Infrastructure Future Proof Data Platformevents.ibs.bg/events/compass2013.nsf/vPagesLookup...IBM...
Transcript of Infrastructure Future Proof Data Platformevents.ibs.bg/events/compass2013.nsf/vPagesLookup...IBM...
© 2013 IBM Corporation 1
Dave Ridley
Technical Product Manager
IBM PureFlex and Flex System, UK & Ireland
Mikhail Saykov
Power Brand Client Rep.
IBM Bulgaria
November 2013
&
© 2013 IBM Corporation 2
Infrastructure
X86 & Power Compute
Virtualised Storage
Optional Management
Fabric
Future Proof Chassis
Data Platform
Delivers high performance data services to transactional and analytics applications
Runs your choice of operating systems, hypervisors, applications and middleware
Application Platform
Built on IBM middleware to accelerate deployment of your choice of applications
Data Platform
Delivers high performance data services to transactional and analytics applications
System for Transactions
System for Analytics
OLTP Data Warehouse Big Data Tools
System for Hadoop
© 2013 IBM Corporation 3
Flex System Design Goals
BladeCenter •Density Focus
•7U Tall with 14 Blades (twice density of 1U Servers
•Network Consolidation
BladeCenter H Density Focus
9U Tall with 14 Blades
Network Consolidation
4 High Speed I/O channels per Blade
•10U tall with 14 Nodes larger Nodes
•Each slot supports 2 High Powered CPU’s + DIMMs
•Improved Efficiency – larger node reduces airflow, lower
resistance for lower fan power
•Power and Cooling for future growth
•Range of higher bandwidth and low latency fabrics
Today’s world requires much more performance, much more memory,
much more bandwidth and is increasingly virtualised….
© 2013 IBM Corporation 4
Infrastructure Components
• Full chipset CPU and memory support
• Future-proofed networking bandwidth
• Flexible compute node storage options
• Exceptional in-chassis network latency
• Chassis
• Compute Nodes
• Storage
• Networking
• Flex System Manager
• Expansion
Open choice of architectures, hypervisors, environments –
designed for multiple generations of technology
© 2013 IBM Corporation 5
Compute Node Choices: Leading edge compute technologies deliver an open architecture, operating system and hypervisor choice
IBM Flex System Advantages: Flexibility and Openness
IBM Flex System p260 >16 Cores
IBM Flex System x220
>16 cores
IBM Flex System x240 >24Core IVB
IBM Flex System x440 >32 Cores
IBM Flex System PCI expansion node
IBM Flex System p460
>32 Core
IBM Flex System Storage expansion node
IBM Flex System p270
IBM Flex System x222
>2x 16Core
P7+ DCM
>24 Cores
P7+
P7+
© 2013 IBM Corporation 6
6
Database /
Consolidation
Enterprise
Performance
Business
Applications
(SAP, ERP,
Small DB)
Infrastructure
Applications
(file/ print/
collaboration)
Flex System: POWER compute node Positioning
p460
AIX/IBM i/Linux
4-sockets
16 to 32-cores
Flexible
configuration
Leadership
Virtualization
Demanding
Database
p24L
Linux only
2-sockets
8 to 16-cores
Optional
High-Value
features
Competitive
with x86
2 Socket 4 Socket
p460
High-performance, reliable, secure system that is cloud-enabled. Outstanding offering for mid-market clients Ideal for server consolidation or a high-performing database server
p270
Combines secure, reliable computing and energy efficient virtualization. Workload-optimizing capabilities to enable companies to get the most out of their systems by increasing utilization and performance while helping to reduce infrastructure and energy costs.
