© 2012 IBM Corporation
DTU – Distributed Systems
IBM PureFlex SmartCloud Entry Solution Technical Walkthrough
Lyngby
April 18
th, 2013
Ib F. Kristensen, ibfk@dk.ibm.com
IT Architect
Nordic PureFlex Product Manager
Agenda
IBM PureFlex SmartCloud Entry Solution
IBM
Concept overview of the system
Design issues/requirement for the Flex Chassis The physics of the Flex Chassis
The networking in the Flex Chassis
The Compute power in the Flex Chassis Management of the hardware
The Cloud software stack
Wrap up
IBM
IBM Corporation
* source: ibm.com/investor/4q12/presentation/4q12.pdf
Key figures
Headquarters
Armonk, New York, USA CEO
Virginia M. Rometty Employees 2012 434,246
Revenue 2012
104.5 billion US-Dollar Homepage
ibm.com
With revenues of 104.5 billion US-Dollar in 2012, IBM is one of the world’s largest provider of information
technology (hardware, software and services) and B2B solutions. The company has more than 430,000
employees worldwide and operates in more than 170 countries.
For more financial information please visit ibm.com/investor/financials/index.phtml
IBM revenue by industry* (in billion USD)
IBM revenue
Revenue in million USD 2012 2011 2010
Global Services
Global Technology Services Global Business Services
58,802 40,236 18,566
60,163 40,879 19,284
56,424 38,201 18,223
Hardware 17,667 18,985 17,973
Software 25,448 24,944 22,485
Global Financing 2,013 2,102 2,238
Other 577 722 750
Total revenue 104,507 106,916 99,870
Other: 5.9 Small and medium
business: 24.2
Communications: 9.5
Distribution: 9.2
Industrial: 9.7
Financial services: 30.7
Public: 15.3 Total revenue
104.5 billion USD
5
Introduction to the “New” IBM
Concept overview of the system
What is Cloud Computing?
A user experience and a business model
Rapidly provisioned standardized offerings Usage based pricing
An infrastructure management and services delivery methodology
Virtualized resources with elastic scaling
Monitor & Manage Services & Resources
Cloud
Administrator
Datacenter Infrastructure
Service Catalog, Component Library
Service Consumers
Component Vendors/
Software Publishers
Publish & Update Components, Service Templates
IT Cloud
Access Services
The IBM CCRA (Cloud Computing Reference Architecture)
defines 3 different roles: the Cloud Service Consumer, the Cloud Service Provider and the Cloud Service Creator.
Governance
Security, Resiliency, Performance & Consumability
Cloud Service Creator Cloud Service
Consumer Cloud Service Provider
Common Cloud
Management Platform (CCMP)
Operational Support Services (OSS) Cloud Services
Inf rastructure-as-a-Service Platf orm-as-a-Service
Sof tware-as-a-Service Business-Process-
as-a-Service
Business Support Services (BSS) Cloud
Service Integration
Tools
Consumer In-house IT
Service Creation Tools
Inf rastructure Existing & 3rdparty
services, Partner Ecosystems
Infrastructure-as-a-Service Platform-as-a-Service Application-as-a-Service
Servers Networking Storage
Middleware
Collaboration
Financials
CRM/ERP/HR Industry
Applications
Data Center Fabric Shared virtualized, dynamic provisioning
Database
Web 2.0 Application Runtime
Java Runtime Development
Tooling
Four major categories of Cloud Computing services are
emerging
ExamplesBusiness Process-as-a-Service
Employee Benefits Mgmt.
Industry-specific Processes
Procurement
Business Travel
The IBM CCRA (Cloud Computing Reference Architecture)
defines 3 different roles: the Cloud Service Consumer, the Cloud Service Provider and the Cloud Service Creator.
