06 z13 Retrospektive und Ausblick - Thomas Wolf · PDF fileaccess new z/OS JSON services ... IBM Java 8: CPU-Intensive Benchmark (Controlled measurement environment, results may vary)

© 2016 IBM Corporation 1

Agenda

• z13 Retrospective• Positioning and goals• Key design changes

• z13 Experience and outlook for future generations• Performance• System Stack

• Summary


IBM z13 Platform Positioning

• The world’s premier transaction and data engine now enabled for the mobilegeneration

• The integrated transaction and analytics system for right-time insights at the point of impact

• The world’s most efficient and trusted cloud system that transforms the economics of IT

Platform Core Capabilities:

Transaction Processing

Data Serving

Mixed Workloads

Operational Efficiency

Trusted and Secure Computing

Reliable, Available, Resilient

Virtually Limitless Scale


“The end of the GHz Race”

§ x86 and POWER already transitioned to a throughput-centric model§ Frequency peaked for x86 in 2005, POWER in 2008

§ In System z we have held this off with unique cooling, packaging, technology, and design solutions

The end of the GHZ race


0

1000

2000

3000

4000

5000

6000

z900 z990 z9ec z10ec z196 zEC12 zNextEC

770 MHz1.2 GHz

1.7 GHz

4.4 GHz

5.2 GHz 5.0 GHz

5.5 GHz

2000z900

189 nm SOI16 Cores**Full 64-bit

z/Architecture

2003z990

130 nm SOI32 Cores**Superscalar

Modular SMP

2005z9 EC

90 nm SOI54 Cores**

System level scaling

2012zEC12

32 nm SOI101 Cores**

OOO and eDRAMcache

improvementsPCIe Flash

Arch extensionsfor scaling

2010z196

45 nm SOI80 Cores**OOO core

eDRAM cacheRAIM memoryzBX integration

2008z10 EC

65 nm SOI64 Cores**

High-freq core3-level cache

MH

z/G

Hz

1000

0

2000

3000

4000

5000

6000

1695*+12%GHz-9%1202*

+33%GHz

+18%

1514*+26%GHz+6%

902*+50%GHz

+159%

* MIPS Tables are NOT adequate for making comparisons of z Systems processors. Additional capacity planning required** Number of PU cores for customer use

CMOS Mainframe Evolution

2015z13

22 nm SOI141 Cores**SMT &SIMD

Up to 10TB of Memory


z13 Balanced System Design

* No server can fully exploit its maximum I/O bandwidth

PCI – Processor Capacity Index (IBM MIPS) 141-way

Customer Processors

PCI for1-way1695

Memory10 TB

System I/O Bandwidth832 GB/Sec*

80-way

64-way54-way

1.5 TB

512 GB 1202902600

288 GB/sec*

172.8 GB/sec*

1514

3TB

384 GB/Sec*

101-wayz10 EC

z9 EC

z196

zEC12

z13


z13 Highlights

10% Single thread capacity improvement over zEC12

Up to40%

Total capacity improvement over zEC121

Up to10 TB

3X more available memory to help z/OS or Linux workloads

2SODs

zKVM * and GDPS virtual appliance for Linux on z* opens the door for more Linux

Up to141

Configurable cores – CP, zIIP, IFLs, ICFs, SAP

• Performance, scale, intelligent I/O, and security enhancements to support transaction growth in the mobile world

• More memory, new cache design, improved I/O bandwidth, and compression help to serve up more data for analytics

• Enterprise grade Linux solution, open standards, enhanced sharing, and focus on business continuity to support cloud

IBM z13Machine Type: 2964

Models: N30, N63, N96, NC9, NE1

Upgradeable from z196 and zEC12

* All statements regarding IBM's future direction and intent are subject to change or withdrawal without notice, and represent goals and objectives only.

1 Based on preliminary internal measurements and projections. Official performance data will be available upon announce and can be obtained online at LSPR (Large Systems Performance Reference) website at: https://www-304.ibm.com/servers/resourcelink/lib03060.nsf/pages/lsprindex?OpenDocument . Actual performance results may vary by customer based on individual workload, configuration and software levels


z13 Architecture Extensions

§ Simultaneousmultithreading(SMT)operation– Uptotwoactiveexecutionthreadspercorecandynamicallysharethecaches,TLBsandexecutionresourcesofeachIFLandzIIP core.

