BladeCenter™
Transcripción
BladeCenter™
BladeCenter™ Barcelona Supercomputing Center 26 de septiembre de 2005 Albert Valls Badia I & I solutions architect © 2005 IBM Corporation IBM Systems Group Agenda Tendencias de mercado BladeCenter ► Chasis ► Blades Opciones de conectividad ► Switches GB/CPM ► Switches FC ► OPM Gestión del sistema Cosumo de potencia © 2005 IBM Corporation Tendencias de mercado © 2005 IBM Corporation IBM Systems Group Today systems view Efficiency increase by consolidating servers High density blades for infrastructure migration Performance and manageability increase Big SMP systems Blades Rack optimized and blade servers fulfill traditional and new customer requirements. © 2005 IBM Corporation Rack optimized Tower IBM Systems Group Blade market view RISC and Intel Server Units 7.9 4.4 5.1 5.7 6.4 $60 7.0 $50 Blades Rack Tower 4,0 Billions Millions 8,0 6,0 RISC and Intel Server Revenue 2,0 0,0 Blades Rack Tower $30 $20 $0 2002 2003 2004 2005 2006 2007 2,203 2000 1,480 1500 868 1000 470 39 209 2002 2003 2004 2005 2006 2007 2002 2003 2004 2005 2006 2007 RISC and Intel Blade Revenue RISC Intel Millions Thousands 44 $10 2500 0 43 52 $40 RISC and Intel Blade Units 500 42 49 47 $7.000 $6.000 $5.000 $4.000 $3.000 $2.000 $1.000 $0 6,288 4,275 2,599 RISC Intel 1,454 98 640 2002 2003 2004 2005 2006 2007 Source: IDC September 2003 Forecast © 2005 IBM Corporation IBM Systems Group BladeCenter Revenue Share Momentum 44% 32% IBM #1 revenue/volume share position last 4 quarters Source: IDC and Systems and Technology Group Market Intelligence © 2005 IBM Corporation IBM Systems Group IBM eServer architectures Big SMP systems Scale Up / SMP Computing i890 / p595 / z9 virtualization Clustering Highly populated racks BladeCenter™ x460 e326 Scale Out / Distributed computing © 2005 IBM Corporation IBM Systems Group Service Providers Intranets Extranets Edge Servers BladeCenter: Target environments Web Presentation Services Transaction Servers Workloads ►Database Application Servers Data Servers Infrastructure Servers Storage Internet Enterprise Server ●DB2, SQL, Oracle ►Transaction ●EAS applications Application Server Workloads ►Collaboration ●Exchange, Domino, SendMail, Bynari ►Advanced Edge Server Workloads ►Edge of network apps ●DNS, caching, load balancing Web Serving ●WebSphere, MS IIS, MS Content Server, BEA WebLogic, appliances ►Workgroup Infrastructure ●File/print: Novell, MS, Samba ●Terminal serving: Citrix MetaFrame ●Small homegrown apps ●SteelEye LifeKeeper ►S&TC applications ●EDA: Cadence, Mentor ●Digital rendering farm ●Homegrown apps © 2005 IBM Corporation Application Server Workloads II ►Commerce ►EAS front-end app servers ●SAP, PeopleSoft, Siebel, Kana ►Data Marts ►Point industry apps IBM Systems Group BladeCenter key considerations: eServer BladeCenter technology is not only a new product: - Cornerstone for “scale-out” and distirbuted computing offering On Demand platform open standards based: - Automation (IBM Director, PFA, EXA...) n io t a ov n IBM ^ In BladeCenter - Virtualization (VMware, HPC Linux clusters...) Core value added offering is INTEGRATION: - Servers, network, storage and applications. - Easy deployment with Advanced systems management. © 2005 IBM Corporation Au teLiza on om ic s ard tand nS Ope - Integration (Intel, PowerPc, Linux, Microsoft, Novell...) IBM Systems Group Step 1 Consolidate Servers Firewalls Routers (Layer 3 Switches) Layer 4-7 Switches SSL Appliances Public Internet/ Intranet Clients SSL Appliances Clients Layer 2 Switches Caching Appliances Network Servers Application Servers Security Servers Application Servers Storage Fibre Switches Storage Fibre Switches SAN © 2005 IBM Corporation Caching Appliances Web Servers Security Gateway File Servers WebSphere Application Servers IBM Systems Group Step 2 Integrate First Layer of the Network (L2) SSL Appliances Caching Appliances Firewalls Layer 4-7 Switches Layer 2 Switches Storage Fibre Switches Storage Fibre Switches SAN © 2005 IBM Corporation Routers (Layer 3 Switches) Public Internet/ Intranet Clients SSL Appliances Caching Appliances IBM Systems Group Step 3 Integrate Storage Fabric SSL Appliances