Governance
Security, Resiliency, Performance & Consumability
Cloud Service Creator Cloud Service
Consumer Cloud Service Provider
Common Cloud
Management Platform (CCMP)
Operational Support Services (OSS) Cloud Services
Inf rastructure-as-a-Service Platf orm-as-a-Service
Sof tware-as-a-Service Business-Process-
as-a-Service
Business Support Services (BSS) Cloud
Service Integration
Tools
Consumer In-house IT
Service Creation Tools
Inf rastructure Existing & 3rdparty
services, Partner Ecosystems
High-level view of IBM SmartCloud Entry for software cloud solution
IBM Flex System Enterprise Chassis
Support technology advancements through the next several years – Future processors, memory, I/O, storage, specialty hardware – Faster internal bus speeds
– Faster I/O speeds
Energy efficient cooling and power system – High efficiency components
• Sophisticated control systems
• Compatible with future, high efficiency data center infrastructure – New Compute and Storage nodes (Leadership Intel Performance) – New scalable switch elements and IO adapters
– New integrated management node and chassis management elements
• Ability to manage multiple chassis
Design issues/requierment for
the Flex Chassis
Some of the design issues/requirement for the Flex System
Based on standard architecture and components (x86, Power, FC, ethernet...) Lifetime of appx. 10 years
Low cost Simplicity Flexibility Integration
Reliability, redundancy upgradeability
Infrastructure Application Platform Data Platform
Delivering Cloud
Infrastructure Services Delivering Cloud Application Platform
Services
Delivering Big Data Platform Services
New Analyt
ics
Model
The physics of the Flex Chassis
2 nodes each with 2 CPU sockets, 24 DIMMs and 2 Hot- Swap Hard Drives 2 node Fillers
2 node Fillers
2 nodes without Hot- swap HDDs
Full Width node
Full Width node with up to 4 processor sockets, 48 DIMMs and Hot- Swap HDDs
2 CPU node with PCIe adaptor expansion node
Bay 1
Bay 14
Chassis Front View and Compute Node examples
Power Supplies (6X) IO Modules (4X) 40 mm Fans (2X)
CMM (2X) 80 mm Fan
Packs (8X)
Fan Mux Card (2X)
Chassis Rear View
4 2
3 1
1 6
1
10
Node Bay Shelf
Remove for Full Wide node Remove Two Shelves for Full Wide, Double High node
Chassis Shuttle
Chassis Bay Shelf Removal
Flex System Power – Overview
Flex chassis is optimized for 3-phase, 60A, 200-208VAC (US/Japan)
– Efficient use of 3-phase, 32A 220-240VAC / phase (International) Chassis contains up to six 2505W power supplies
– Supports both N+N and N+1 power redundancy
– Efficient balancing of N+N power across all 3 phases
– Power supply unit (PSU) is sized to efficiently use PDU power High Efficiency
– PSU certified to 80 PLUS Platinum
– Low power loss mid-plane and connectors
Cooling Paths – Zoned Cooling Between Subsystems
Switch Cooling
Node Cooling Power Supply Cooling
Zoned Cooling:
• Nodes
• Left / Right
•
•Subsystems
• Node, Switch, and PSU Modules
The networking in the Chassis
I/O Adapter
1 I/O
Adapter 2
Node I/O: LAN-on-Motherboard (LoM) and I/O Adapter slots
x86 nodes have models with LoM and models without LoM, Power nodes only have models without LoM
CMM
L2 Switch
Flex System Manager Bay 1
IMM2
Eth LOM
L2 Switch
Eth
Compute Node x86 Bay 2
IMM2
CMM
L2 Switch Adapter 1 Adapter 2
Compute Node POWER Bay 14
FSP Adapter 1 Adapter 2
Switch Bay 1
Switch Bay 2
Switch Bay 3
Switch Bay 4
Management Network Data Network
I/O Bandwidth: “Future-proofed” mid-plane capability
CMM
L2 Switch
Flex System Manager Bay 1
IMM2
Eth LOM
L2 Switch
Eth
Compute Node x86 Bay 2
IMM2
Management Network Data Network
CMM
L2 Switch Adapter 1 Adapter 2
Compute Node POWER Bay 14
