pgBench benchmark on PostgreSQL 9.6.1 with up to 1.7x more throughput per core1 than x86
YCSB benchmark on MongoDB 3.4.1 with up to2.2x more throughput per core2 than x86
DayTrader benchmark on WebSphere Application Server 8.5.5.9 with up to 1.6x more throughput per core3
than x86
DayTrader Apache TomEE 1.7.1 with up to 2x more throughput per core4 than x86
Acme Air benchmark on Node.js 6.10 with up to 2.2x more throughput per core5 than x86
MicroBM CPU benchmark on InfoSphere® DataStage®
11.5 with up to 2.5x more throughput per core6 than x86
Your business logic runs fasterJava on LinuxONE performs 1.5x faster7 than on x86 platforms
1,2,3,4,5,6,7 All claims noted on this slide are based on IBM Internal measurements. Results may vary. See Footnote page at end of presentation. Additional information is available upon request
Value Proof
Scale that maps to your growth vision
Value Proof
Grow your data serving and your workloads without growing your data center:
• Standard 19” Rack• 16U Reserved space can be used for
storage or other devices
240 concurrent databases executing a total of 58 billion database transactions per day1 on a single LinuxONE Emperor II system
Scale out to 30 cores running MongoDB databases under z/VM 6.4, each with a constant throughput and not more than 10 μs latency increase per additional MongoDB instance2
Scale-out to 330,000 Docker containers3 in a single LinuxONE Emperor II system, no application server farms necessary
Scale-up a single MongoDB instance to 6.8 TB in a single system without database sharding and get 2.4x more throughput and 2.3x lower latency4 compared to LinuxONE Emperor, which was already better than x86
Your business logic runs more predictably and consistently as you grow
Emperor II has hardware-based pause-less garbage collection5, which minimizes application pauses that can disrupt response times and limit scale
• Virtualization is built into the DNA of LinuxONE• PR/SM™ manages and virtualizes all the installed and enabled system resources as a single large SMP system• Full sharing/partitioning of the installed resources with the highest levels of efficiency and utilization• Scale up or scale out on demand with support for up to 85 partitions on Emperor II or up to 40 on Rockhopper II• IBM Dynamic Partition Manager is PR/SM with a simplified provisioning and management experience
– New in DPM: FICON® ECKD™ support - Guided storage setup, provisioning and management• Assured workload isolation, designed to meet the highest EAL5+ security certification• Dynamic optimization and scalability
z/VM v6.4 and v7.1
• Enables extreme scalability, security and efficiency – Support for 2TB of memory and improved z/VM paging, enabling workload consolidation, growth in memory-intensive applications, and superior levels of elasticity
• Operational improvements by providing Guest Large Page Support and Guest TX (Transactional eXecution) support• Increase efficiency and reduce complexity – Reliability, availability and serviceability enhancements remove the
need for a SAN Volume Controller, allowing low-end devices such as Storwize, SAN Volume Controller, V7000, V840 and V9000 to be directly connected to z/VM host use
• HW Security features – Pervasive encryption support: encrypting data moving between active memory and z/VM paging volumes. Enable guests to improve application program security by exploiting LinuxONE encryption features
• HW Performance features – Allowing guests to improve the performance of Java garbage-collection processes• Introducing z/VM 7.1 – Delivering a new release cadence that improves z/VM’s Continuous Delivery (CD) model.