§ Coremicro-architectureradicallyalteredtoincreaseparallelism– NewbranchpredictionandinstructionfetchfrontendtosupportSMTandtoimprovebranchpredictionthroughput.

– Widerinstructiondecode,dispatchandcompletionbandwidth:IncreasedtosixinstructionspercyclecomparedtothreeonzEC12

– Largerinstructionissuebandwidth:Increasedtoupto10instructionsissuedpercycle(2branch,4FXU,2LSU,2BFU/DFU/SIMD)comparedto7onzEC12

– Greaterintegerandfloatingpointexecutionbandwidth

§ SingleInstructionMultipleData(SIMD)instructionsetandexecution– Datatypes:Integer:bytetoquad-word;String:8,16,32bit;binaryfloatingpoint– Newinstructions(139)includestringoperations,vectorintegerandvectorfloatingpointoperations– FloatingPointInstructionsoperateonnewlyarchitectedvectorregisters(32new128-bitregisters)


Acceleration through special purpose hardware

§ On-processor– Crypto (CPACF), Compression, SIMD– Tight, synchronous integration with instruction stream

§ PCIe Gen3− Accessible and sharable by all processors− Faster time to market for new functions− Compression (zEDC), Crypto, Flash Express

§ Network Acceleration– RDMA over Converged Ethernet SMC-R (RoCE)

§ Integrated External Accelerators− Integrated by Software− IBM DB2 Analytics Accelerator for DB2 Query

Acceleration

§ Specialty Engines and Firmware Partitions− Leverage Flat SMP Design, enable price flexibility− zIIP for DB2 and Java, IFL for Linux on z Systems− IBM zAware

IBM DB2 Analytics Accelerator built on Netezza Technology


z13 PU and SCM Assembly

PU Chip

SC SCM withThermal Module

CappedPU

Fully assembled CPC Drawerwith the chilled water

supply manifold lifted to the left

2x SC SCMs(Air Cooled)

6x PU SCMsunder the cold-plates

Front of CPC

Draw

erfor Fanouts/FSPs

6x PU SCMs

SC ChipCappedSC

4 CPC Drawer System connectivity

NN

NN

NN

NN

Draw

er 2D

rawer 1

Draw

er 3D

rawer 4


Frame and Packaging

InternalBatteries(optional)

PowerSupplies

N+2 Radiator Pumps

PCIe I/O drawers numbers 1 to 4

(Note: for an upgraded System, drawer slots 1 and 2 are used for the

I/O Drawer

2 x 1U Support Elements

CPC Drawers, PCIeFanouts, Cooling water manifold and pipes, PCIeI/O interconnect cables, FSPs and ethernet cables

PCIe I/O drawer number 5

System Control Hubs (used to be BPHs)

Overhead Power Cables

(option)

Overhead I/O featureis a co-req foroverhead power option

Displays and keyboards for Support Elements


z13 Physical Planning

• Extend / Maintain zEC12 Datacenter Characteristics– 2 frame base system (CEC, I/O, service system and PP&C)

– No significant increase in weight

– Maintain floor tile cutouts for raised floor system • Same as z10 EC, z196, and zEC12

– Same number of power cords (2 or 4) as “equivalent” zEC12 config

– Maintain 27.5 kW box max input power • Same as z10 EC, z196, and zEC12

– Maintain DC input power capability, overhead I/O cabling option, add overhead power option

• Better control of energy usage and improved efficiency in your data center– Support for relaxed datacenter requirements

ASHREA A2 - maximum 35°C air-inlet temperature (zEC12 32°C), water temp 24°C (zEC12 20°C)

– Upgraded cooling systems compared to zEC12 with N+2 pumps and blowers


Agenda

• z13 Retrospective• Positioning and goals• Key design changes

• z13 Experience and outlook for future generations• Performance• System Stack

• Summary


Experiences with z13 architecture

§ Hardware/Software co-optimization is increasingly important to performance– Performance gains from technology scaling have ended– Hardware performance gains are coming from design