FSP Adapter1 Adapter 2
Switch Bay 1
Switch Bay 2
Switch Bay 3
Switch Bay 4
In summary … a very capable multi-layered mid-plane design
S y s te m i n fr a s tr u c tu re
Flexible networking solution, allowing for best price/performance
IBM 10Gb Switch: Wired for up to 16 10Gb ports per node and twenty two external ports
Networking
NIC
Mezz card-1
Mezz card-2
Midplane CPU-Node
Up to 4 KR Ports Per Switch Bay
NIC
Scalable Switches 5,6
Scalable Switches 7,8 Scalable Switches 3,4 Scalable Switches 1,2
4-port Mezz
4-port Mezz
Uplink Ports
ScS – Bay 1
ScS – Bay 2
(A) (B)
(A) (B)
(A) (B)
(A) (B)
2 going to 4 per physical bay
8 going to 16 ScS – Bay 3
ScS – Bay 4
NIC
Ethernet & FCoE Fibre Channel InfiniBand – 52 port 1Gb Switch
• Base: 14/10 (internal/external)
• Upgrade: 14/10
• Upgrade: four 10Gb uplinks
– 64 port 10Gb Ethernet Switch
• Base: 14/10
• Upgrade: 14/8 (two 40Gb uplink)
• Upgrade: 14/4 – 1/10Gb Pass-Thru
– 20 port 8Gb (Qlogic) – 20 port 8Gb Pass-Thru
(Qlogic)
– 12 port 16Gb (Brocade) – 24 port 16Gb w/ ESB
(Brocade)
QDR Switch (Mellanox) – Upgrade: FDR
– 4 port 1Gb (Broadcom) – 4 port 10Gb (Emulex) – 2 port 10Gb (Mellanox)
– 2 port 8Gb (Qlogic) – 2 port 8Gb (Emulex) – 2 port 16Gb (Brocade)
QDR & FDR Adapter (Mellanox)
S w it c h A d a p te r
FlexChassis I/O Portfolio view
IBM 10Gb Scalable Switch for IBM Flex System Chassis
One, two, or three 10G ports per server – selectable by software license
Base Switch: 14 x 10Gb server port and 10 x 10Gb uplinks Switch upgrade 1: 28 x 10Gb server ports and 16 x 10Gb uplinks. 40Gb uplinks
enabled.
Switch upgrade 2: 42 x 10Gb server ports and 22 x 10Gb uplinks. 40Gb uplinks
enabled.
1.28 Tbps – first 1Tbps+
blade switch
5+ Tbps per chassis
IBM Flex System 10Gb Virtual Fabric Scalable Switch
Two 40 G uplink ports. Each port can also be converted to 4*10G
using QSFP to SFP+ cable
Full featured, Scalable bandwidth bringing convergence and simplicity to Datacenter
applications
The Compute power in the
Chassis
S y s te m i n fr a s tr u c tu re
Compute
Diverse offerings to match the diverse workloads
System Portfolio tuned to workloads
◊
Reduce acquisition costs through virtualization
consolidation
◊
Maximum platform capability provides deployment
flexibility
IBM Flex System x220
IBM Flex System p260
IBM Flex System p24L IBM Flex System x240
IBM Flex System p460 IBM Flex System
x440 Compute Node
IBM Flex System PEN
(PCIe Expansion Node)
• Standard Compute Node Form Factor
• 2-Socket Sandy Bridge-EP (Xeon E5-2600 Series Processor) 135W, 130W, 115W, 95W, 80W, 70W, 60W (4 – 8 core)
• 24 LP DDR3 DIMMs / Up to 1600MHz
• Up to 384GB Memory capacity with 16GB RDIMMs
• Up to 768GB Memory capacity with 32GB LR-DIMMs
• Chipset – Patsburg B (Intel C600 Series)
• 2x2.5” Hotswap SAS/SATA/HDD/SSD
• Onboard ServeRAID H1135 SAS Controller (RAID 0, 1) 6Gbs + Optional ServeRAID M5115 (RAID 0,1,5,6,10,50) with Flash-to- cache, supports eight 1.8inch SSDs
• SAS or SATA drives
• LSI 2004 connects to Patsburg with x4 PCIe Gen2
• Emulex BE3 10Gb Dual Ethernet LOM using Periscope connector
• LOM supports iSCSI and FCoE with FoD (feature On Demand) option
• Models with LOM and LOM-less
• Up to 2 Fabric Mezzanine Cards (PCIe Gen3 – 8GTs capable)
• Each Mezz connector: 1x16 PCIe port and 1x8 PCIe port
• Installing Mezz 1 card requires removal of LOM Periscope connector
• uEFI / IMMv2 / TPM 1.2 Rev 1.03
• Power Management
• xSmartEnergy Control
• Capping w/Pstate and SCI, Sys Power Maximizer
• Embedded Hypervisor
• ESXi on Flash key option (2 USB Keys for redundant boot option supported on unique DIMM air baffle)
• Front panel – one USB connector and one dongle