p260 Highly flexible node with large memory capacity, outstanding performance, industrial-strength virtualization and workload-optimizing capabilities. Ideal for small-to-midsize database servers, and consolidation of virtualized application servers
p24L Compute node which runs industry-standard Linux from Redhat or SUSE. Exploits the advanced hardware and software capabilities of POWER7 to provide high qualities of service for Linux workloads
p270
AIX/IBM i/Linux
2-sockets
24-cores only
Built-in
Dual VIOS
Leadership
Virtualization
Cloud Ready
p260
AIX/IBM i/Linux
2-sockets
4 to 16-cores
Flexible
configuration
Leadership
Virtualization
Application
Workloads
NEW models Aug 6th
POWER7 POWER7+ DCM POWER7+ SCM
© 2013 IBM Corporation 7
DIMMs
DIMMs
DIMMs
DIMMs
DIMMs
DIMMs
DIMMs
DIMMs Mezz 2
Mezz 1
POWER7Socket
IO
Hub
IO
Hub
POWER7Socket
Sy
ste
m i
nfr
as
tru
ctu
re
Compute
2 x SAS 2.5” HDD
or 2 x 1.8” SDD
drives
RAID Controller
Standard Width node
◊
2-socket POWER7+ processor
◊
Supports AIX, IBM i, Linux
◊
16 core : 2 Socket x 8 core Multiple Speeds 4.1 or 3.6GHz
◊
8 core : 2 Socket x 4 core New 4.0GHz speed
◊
16 DIMMs DDR3, 512GB Max ◊
Double the number of VM’s per
core
Flex System p260 compute node
Integration without compromise, designed for the next decade
Delivers over 30% greater performance with
same density and energy use as existing p260
nodes
Flex System p260 compute node – Power7 & Power7+
© 2013 IBM Corporation 8 8
Ideal for Highly Virtualized Application Environments IBM Flex System p270 DCM compute node
DIMMs
DIMMs
DIMMs
DIMMs
DIMMs
DIMMs
DIMMs
DIMMs Mezz 2
Mezz 1
POWER7+ Socket
IO
Hub
IO Hub
POWER7+ Socket
Sy
ste
m i
nfr
as
tru
ctu
re
Compute
2 x SAS 2.5” HDD or
2 x 1.8” SDD drives
Standard compute node
◊
2-socket POWER7+
◊
24 core : 2 Socket x 12
cores (Dual Chip Module)
◊
16 DIMMs 512GB Max
◊
Built-in support for Dual
VIOS (with optional adapter)
◊
Double the number of VM’s
per core
◊
IBM p270 compute node
Enables over 6,000 POWER Virtual Machines per chassis
© 2013 IBM Corporation 9 9
Ideal for demanding Database and Analytics Workloads Flex System p460 compute node – Featuring POWER 7+
DIMMs
DIMMs
DIMMs
DIMMs
DIMMs
DIMMs
DIMMs
DIMMs Mezz 2
Mezz 3
POWER7+ Socket
IO
Hub
IO
Hub
POWER7+ Socket
POWER7+ Socket
POWER7+ Socket
DIMMs
DIMMs
DIMMs
DIMMs
DIMMs
DIMMs
DIMMs
DIMMs Mezz 4
Mezz 1
IO
Hub
Sy
ste
m i
nfr
as
tru
ctu
re
Compute
*HDD or SSD – Mounted on cover (located over memory)
Flex System p460 compute node
Double Width compute node
◊
4-socket POWER7+
processor
◊
Supports AIX, IBM i, Linux
◊ 32core : 4 Socket x 8core
New 3.6 and 4.1GHz speed 16 core : 4 Socket x 4core
New 4.0GHz speed
◊
32 DIMMs 1TB Max
◊
Double the number of VM’s
per core
Delivers 39% better 32-core SPECint Rate performance versus the HP DL560 G811
© 2013 IBM Corporation 10 10
Database /
Consolidation
Enterprise
Performance
Business
Applications
(SAP, ERP,
Small DB)
Infrastructure
Applications
(file/ print/
collaboration)
Flex System: x86 Compute Nodes Positioning
x440
Price-
Performance
Optimized
4 socket
Mainstream
Database
High-End
Virtualization
Memory
Intense
x240
IvyBridge
Enhanced
Platform
Dual Proc
High-Density
Virtualization
Standard
High-Value
features
x220
Standard
Platform
Dual Proc.
Optional
High-Value
features
Energy
Efficient
2 Socket 4 Socket
x440
High performance compute node with leadership compute, memory and I/O capacity 48 DIMMs for memory intense Performance optimized with
leadership IO capacity
x240
High Density, high performance optimized for virtualization 24 DIMM slots Integrated HW RAID 0,1
& HS HDDs Performance optimized with
flexible IO capability
x222 Double dense computing and virtualization Optimized for solution cost Efficient infrastructure utilization Flexible I/O to integrate into
existing datacenters
x220
Versatile, easy to use, optimized for performance, power and cooling. Flexible design for Virtualized and Native workloads Great price/performance No compromise blade Cost-optimized platform
x222
Virtual Desktop
computing
Flexible I/O
Dense Cloud
Deployments
© 2013 IBM Corporation 11
Sy
ste
m i
nfr
as
tru
ctu
re
Compute
Standard Width compute
node
◊
2-socket Sandy Bridge-EN
◊
12 LP DDR3 DIMMs /
1333MHz / 1600MHz
◊
1Gb base with expansion
slots
◊
2 hot swap 2.5” SAS/SATA
SSDs or HDDs
IBM Flex System x220
Entry cost optimized compute, designed for
energy-efficiency, ideal for native and point
application workloads
2x Hot Swap, Small
Form Factor HDDs
12x LP DIMMs
2x IO Mezzanine
Cards
2x Intel E5 2400
Processors
x220 2S EN Compute Node
© 2013 IBM Corporation 12
Syste
m i
nfr
astr
uctu
re
Compute
Double the density:
28 servers per 10U chassis
◊
2-socket Xeon E5-2400 per
twin – each up to 8 Core
◊
12 LP DDR3 DIMMs /
1333MHz / 1600MHz per twin
◊
2x 10GbE ports standard per
twin. Optional
8Gb/16Gb Fibre Channel
or QDR/FDR Infiniband
◊
1x 2.5” SATA HDD
or 2x Hot swap 1.8” SSD per
twin
Increase server density & business performance:
• Double-dense design that can support 28 servers per
IBM® Flex System™ Enterprise Chassis
• More business and floating point operations per
chassis
• Over 150% increase in server density compared to a
std 1U rack server in 10U of rack space2
• 58% better compute throughput per chassis vs. HP
BL460c3
1. Based on IBM testing for medium task workers
1. x240 with two E5-2680 (2.7 GHz) CPUs can support 206 users, and a total of 2884 users (206 x14) in 14U chassis
2. x222 with E5-2470 (2.3 GHz) CPUs can support 174 users (or 348 per x222), and a total of 4872 users in a 14U chassis (14x2x174)