• Micro-architectural innovation (and complexity): Instr. pipeline, Branch Prediction…• New instructions and architected features: SIMD, SMT…

– Coding practices and software exploitation needed to get the full value of the hardware

§ Compilers biggest single performance lever for many applications – Aggressive use of the latest compiler technology– ARCH and TUNE options optimize for current hardware designs

• ARCH needs to match oldest hardware level supported • TUNE should match hardware level for which you care most about performance

– Use higher levels of OPT to get the best performance:• At least on performance-sensitive components

§ IF compilation not possible or not appropriate: ABO for zOS – Performance improvements for compiled Cobol programs


IBM Automatic Binary Optimizer (ABO) for z/OSImprove Performance of Compiled COBOL Programs

ABO FeaturesInternal &CustomerPerformanceImprovementsMeasuring~15% ü

No SourceCode,MigrationorPerformanceOptionsTuningRequired ü

Targets LatestIBMzSystems:zEC12,zBC12,z13,z13srunning z/OS2.1or z/OS2.2 ü

AllIBMEnterpriseCOBOLv3 &v4Compiled ProgramsAreEligibleForOptimization ü

OptimizedPrograms GuaranteedToBeFunctionallyEquivalent ü

IBM ProblemDeterminationToolingSupport+WorkingWith SeveralKey3rd PartyToolingVendorsInOurBetaProgram

ü

Leverages newz/OS2.2InfrastructureToTargetMultipleHardwareLevelsAutomatically ü

Optimizer

Original Program Binaries(Base ESA390)

Optimized Program Binaries(Latest z Systems)


OS focused on exploiting HW innovation

z/OS Version 2.1

§ Improved price performance for zIIP workloads with SMT§ Support new analytics workloads with SIMD§ New Crypto capabilities for faster encryption§ Large memory to improve performance and enable new

applications

z/VMVersion 6.3

§ Improved price performance with support for multi-threading technology

§ Support for twice as many processors (z13 only)§ Improved systems management and economics § Embracing Open Standards and Open Source Interoperability § Supports more virtual servers than any other platform in a

single footprint

z/VSEVersion 5.1

§ Reduced risk of access from unauthorized users § Reduced memory constraints§ Wide portfolio using Linux on z Systems § Continued system usability enhancements with CICS Explorer§ More efficient communications

Linux on z System

§ Multithreading allows for per core software savings§ Ability to host and manage more workloads efficient and cost-

effective§ Automatic identification of unusual messages§ Integrated continuous availability & disaster recovery solution


Compilers on z Systems• IBMcontinuestoinvestinthecompilerportfolioonz:

• Increasedfocusonapplicationprogramperformanceinrecentyears• Continuedadvancementsinlanguagesandoperatingsystems

• Java/JIT,C/C++,COBOL,PL/I,Linux,z/OS

• Optional feature of z/OS 2.2

• Provides system programming capabilities with Metal C option

• Fully Supports z/Architecture, including z13 & z13s processors

• Ships with High performance Math Libraries tuned for z13

z/OS V2.2XL C/C++

Enterprise PL/I for z/OS V4.5

• Critical Business Language – Committed to invest in leading-edge technology

• Shipped a new release every year since 1999


• Provide full support for JSON (Parse, Generate, and Validate)

*Up to 17% reduction in CPU time*

*Up to 24% increase in throughput*

XL C/C++ for Linux on z Systems V1.2

• New compiler based on Clang and IBM optimization technology


• Provide easy migration of C/C++ applications to System z

*Up to 14% increase in performance over GCC*

* The performance improvements are based on internal IBM lab measurements. Performance results for specific applications will vary, depending on the source code, the compiler options specified, and other factors

Enterprise COBOL for z/OS V5.2

• Leverage SIMD instructions to improve processing of certain COBOL statements.