• Dongle includes two USB, video, and one serial port
• Front panel connector does not have enough power to run external USB drives
• Management
• iMM V2 Management Controller
• RTMM for Power Exec and Power Sequencing
x240 Compute Node Overview
PCIe x1 LPC
USB
iMM v2
QPI Links
DDR3 DIMMS
Intel SandyBridge
EP Processor 2
DDR3 Bus G DDR3 Bus H DDR3 Bus E DDR3 Bus F
DDR3 DIMMS DDR3 DIMMS DDR3 DIMMS DDR3 DIMMS
DDR3 Bus C DDR3 Bus D DDR3 Bus A DDR3 Bus B DDR3 DIMMS
DDR3 DIMMS DDR3 DIMMS
NC (8 PCIe lanes + unused x4 ESI) x4 ESI
Cntl, Misc
HS HDD/SSDs
Patsburg SSB
(C600 Series)
LSI 2004 SAS
Management to Midplane
x16 PCIe Gen3 x16 PCIe Gen3
Intel SandyBridge
EP Processor 1
FRM or PME/SME Interface
ETH Out ETH Out PCIe x2
Internal USB
USB USB(2)
Dual KR to Midplane x4 PCIe
Gen2
x8 PCIe Gen3
Mezz 0
10Gb LOM
ETH In ETH In
Video (RGB)
Front USB
Front DongleKVM
USB(2)
Mezz1 Mezz2
PHY PHY Serial Port
USB
QPI 1.1 – 8GT/s DDR3 Ch – up to 1600MT/s
PCIe Gen3 – 8GT/s
SPI
TPM1.2
System BIOS 16MB
SPI Parallel
Exp.
NAND Flash 2GB pbDSA iBMC Flash
8MbitBoot
RTMM Power Seq.
Power Exec
DDR3 Video 256MB
I2C
I2C LPC LPC
TPM IMM
x8 PCIe Gen2
QPI – 8.0GT/s
16GB/s/dir
16GB/s/dir 8GB/s/dir
4GB/s/dir
DDR3 – 1600MT/s/Ch DDR3 – 1600MT/s/Ch
Pop/No-Pop for FRM
X240 Compute Node Block Diagram
Flex x86 Node Relative Performance
S y s te m i n fr a s tr u c tu re
Storage Node Design Overview
Storage
In Chassis Enterprise Storage Double high / double wide node Dual hot swappable controllers for HA Shared storage for multiple CPU nodes High performance/ high function storage
Midrange class performance FCoE & iSCSI block storage
FC block storage as well NAS file storage (future)
24 HDDs/SSDs – 2.5” hot swap disk trays Via dual SAS paths to both controllers Add’l storage via Storage Disk Expansion Automated FW for hot spot data migration
Migrate content between SSDs and HDDs
Chassis and Management Appliance Integration Tight integration with chassis mgmt
Tight integration for storage and fabric mgmt VM provisioning, copy svcs, mirroring, etc
34
Enclosure Internal layout
Full Wide
24, 2.5” HDD Bays Controller
Full Wide
Double High
24, 2.5” HDD Bays Controller
Node Canister
Enclosure
Hard Disk Bay
HIC card
Management of the hardware
S y s te m i n fr a s tr u c tu re
Management
Simplified Virtualization Management experience
P h y s ic a l V ir tu a l W o rk lo a d
Physical Consolidation
& Setup Management Integration Resource Utilization Resource Pooling Intelligent Automation
Choice of Hypervisors
Automated virtual machine placement Dynamic allocation of virtual server, storage and network resources
Network QoS
Virtual workload definition
Placement services and advisors
Pooling of all network switch resources to enable consistent network policy application
Monitors VM network utilization within the fabric Virtual Machine relocation for compute, storage and networking
Monitors the Network Fabric for congestion
FSM Flex System Explorer
Dashbar Explorer Hardware View
Chassis only, otherwise Explorer shows a table
Rich Menus
Finder Status Pods
Framework Page
The Cloud software stack
IBM SmartCloud Entry solution architecture, concept
Power Systems Reference Configurations
IBM Systems Director 6.2
PowerVM ESXi
VMControl
2.3.1 VMware
vCenter VMControl API vCenter API
Cloud administration and management
Tivoli Provisioning
Manager for Images 7.1.1
REST API Self-service UI
IBM SmartCloud Entry management stack
Customer integration
• Approval policy, project management, users and roles
• Events and auditing
• Metering
• Image library
Common functionality Power Systems only System x only
IBM Systems Director 6.2
System x