3. This represents an approximate 60% increase in VDI density with the new x222 compute node.
2. Based on double dense design compared to 1U rack optimized server design in a standard rack
3. Based estimate of SPEC_int performance comparing x222 configured with E52400 CPU and BL460 configured with E5-2600 procs and twice the number of servers per chassis with IBM offering
x222 provides clients with 60% increase in VDI user
density. Supports over 2800 users per chassis
compared to standard density blade offerings1
x222 Compute Node High Density EP based Node Optimized for Virtualization, Virtual Desktop Solutions, Cloud and Infrastructure Consolidation
© 2013 IBM Corporation 13
• Each Flex System x222 Compute Node has two independent “twin” servers
• Each twin server:
• Is a independent 2 socket system (Intel Xeon E5-2400 processors)
• Maximum number of memory slots or up to 384GB memory per x222
• Two 10GbE network connections enabled for FCoE w/ FoD key
• Support for one 2.5” SATA HDD/SSD or two 1.8” SATA SSDs
Note: The twin servers in the x222 Compute Node must be identically configured
Understand the uncompromised density of Flex System x222
Server 1 with:
• 2 x E5-2400 processors
• 12 x DDR3 memory
• 2 x 10Gb networking ports
Server 2 with:
• 2 x E5-2400 processors
• 12 x DDR3 memory
• 2 x 10Gb networking ports IBM Flex System as shown fits in standard width bay
© 2013 IBM Corporation 14
IBM Flex System x222 Highest Compute Density for VDI, Virtualization, Cloud and Infrastructure
consolidation
Standard width node, with 2 servers for dense computing
Twin board 2-Socket Intel Xeon E3 (Sandy Bridge-EN)
12 LP DDR3 DIMMs / 1333MHz per board
10Gb base with option to add additional adapter
1 non-hot swap 2.5” SAS/SATA HDDs or 2 hot swap 1.8”
SSDs
12x2 DIMMS 2x2 E5 2400 Proc
1x Non-Hot Swap, 2.5” HDD or 2x Hot Swap 1.8” SSD
Unique mezz card
© 2013 IBM Corporation 15
Sy
ste
m i
nfr
as
tru
ctu
re
Compute IBM Flex System x240
2x Hot Swap, Small
Form Factor HDDs
24 LP
DIMMs
2x Intel Xeon E5 2600
V2 Processors
2x IO Adapter Cards
Keyboard, Mouse,
Video Dongle
connector
Release latch
Standard Width compute node
◊
2-socket Ivy Bridge to 130W &
Sandy Bridge-EP to 135W
◊
Up to 12 Core IVB
◊
24 LP DDR3 DIMMs
◊
Up to 1866MHz Memory IVB
◊
10Gb LOM or 10Gb I/O Adapter
or LOMless
◊
2 hot swap 2.5” SAS/SATA SSDs
or HDDs
USB for ex: ESXi
Uncompromised Compute, IO, and Storage
performance, designed for mainstream
virtualization, and a broad range of workloads
x240 2 Socket Xeon EP Compute Node
© 2013 IBM Corporation 16
Sy
ste
m i
nfr
as
tru
ctu
re
Compute
Delivers maximum CPU, memory, and IO
performance, ideal for database and large VM
deployment
IBM Flex System x440
Full Width compute node
>32 Cores
◊
4-socket Xeon EP E5
◊
48 LP DDR3 DIMMs / 1.5TB
◊
2 x 10GbE LOM
◊
2 hot swap 2.5” SAS/SATA
SSDs or HDDs
2x Hot Swap, Small
Form Factor HDDs
48x LP DIMMs
4x IO Adapters
4x Intel E5 4600
Processors
x440 4 Socket Xeon EP Compute Node
© 2013 IBM Corporation 17
Compute Node
PEN Attached to
Compute Node
Interposer
Cable
PCIe Bridge
Chip
PCIe Gen2
Low Profile
Slots (x2)
PCIe Gen2 Full
Width, Full Height
Slots (x2)
2 Extra
Mezz slots
x86 Compute Node with IBM Flex System PCIe Expansion Node
© 2013 IBM Corporation 18
GPU Support today…
Supported in a PCI Expansion Node
TODAY!
