z196 architectures?
zSeries logical structure
zSeries MCM
I/O books
FICON Express2, Express4, Express8, and Express8S Channels
zHPF channels
Coupling facility
z10 architecture
I/O elongation
Basic Queueing Theory
What are the basic queueing concepts that are required for I/O
performance analysis?
Utilization law
Little's law
Response time equation
Enterprise Storage Devices
The characteristics of rotating and flash storage will be examined
and the seminar will explore how the technologies can best be employed to
address the performance requirements of the enterprise. What are Unrecoverable
Errors and why are they important? Specific attention will be placed on
the long term trends for reliability, capacity, power consumption, and price.
Overview
SATA, SAS, and FC
3.5" and 2.5" form factors
FBA encryption
Reliability
Flash device architecture
Real world reliability experiences
Virtual or Virtualize
Virtualization is the fundamental building block of contemporary
storage subsystems. The basic concepts of virtual memory are reviewed and
a simple model is proposed for virtualizing CKD volumes to FBA devices.
Fundamentals
Paging and segmentation with paging
CKD volume structure
Mapping CKD volumes to FBA devices
Fixed and thin provisioning
Counters and reference bits
Storage Protection Schemas
What storage protection schemas are available for enterprise storage
subsystems? Given the increasing statutory and regulatory requirements for
data retention, it is imperative to understand the available protection
schemas.
Forever is a long long time!
RAID concepts
MAID concepts
XIV storage architectures
Logical Volume Skewing
What is Logical volume skewing and how may it be characterized? What are
its influences on cache size?
What is skew?
Sources of skew
Characterizing DASD skew
RMF Device Activity Report
Caching Concepts
Why is caching a good idea and what caching strategies can maximize
the benefits of caching for different reference patterns? What is locality
of reference and what is the role of the define extent CCW in cached subsystems?
Locality of reference
The define extent CCW
Normal cache replacement
Sequential
Detected sequential
Record level caching
Predicting cache behavior
Tiered Storage Architectures
Contemporary storage subsystem designs incorporate multiple levels
of storage devices ranging from SATA to flash. How can allocations be controlled
or logical volumes and/or extents be migrated to improve subsystem performance?
The storage pyramid
Controlling initial allocations
SMS allocation extensions
Recognizing volume and extent misplacements
Algorithmic priorities
Relocating volumes and extents
FICON Connectivity and Performance
FICON represents the next step in I/O subsystem connectivity. The logical
and physical characteristics of FICON connectivity are introduced. Particular
emphasis is placed on how FICON addresses ESCON's distance related performance
problems and how cascaded FICON directors will simplify GDPS design.
FICON and Fibre Channel
FICON building blocks
FC-SB-2 and FC-SB-3
FICON topology and terminology
FICON architecture and addressing
FICON I/O definitions
Cascaded directors and ISLs
Configuration design
Managing Complex FICON Configurations
Performance
Implementation considerations
Extended Address Volumes
EAVs offer an elegant solution to historical problems of logical volume
management. This section will examine the characteristics of addressing
schema employed for EAVs.
Traditional CCHH addresses
EAV acrchitecture
Dataset support
Dynamic volume expansion
High Performance FICON for z
zHPF offers the potential of exploiting the performance advantages of native
fibre channel to enhance the performance of FICON connectivity. Emphasis
will be placed on understanding how existing channel program structures
are encapsulated to reduce channel microprocessor utilization and increase
effective data transfer rates.
FC-SB-4
Channel program encapsulation
Sample CCW and TCW sequences
Performance Considerations
PAVs, Multiple Allegiances, and HyperPAVs
PAVs and multiple allegiances address the historical problem of UCB serialization.
Special emphasis is placed on the development of a back-of-the-envelope
model to predict the potential performance benefits of PAVs and HyperPAVs.
Architectural overview
Back-of-the-envelope model
HyperPAV management
What about EAVs?
Long DISC Time
What causes long DISC times on contemporary storage subsystems?
How are the cache-hit ratio and the RMF average DISC time related?
Sherkow's algorithm
Recomputing DISC times
Why do long DISC times occur?
Addressing DISC time with flash
Implications of Logical and Physical Volume Size
Perhaps one of the most common concerns of analysts is the impact of logical
and physical volume size on the performance of subsystems. An analysis based
on simulation modeling is employed to examine the impacts of logical and
physical volume size as we enter the era of EAVs. Emphasis will also be
placed on the identification of candidate datasets for flash storage.
The evolution of logical volume size
Modeling logical volume performance
Identifying flash candidates
Implications of SCSI device size
Storage architecture design using HyperPAVs and EAVs
Recommendations
Remote Copy Planning
As a result of increased security concerns and regulatory requirements,
more installations are considering the implementation of Metro Mirror, Global
Mirror, XRC, HRC, or SRDF solutions for real time data replication. What
are the characteristics of these solutions and how can installations identify
and respond to problem volumes and datasets, distance concerns, and management
issues? How can the new GDPS distance limits be exploited to provide enhanced
protection?
Local Mirror architecture
XRC architecture
XRC exploitation of zIIPs
SRDF architecture
Problem volumes and datasets
Protocol and distances impacts
Fibre channel and IP connections
Management issues
Tape and Virtual Tape
New physical tape architectures and media encryption offer solutions to
application data transfer requirements and the secure encryption of media
that is transported beyond the bounds of the data center. Virtual tape subsystems
offer potential solutions for the small tape management problem. Emphasis
will be placed on a review of the currently available hardware and software
solutions.
High performance tape architectures
Encryption
Why virtual tape?
IBM VTS 7720
STK VSM6
De-Duplication
Remote Replication
Logical Control Units
What is a logical control unit? How are they defined?
What is an LCU?
How do LCUs and LSSs relate?
I/O Configuration Definition
The use of HCD and IOCP to define the physical connectivity of I/O subsystems
is discussed in detail. Examples are provided for contemporary DASD subsystems.
Particular emphasis is placed on performance impacts of IOP and channel
path assignment.
IOCP and HCD concepts
Multiple subchannel sets
Defining complex subsystems
Channel subsystem and storage subsystem interfaces limitations
Path Selection
The path initialization process is discussed in detail. Specific emphasis
is placed on how the channel subsystem identifies and interrogates the control
unit, the role of the logical path table in the control unit, and how the
outbound and inbound path selection processes work.
Channel and subsystem initialization
Step by step analysis of path selection
zHPF and path selection
Storage Tiering
Storage tiering offers the opportunity automatically assign volume segments
to devices with disparate performance expectations within a subsystem. This
approach offers the opportunity to improve performance and reduce cost.
The SMS management facilities for tier management under z/OS 2.1 are discussed.
The objective
Thermal models
Paging or swaping
Implementations
Contemporary Storage Subsystems
A detailed comparative analysis of contemporary storage subsystems for z/OS
environments is provided.
EMC V-Max 40K
HDS VSP G1000
IBM DS8880