Includes Single System Image (SSI) support in the base product, improving availability of z/VM systems by allowing clients to schedule outages without disrupting business critical applications
KVM running on IBM LinuxONE
Supports up to 8TB memory per guest for greater processing scale and performance Support new analytics workloads with Single Instruction Multiple Data (SIMD) for competitive advantage Improved performance for guests, Java workloads, and cryptographic processing Secure and protected business data with crypto exploitation Technology developed by IBM ... product offered by Linux Distribution partners
Trust, Growth and Digital Transformation with LinuxONE
A scalable data platform for energizing customer
experiences
Centrally manage data at scale without sharding to enable high performance
A single version of the truth delivered to services across the business
Leverage data in place to drive insight and value within the current transaction
Pervasive encryption and workload isolation at the core of
a robust cyber security
Stringent user identification, access control and Protection against malware
Encrypt data by policy without application change and at low cost
Enforced isolation to protect against internal and external threats
A robust cloud platform in a multi-cloud infrastructure
A develop anywhere, deploy anywhere approach for new services
High performance to meet service levels even at peak transaction rate
High availability and disaster recovery solutions for continuous availability
With smaller I/O configurations – a new 16U Reserved feature code can be added that tags 16U of space in rack as “Available”
This creates a new opportunity to build a fully customized, comprehensive solution to meet your infrastructure needs
Populate with your choice of server, switch or storage elements 1
1 Hardware service to IBM or non IBM options may be provided by a 3rd party – but they must be able to provide analysis of power, thermal, air flow and other requirements as listed in the IMPP guide. Requirements for physical structures as well as interactions with the ‘mainframe server’ will be provided
Topics
Rockhopper II 19” Rack configurations with (non-removeable) 16U-Reserved Feature ( FC 0617 )
To stay within warranty boundaries of the Rockhopper II, the externally provided hardware content being considered for installation shall follow these requirements (additional detail available with feature documentation):
• Fit within 19” rail-to-rail width, and 28 1/4” (717 mm) front-to-rear depth.
• Shall employ front to rear airflow.
• No more than 45 lbs. / 20.4 Kg) weight per EIA location. (e.g. a 4U drawer can weigh no more than 180 lbs. 81.6 Kg.)
- FC 0617 will include requirement for ballast to be added in the rack base (both, new build and during MES with 0 or 1 I/O drawers)
• Components to be installed must show safety certification labels required for server components in the country where the system is used (e.g. UL, TUC)
• New components should be populate from lowest open EIA and move up to avoid tipping hazards.
• No water/liquid based cooling systems/components to be added, as thermal/humidity monitoring is active.
• FC 0617 must be ordered to reserve the space for planning and stay under warranty.
Consolidate file & object workloadsFaster data analysisGlobal sharing
DS888xF
Business Critical
z/OS / AIXPower HAPower i HA
Business critical, deepest integration with z SystemsSuperior performance and reliabilityThree-site / Four-site replicationDS8888F, DS8886F, DS8884F, DS8882F
IBM FlashCore™ Technology OptimizedFlashSystem 900 Application acceleration • Extreme performance
• Hardware Compression• Targeting database
acceleration
Simplified managementFlexible consumption model
Large Grid scaleFull time data reduction
Enhanced data storage functions,economics and flexibility withsophisticated virtualization
StorwizeV5100/F
Scale-out clusteringSimplified managementFlexible consumption modelVirtualized, flash-optimized, modular storageEnterprise heterogeneous data services and selectable data reduction
TS1150, TS4500 FICON, FCP z/VM • Support for z/VM via FICON if TS7700 is front end
1 GDPS support for HyperSwap via the GDPS Virtual Appliance for Linux running under z/VM or KVM (GDPS appliance not available for native Linux on LinuxONE)
IBM LinuxONE and Storage – Key things to remember
19
Designing, developing and testing together is key to unlocking true value
• Direct Attach between the Storage and LinuxONE is only supported
with FICON• This means DS888x systems connected via FICON would be
the primary option if Direct Attach storage is required
• ANY storage being connected via FCP will require a Storage
Attached Network (SAN)• This requirement is dictated by N_Port ID Virtualization (NPIV)
• This SAN can be switches integrated into the 16U of space or
existing customer switches
• There is a broad range of switches with varying port counts to meet attachment requirements
• SAN attached coupled with any Spectrum Virtualize product
allows for the possibility to virtualize existing customer storage products for re-use
IBM LinuxONE and Storage – NPIV
20
Designing, developing and testing together is key to unlocking true value
• NPIV is the ability to have multiple WorldWide Port Names(WWPNs)
‘login’ to the SAN fabric sharing a single, physical Fibre Channel Port ID
• On Storwize products NPIV is also referred to as “Target Port Mode”
• Target Port Mode enables additional NPIV WWPNs on the
Storwize controller Ports• These additional WWPNs are used to split out traffic for host
usage and system uses, including virtualization of external
storage arrays
• Care must be taken to use the correct WWPN when zoning to ensure the NPIV target WWPN on the Storwize is utilized for
host connectivity. If this is not done then the host will not be
able to see the storage.• NPIV is also used on the LinuxONE to allow for multiple LPARs and
guests to share a single FICON adapter running in FCP mode• Each virtual device set will get it’s own NPIV WWPN