• Increased use of DFP instructions for Packed Decimal data

• Support COBOL 2002 language features: SORT and table SORT statements

• Allows applications to access new z/OS JSON services

*Up to 14% reduction in CPU time*


IBM Java 8: CPU-Intensive Benchmark

(Controlled measurement environment, results may vary)

Common Java CPU Intensive Benchmarksz13 with SMT zIIPs Java 8 vs. zEC12 Java 7 SR4

IBM z13 and IBM Java 8 show a composite improvement of 61% over zEC12 and Java7 SR4 running the CPU Intensive benchmark

0%

200%

400%

Compression

Cryptography

Database MP3 Library

Scientific FP

Serialization

Graphics XML

35%

362%

41%7%

44% 59% 61%

Composite

29%59%


Client Use Cases for SMT

European Bank (Germany)• Hosts its Internet Banking solution (WebSphere Portal) on IBM z Systems.

• It utilizes 30 zIIPs

• SMT exploitation for ZIIPs:

• Up to +40% extra capacity during business hours

European Bank (France) • SMT exploitation for ZIIPs and IFLs: + 20% capacity

European Accounting Company (Germany)• SMT exploitation for ZIIPs: + 20% capacity


Client Use Cases for SIMD

Scenario: Rewrite ASM-PGMs to exploit SIMD• Validate that input files and the data in the files have a legal format &

move the data into a different data structure • Co-operation with IBM• Reduction in CPU time by 30%• Effort: high

Challenge: New application from ISV• Application runs streaming videos at all branch offices• Freeware FFmpeg for recoding new video clips, very CPU consuming• Conversion of MPEG file

• zEC12: 15min, offloaded to x86• z13, recompile: 2min, executed on z13

• Effort: low

7x improve

EuropeanAccounting Company

EuropeanBank


More memory makes a difference§ Up to 10 TB available memory

§ Transparently support shrinking batch windows and meet service goals – no change to applications needed to gain benefits

§ Get more work done – online transaction processing can experience up to 70% reduction in response time with more memory

§ Improve system performance, minimize constraints and simplify management of applications with database middleware exploitation of additional memory

§ Achieve faster decision making with the advantages of in memory data


SSI: Online banking workload 12w DB2 V11 z/OS1.13

Memory BP Size CPU % ITR ITR Delta ETR ETR

Delta

Txn

response time(sec)

Response

time delta

Sync

Read IO/sec

Sync IO

delta

256 GB 160 GB 72 992 n/a 709 n/a .695 n/a 38.4k n/a

512 GB 320 GB 73 1124 13.3% 819 15.5% .428 -38% 11.7k -69%

1024 GB 638 GB 79 1237 24.7% 976 37.7% .209 -70% 0.9k -97%

1TB Study

0.00

200.00

400.00

600.00

800.00

1,000.00

1,200.00

1,400.00

0 100 200 300 400 500 600 700BPool (GB)

Met

rics(

adj

uste

d fo

r gra

ph)

ITRETRDBReq/DSsync Read/sec


Client Use Cases for Large Memory

• Tested by several European clients

• Large memory exploitation

• VSAM/RLS, VSAM/SMB• DB2 Buffer Pools• MQ Series V8• CTG – Large Pages• Steady Sort Performance

• All CICS Storage

European Bank: DB2 Workloads• By just increasing the memory and let DB2 use

that for Large pages and increase buffer pool size • Client has seen up to 5% MIPS reduction for DB2• Quote: “We prefer to buy storage to save MIPS”

European Insurance Company: DB2 Workload• By increasing the DB2 Buffer Pools, in fact

quadrupling the amount of memory needed, the client was able to decrease their I/O Rate by 30%

• Simulations indicate the potential for further 36% reduction in I/O. à ordered additional 640 GB


OSA/LAN IP network with SMC-D and SMC-R

23

OSA ISM

TCP

IP

Interface

Sockets

Middleware/Application

z/OS System B

OSAISM

TCP

IP

Sockets


z/OS System A

IP Network (Ethernet)

ISM VCHID (within System z)

System z13

OSA,ISMandROCEhavethesamePNet

IDe.g.PNET1

• 23

SMC-DSMC-D

ROCE OSAROCE

TCP

IP

Sockets


z/OS System C

RDMA network ROCE

System z13 / zEC12 / zBC12

SMC-RSMC-R

Note. This figure is another view of chart 13.