© 2013 IBM Corporation 19
IBM Flex System Storage Expansion Node Increase Storage Capacity or create an entry Network Attach Storage (NAS)
Dedicated Storage Expansion For Flex System 2S x86 Compute Nodes
Increase storage in following workloads:
• Network Attach Storage (NAS)
• Video Security/Surveillance
• Transactional Data Server
Features
• Attaches to standard-width compute nodes
• Up to 12 SAS/SATA or SSD
• Enables Flex System migration platform for
competitive 2U servers
• Provides a competitive alternative to HP’s
StorageWorks D2200sb Storage Blade
• Supports SSD’s, up to 1.6TB each
IBM Flex System x240 With
Storage Expansion Node
Up to
19.2 TB of storage
Faster recovery
with
Hot-Swap
disk capability
© 2013 IBM Corporation 20
IBM Flex System Advantages: Flexibility and Openness
“Tomorrow ready” switches provide the flexibility to upgrade without replacement via Feature on Demand
Available
40Gb Uplinks
Up to
220Gb uplink BW and <1microsec
latency
10Gb Pass Thru 10Gb Switch with FCoE
1Gb Scalable switch
Industry’s first integrated
56Gb Infiniband switch
Flexible and open – Fiber Channel and Ethernet switching Flexible and Converged - FCoE 10/40Gb Ethernet uplinks (OmniPorts allow 10GbE or 4/8Gb FC choice) Flexible, future proof – 40Gb uplinks Flexible and simple – connect to existing infrastructure, no “rip and replace”
© 2013 IBM Corporation 21
Ethernet/Converged Fibre Channel InfiniBand
Pass
Thru
IBM
EN4091
1Gb Switch
IBM
EN2092
10Gb Switch
IBM
EN4093R
10Gb Converged
Switch
IBM
CN4093
8Gb SAN
Qlogic
FC3171
16Gb SAN
Brocade
FC5022
QDR 40Gb
FDR 56Gb Infiniband
Mellanox
IB6131
Network I/O Modules & Adapters:
8 port 10Gb (Emulex)
4 port 10Gb (Emulex)
2 port 10Gb(Broadcom)
4 port 1Gb (Broadcom)
2 port 10Gb (Mellanox)
2 port 40Gb (Mellanox)
2 port 8Gb (Qlogic)
2 port 8Gb (Emulex)
2 port 16Gb (Brocade)
2 port 16Gb (Emulex)
4 port 16Gb (Emulex)
QDR Adapter
FDR Adapter
(Mellanox)
10Gb Systems
Interconnect
IBM
SI4093
40Gb Switch
Mellanox
EN6131
10Gb
Switch Brocade
EN4023
Fabric
Extender Cisco
B22 FEX
© 2013 IBM Corporation 22
I/O
Mo
du
le
1
I/O
Mo
du
e
2
I/O
Mo
du
le
3
I/O
Mo
du
le
4
Chassis I/O Routing • I/O module bays are not wired to work with
only specific I/O modules
• The IO Module in an I/O module bay
determines what kind of I/O is routed out of
that bay
• The adapter slot an I/O card is placed in
determines which bay I/O is routed to
• Each physical adapters port can have up to 4
x1 links or 1 x4 link
• The adapter determines the number and type
of links consumed
• Adapters can use up to 8 ports per I/O
adapter depending on adapter selected
• Multiport adapters mean one has to take into
consideration the number of internal as well
as external ports one has available on an I/O
Module.
© 2013 IBM Corporation 23
CMM Chassis Management
CMM Chassis Management selections deal with components within the Chassis itself:
• PSUs
• Fans
• Switch Modules
• Compute Nodes
© 2013 IBM Corporation 24
Servers
Storage
Network
Virtual Machines
VM VM VM
Integrated Management of Infrastructure at each point in the lifecycle
IBM Flex System Manager
Hardware
Setup
System
Configuration
System
Provisioning
Virtual Machine
Lifecycle
Management
Infrastructure
Lifecycle
Management
Upward Integration
Single Point of Control Management: IBM Flex System Manager
© 2013 IBM Corporation 25
FSM GUI