Reference Configurations
IBM SmartCloud Entry solution architecture, concept
Power Systems Reference Configurations
IBM Systems Director 6.2
PowerVM ESXi
VMControl
2.3.1 VMware
vCenter VMControl API vCenter API
Cloud administration and management
Tivoli Provisioning
Manager for Images 7.1.1
REST API Self-service UI
IBM SmartCloud Entry management stack
Customer integration
• Approval policy, project management, users and roles
• Events and auditing
• Metering
• Image library
Common functionality Power Systems only System x only
IBM Systems Director 6.2
System x
Reference Configurations
Network sample
IBM SmartCloud Entry solution architecture, concept
Power Systems Reference Configurations
IBM Systems Director 6.2
PowerVM ESXi
VMControl
2.3.1 VMware
vCenter VMControl API vCenter API
Cloud administration and management
Tivoli Provisioning
Manager for Images 7.1.1
REST API Self-service UI
IBM SmartCloud Entry management stack
Customer integration
• Approval policy, project management, users and roles
• Events and auditing
• Metering
• Image library
Common functionality Power Systems only System x only
IBM Systems Director 6.2
System x
Reference Configurations
vCenter API
The Vix API is object-based. Most API functions either create objects or operate on the properties of existing objects.
Client applications reference Vix objects with handles. Handles are opaque identifiers (actually integers) that can be passed as parameters to functions. Handles are run-time only and are unique only within a client's address space.
Most functions in the C-language API take a handle as a parameter. Because a handle value represents an object to the API, this document uses the terms "handle" and "object"
interchangeably.
There are several handle types, but a few of the key types are:
– Virtual Machine -- A single virtual machine, which might or might not be powered on.
– Host -- A single host computer, either the local host or a remote host.
– Job -- An object used in managing asynchronous operations.
– Snapshot -- A snapshot of a virtual machine.
IBM SmartCloud Entry solution architecture,concept
Power Systems Reference Configurations
IBM Systems Director 6.2
PowerVM ESXi
VMControl
2.3.1 VMware
vCenter VMControl API vCenter API
Cloud administration and management
Tivoli Provisioning
Manager for Images 7.1.1
REST API Self-service UI
IBM SmartCloud Entry management stack
Customer integration
• Approval policy, project management, users and roles
• Events and auditing
• Metering
• Image library
Common functionality Power Systems only System x only
IBM Systems Director 6.2
System x
Reference Configurations
REST API reference
IBM SmartCloud Entry provides the following services for appliance libraries.
GET /appliances: Retrieve appliances from the cloud.
GET /appliances/{id}: Retrieve specific appliance from the cloud.
PUT /appliances/{id}: Update specific appliance in the cloud.
DELETE /appliances/{id}: Delete a captured appliance that has failed.
GET /appliances/{id}/targets: Retrieve the targets that can handle a workload of this appliance.
GET /appliances/{id}/customization: Retrieve default customization available when deploying this
appliance.
PUT /appliances/{id}/customization: Update the default customization for this appliance (admin
service).
DELETE /appliances/{id}/customization: Resets the customization for this appliance.
GET /appliances/{id}/log: Retrieve a appliance's capture progress logs.
POST /appliances: Create a new appliance by taking a workload capture.
The representational state transfer (REST) application programming interface (API) is
provided by SmartCloud Entry.