Firewalls Layer 4-7 Switches Caching Appliances SSL Appliances Caching Appliances Storage Fibre Switches Storage Fibre Switches SAN © 2005 IBM Corporation Routers (Layer 3 Switches) Public Internet/ Intranet Clients IBM Systems Group Step 4 Integrate Second Layer of the Network (L4-7) SSL Appliances Firewalls Layer 4-7 Switches Caching Appliances SSL Appliances Caching Appliances SAN © 2005 IBM Corporation Routers (Layer 3 Switches) Public Internet/ Intranet Clients IBM Systems Group Step 5 Consolidate Applications Firewalls SSL Appliances SSL Appliances Caching Appliances Caching Appliances SAN © 2005 IBM Corporation Routers (Layer 3 Switches) Public Internet/ Intranet Clients IBM Systems Group Step 6 Result BladeCenter Clients Consolidate Collapses Complexity Firewalls SAN © 2005 IBM Corporation Routers (Layer 3 Switches) Public Internet/ Intranet Clients IBM Systems Group Simplifying Datacenter Topology 2 3 7 5 9 6 4 8 10 1 Typical Datacenter Configuration 1. Ten x86 1U 2-way servers 1. Layer 2 GbE switches 2. KVM switches 2. RISC-based 2-way server 3. Ethernet cables 3. HPQ 4-way server 4. KVM cables 4. Alteon L7 E’net switches 5. Power cables 5. FC SAN switches / Cables © 2005 IBM Corporation Bladed Datacenter Configuration IBM eServer BladeCenter IBM Systems Group BladeCenter: where to use it Phisycal consolidation Rack/ Environment space constrained customers Customers looking for fast and simple deployments High scalability customer requirements High availability and reliability customer requirements Remote management for all IT infrastructure © 2005 IBM Corporation BladeCenter © 2005 IBM Corporation IBM Systems Group What a blade is? A "server on a card" - each "Blade" has its own: ►processor ►ethernet ►memory ►optional storage IBM Blade - in its own ruggedized chassis ►etc. The chassis= BladeCenter provides shared: ►Redundant management module* (KVM) (Keyboard, Video, Mouse) ►Redudndant Power supplies* ►Redundant blowers* ►Redundant Network and Fibre Channel switches, OPM* ►Phisycal* and virtual CD-ROM ►Phisycal* and virtual floppy ►USB port ►etc. * Hotswap devices © 2005 IBM Corporation IBM Blade - with its cover on - ready for insertion into the BladeCenter Chasis IBM BladeCenter - 7U IBM Systems Group IBM eServer BladeCenter Up to 4 processors per blade Up to 14 blades per chassis Up to Six 7U chassis per rack Full performance and manageability of rack-optimized platforms ... ... at TWICE the density of most comparable non-blade 1U servers © 2005 IBM Corporation IBM Systems Group IBM eServer BladeCenter: Chasis Switches Gigabit Ethernet (Layer 2) ► Commodity level networking ► Link aggregation ► VLAN creation and management Switches Nortel Layer 2-7 ► Advanced networking ► Content-based routing Fibre Channel Switches (2Gb FC Fabric) ► Lower cost via integration ► Full support of FC-SW-2 standards Power (4 x 2000W load-balancing) ► Upgradeable as required ► Redundant and load balancing for high availability Calibrated, vectored cooling™ ► Fully fault tolerant ► Allow maximum processor speeds KVM Switches / Management Modules ► Full remote video redirection ► © 2005/IBM Corporation Out-of-band lights out systems management KVM Switch / Redundant Power Nortel Layer 2-7 Switch Ethernet Switch Redundant Blower Management Module IBM Systems Group Redundant Midplane for server connectivity. No Single Point Of Failure (NSPOF) design. © 2005 IBM Corporation IBM Systems Group Sharing Redundant Power to Blades 1 2 1800/2000 watt Power Modules Blades 1 through 6 © 2005 IBM Corporation 3 middle plane 4 1800/2000 watt Power Modules Redundant middle plane Blades 7 through 14 IBM Systems Group BladeCenter System Cooling and Airflow Two Curved Impeller Blowers Capable of 325 Cubic Feet Per Minute (CFM) each 250 CFM each in standard operation Hot Swap, Redundant Predictive blower failure by monitoring the blower RPM Back flow dampers (louvres) Fan speed control Acoustic Attenuation Module Noise Reduction for acoustically sensitive (or regulatory) environments © 2005 IBM Corporation IBM Systems Group BladeCenter Management Module (MM) Service Processor Hot Swap Interfaces via midplane 10/100Mb Ethernet KVM