Shared Memory Communications architecture Faster communications that preserve TCP/IP qualities of service

• Shared Memory Communications – Direct Memory Access (SMC-D) optimizes z/OS for improved performance in ‘within-the-box’ communications versus standard TCP/IP over HiperSockets or Open System Adapter

Typical Client Use Cases:• Valuable for multi-tiered work co-located onto a single z

Systems server without requiring extra hardware• Any z/OS TCP sockets based workload can seamlessly

use SMC-D without requiring any application changes

Up to 61% CPU savings for FTP file transfers across z/OS systems versus HiperSockets*

Up to 9x improvement in throughput with more than a

88% decrease in CPU consumption and a 90%

decrease in response time for streaming workloads versus

using HiperSockets*

Up to 91% improvement in throughput and up to 48%

improvement in response time for interactive workloads versus

using HiperSockets*SMC Applicability Tool (SMCAT) is available to assist in gaining additional insight

into the applicability of SMC-D (and SMC-R) for your environment

* All performance information was determined in a controlled environment. Actual results may vary. Performance information is provided “AS IS” and no warranties or guarantees are expressed or implied by IBM.


SMC-D to Hipersockets comparison

Up to 9x the throughput! See breakout summary on next chart.


SMC-D to OSA comparison

Up to 21x the throughput! See breakout summary on next chart.


z Systems Stack Optimization

Java

Semiconductor Technology

Microprocessor Design

Systems Design

Virtualization & Operating Systems

Compilers & Java Virtual Machine

Optimized Middleware


IBM LinuxONE: Linux without Limits

IBM LINUXONE SYSTEMS

IBM LINUXONE SOLUTIONS

Cloud DevOps

Simplification: Dynamic Partition Manager

zKVM

Blockchain within Secure Service Containers

Mobile Analytics

IBM LinuxONE Elastic Pricing


The world without Appliances


The mainframe approach: zACI

Hardware

Firmware z13GA2ü

Mgm

tUI

Management Backend

Application Interfaces

Base MCP Linux

Application

zACI SW Framework

zACI Installer zACI Loader


Our Approach: zACI

Hardware

Firmware

zACI Installer zACI Loader

z13GA2ü

any Appliance


IBM Secure Service ContainerThe base infrastructure to host and build software appliances

• Provides simplified mechanism for fast deployment and management of packaged solutions

• Provides tamper protection during Appliance installation and runtime

• Ensure confidentiality of data and code running within the appliance – both at flight and at rest

• Management provided via Remote APIs (RESTful) and web interfaces

• Enables appliances to be delivered via distribution channels

Services

Applications

Operating SystemManagem

ent


Summary§ GHZ race is over but significant performance and capacity increases will continue

§ Special accelerators: On-chip, on-board/drawer, external ….§ Special PCI based IO adaptors: RoCe, Crypto, EDC…§ New architectural designs: SIMD, SMT…

§ Balanced System: MP Capacity, IO, Large Memory growth

§ Total stack optimization to leverage new architectural functions − New compilers to exploit hardware features − Run-time optimizer tools − OS and middle-ware exploitation

§ Continued focus on Simplification and New Workloads − Simplification to enable new customers and/or to mitigate skill issues

− Opportunities in new growth areas: Mobile, Analytics, Security …

§ Secure Service container vs firmware delivery − Enables new ways of business with rapid and simplified install/deployment

and secure execution− Enable new business models and new workloads: zAware, zaaS,

Blockchain

Semiconductor Technology

Microprocessor Design

Systems Design

Virtualization & Operating Systems

Compilers & Java Virtual Machine

Optimized Middleware

Services

Application

OperatingSystemM

anagem

ent


ThankYou!