IBM SmartCloud Entry solution architecture, concept
Power Systems Reference Configurations
IBM Systems Director 6.2
PowerVM ESXi
VMControl
2.3.1 VMware
vCenter VMControl API vCenter API
Cloud administration and management
Tivoli Provisioning
Manager for Images 7.1.1
REST API Self-service UI
IBM SmartCloud Entry management stack
Customer integration
• Approval policy, project management, users and roles
• Events and auditing
• Metering
• Image library
Common functionality Power Systems only System x only
IBM Systems Director 6.2
System x
Reference Configurations
49
IBM SmartCloud Entry: Comprehensive cloud capabilities
Administrator Functions
• Configure workload to host
• Configure workload targets to system pool
• Configure Virtual Appliance Parameters
• Register Virtual Appliance for User Selection
• Approve/Reject New Workload Requests
• Approve/Reject Workload Resize Requests
• Configure Billing
• Charging accounts, account assignment
• Configure approvals
• Configure to generate metering records
• Create & Manage Projects
• Add users to projects as admin, deployer, viewer
• Create Users
• Configure Network Pools
• Configure LDAP environment
• Cloud Configuration to VMControl or VMware
• Manual Intervention
• Review event logs & failures
• Initiate workload movement (between projects or via virtualization manager)
End User Functions
• Request access to a project
• Request CFS access from login panel
• Request workload creation – appliance deploy
• Set user instance parameters (CPU, memory)
• Resize Running Workloads
• Delete a Workload
• Clone a workload
• Start/Stop a workload
• Review workload properties
• Add additional disk (VMControl)
• Set target disk size at deploy
(VMware)
IBM SmartCloud Entry delivers cloud experience to users
Easy to access, easy to use Service Request Catalog
Hides underlying infrastructure from user and shifts focus to services delivered Enables the ability to provide standardized and lower cost services
Facilitates a granular level of services metering and billing
Workload standardization eases complexity
The IBM SmartCloud Entry end-user self-service scenario
End
Users Service Portal
Service Request Catalog
Provisioning Engine Workflows
Expert Systems Scripts
Service Modules Metering/Usage Billing
Approvals
Virtualized Cloud Infrastructure
Easy to access, easy to use Service Request Catalog Hides underlying complex infrastructure from user and shifts focus to services provided
Enables the ability to provide standardized and lower cost services
Facilitates granular level of services metering and billing
Workload standardization eases complexity
Self-service Web UI Basic usage metering
IBM SmartCloud Entry delivers cloud benefits to managers
Wrap up
IBM SmartCloud Entry solution architecture, concept
Power Systems Reference Configurations
IBM Systems Director 6.2
PowerVM ESXi
VMControl
2.3.1 VMware
vCenter VMControl API vCenter API
Cloud administration and management
Tivoli Provisioning
Manager for Images 7.1.1
REST API Self-service UI
IBM SmartCloud Entry management stack
Customer integration
• Approval policy, project management, users and roles
• Events and auditing
• Metering
• Image library
Common functionality Power Systems only System x only
IBM Systems Director 6.2
System x
Reference Configurations
ISV and IBM Patterns
SmartCloud Provisioning
Compute Network
Storage
SmartCloud Provisioning
SmartCloud Entry
Compute Network
Storage
SmartCloud Provisioning
SmartCloud Entry
Pre-NGP or Non IBM HW
Flex System PureFlex PureApplication
Compute Network
Storage
Compute Network
Storage
Pre-integrated entry point Upgrade opportunity
Platform and Infrastructure Management SmartCloud
Orchestration
SmartCloud Orchestration ISV and IBM
Patterns
ISV and IBM Patterns
ISV and IBM Patterns ISDM or
SmartCloud Orchestration
Orchestrates
Foundation
Patterns
Links
IBM SmartCloud Entry Documentation Representational state transfer - REST
Redbook - Implementing SmartCloud Entry on IBM PureFlex System
Redbook - ReadyPack for Cloud with Hyper-V on IBM Flex System
QUESTIONS ?
Ib F. Kristensen
IT Architect
Nordic PureFlex Product Manager Systems & Technology Group
Bytoften 1 8240 Risskov +45 2880 6217 ibfk@dk.ibm.com