© 2014 IBM Corporation

Backup

35

© 2014 IBM Corporation© 2016 IBM Corporation 36

Other resources

Like this stuff? There’s lots more available:§ Microprocessor Optimization Primer

– Available under IBM Developerworks’ LinuxOne community• https://www.ibm.com/developerworks/community/groups/community/lozopensource

§ CPU Measurement Facilities– User-accessible hardware instrumentation data to understand performance characteristics– Documentation and education materials can be found on online, some references:

• For z/OS http://www-03.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/TC000066– (supported under Hardware Instrumentation Services - HIS)

• For z/VM http://www.vm.ibm.com/perf/tips/cpumf.html

§ Other related references– “z/Architecture: Principles of operation,” Int. Bus. Mach. (IBM) Corp., Armonk, NY, USA,

Order No. SA22-7832-10, Feb. 2015. [Online]– Dan Greiner’s presentations of z/Architecture features with SHARE– John R. Ehrman's book: Assembler Language Programming for IBM z System Servers– “The IBM z13 multithreaded microprocessor,” in IBM J. Res. & Dev., pp. 1:1–1:13, 2015

§ ABO: http://www-03.ibm.com/software/products/en/z-compilers-optimizer


zSystems Announcement Summary

OperatingSystemsNEWz/OSV2.2SPE

NEW z/VM6.4PreviewNEW IBMWave1.2forz/VMSOD

UPDATE KVMforIBMzSystemsv1.1.1

HYBRIDCLOUD•NEW z/OSCloudBeta(MiddlewareasaService- WAS,DB2,CICS,IMS)

•UPDATEz/OSConnectEnterpriseEdition•NEW StrongLoop &LoopbackConnectors•NEWAPIConnect•NEW Cloudant•RECENTSUSECloudOpenStack 6

DEVOPS/MODERNIZATION•NEW ApplicationDeliveryIntelligence(TestOptimization)

•UPDATE COBOLv6.1•NEWApplicationDeliveryFoundation(RDz +PDTools)

•RECENTAutomaticBinaryOptimizer•RECENT CICS5.3(withLiberty);IMS14•NEWWebSphere ApplicationServerforz/OS•NEW HACPv12•UPDATE RationalProgrammingPatterns•UPDATE IBMWorkloadScheduler•NEW IBMTivoliOMEGAMONXEonz/OSMonitoringFeatureforJVM

•UPDATE IBMServiceMgmt Suiteforz/OS•UPDATE IBMAdvancedStorageMgmt Suite•UPDATE IBMPerformanceMgmt Suiteforz/OS•UPDATEIBMServiceManagementUnite•UPDATE UrbanCode Deploy6.2.1•NEW IBMExplorerforz/OS3.0Aqua

ANALYTICS• RECENT Zementis in-transactionanalytics• RECENT Sparkforz/OSandMainframeDataServiceforSpark

• RECENT DB2AnalyticsAccelerator• RECENTODMDecisionServerInsightsforz• NEW IBMBigInsights forHadoop• RECENTIBMBigInsights BigIntegrate andBigQuality forzSystems

• NEWIBMOperationsAnalytics• NEW IBMzOperationalInsightsv.Next Beta

CYBERSECURITY•NEWMultifactorAuthenticationforz/OS•NEWCyberSecurityAnalyticsBeta•NEW ReadyforSecurityIntelligencePartnerProgram

•NEW SecurityWorkloadPricing•UPDATE IBMSecuritySWforzSystems(Guardium,QRadar,zSecure)

37

New!IBMz13s

Update!IBMz13GA2

© 2014 IBM Corporation

The following are trademarks of International Business Machines Corporation in the United States, or other countries, or both.

Other company, product, and service names may be trademarks or service marks of others. All information contained in this document is subject to change without notice. The products described in this document are NOT intended for use in implantation, life support, space, nuclear, or military applications where malfunction may result in injury or death to persons.The information contained in this document does not affect or change IBM product specifications or warranties. Nothing in this document shall operate as an express or implied license or indemnity under the intellectual property rights of IBM or third parties. All information contained in this document was obtained in specific environments, and is presented as an illustration. The results obtained in other operating environments may vary.

THE INFORMATION CONTAINED IN THIS DOCUMENT IS PROVIDED ON AN “AS IS” BASIS. In no event will IBM be liable for damages arising directly or indirectly from any use of the information contained in this document.

IBM Systems & Technology Group2070 Route 52, Bldg. 330Hopewell Junction, NY 12533-6351

38

® ®®®®

Documents

06 z13 Retrospektive und Ausblick - Thomas Wolf · PDF fileaccess new z/OS JSON services ... IBM Java 8: CPU-Intensive Benchmark (Controlled measurement environment, results may vary)