not supported on JS20 (SoL) RS-485 interface I2C interfaces (serial interconnect daisy chain technology used for hardware level functions) Optional redundant Management Module © 2005 IBM Corporation IBM Systems Group Blade Servers portfolio HS20 2-way Xeon Features HS40 4-way Xeon Intel Xeon DP Intel Xeon MP processors processors EM64T Delivers bladed 4-way Mainstream rackdense blade server SMP capability Supports Windows and Linux Edge and mid-tier Target Apps workloads JS20 POWER-based Two PowerPC 970 processors 64-bit performance at IA32 price Performance for VMX Back-end workloads deep computing clusters 64-bit HPC Large mid-tier apps Web Serving Collaboration Web serving Infrastructure One Common Chassis and Infrastructure © 2005 IBM Corporation IBM Systems Group IBM BladeCenter HS20 Overhead View SCSI Expansion Processor slot 2 (air baffle connector terminator IDE connector 1 IDE connector 2 Upper midplane connector Daughter Card connectors Lower midplane connector Processor slot 1 © 2005 IBM Corporation DIMM 4 DIMM 1 IBM Systems Group Blade HS20 –Intel Xeon DP processor Up to Two Intel High Performance Xeon Processors ► 2.8Ghz, 3.06Ghz, 3.2Ghz-1M cache, all with 533Mhz Front Side Bus ► 3.2Ghz-2M cache (coming soon!) Integrated Mirroring for Local IDE Drives (80GB total capacity) Support for Optional pair of local HS SCSI drives with Mirroring Support for one optional Expansion Card (FC, GBE, Myricom) Integrated pair of GBE connections Dedicated Systems Management Connection OS Support ► Windows ► Red 2000, 2003 Hat and SUSE Linux ► VMWare and Novell Netware © 2005 IBM Corporation IBM Systems Group BladeCenter HS20 - Intel Xeon Processors IBM first blade vendor to ship an EM64T-enabled blade offering HS20 2-way Xeon ► Intel Xeon 3.06GHz/1MB-L2 533MHz FSB ► Based on same proven Xeon HS20 architecture Substantial performance increase over Xeon 533MHz FSB Full complement of supported options Support for integrated networking and storage connectivity such as Cisco, Nortel, Brocade, etc. Dedicated systems management connection Concurrent Serial Over LAN connectivity Complete list of supported OS’s including several 64-bit enabled systems Price parity with current 533MHz speed bins Bottom Line: HS20 with EM64T has the performance, compatibility, and pricing to make it production ready today and 64-bit enabled for tomorrow © 2005 IBM Corporation IBM Systems Group HS20 Feature Comparison IBM eServer BladeCenter HS20 (533MHz) (8832) Dual Intel Xeon 2.8GHz/3.06/3.21MB/3.2-2MB with 533MHz Front Side Bus 14 Blades per Chassis (30mm blade width) 2 Gb Ethernet Ports standard 4 DIMM slots Up to (2) 40GB IDE with IDE RAID 1 standard Internal switches (Enet/FC/KVM) Redundant/hot swap fans standard Hot swap power optional for bays 7-14 Redundant/hot swap mgmt optional Support for internal IDE and SCSI Storage Expansion Unit Support for IBM Director/RDM © 2005 IBM Corporation IBM eServer BladeCenter HS20 (800MHz) Dual Intel Xeon EM64T 3.2GHz / 3.4GHz / 3.6GHz with 800MHz Front Side Bus 14 Blades per Chassis (30mm blade width) 2 Gb Ethernet Ports standard 4 DIMM slots Up to (2) 73GB SFF SCSI with RAID 1 stnd Internal Switches (Enet/FC/KVM) Redundant/hot swap fans standard Hot swap power optional Redundant/hot swap mgmt optional Support for NEW SCSI Storage Expansion Unit Support for dual SCSI drives and Expansion Card Support for IBM Director/RDM IBM Systems Group HS20 Improvements - Performance 800 MHz Front Side Bus 1.5 times the system bus bandwidth when compared to 533Mhz Front Side Bus ► Helps support faster Web site response times, more users, and greater business ► With Intel® EM64T 64-bit CPU core extensions (EM64T) Improved throughput in targeted applications ► Full support for 64-bit OS with legacy support for 32-bit and 16-bit ► DDR2 400 Memory 20% increase in memory bandwidth over DDR333 ► 40% reduction in the power required to run the memory ► PCI-Express expansion capability Bottom Line: operational enhancements to increase performance, efficiency and timing margins for high performance computing © 2005 IBM Corporation IBM Systems Group HS20 Improvements - Flexibility On-board SCSI HDDs replaces IDE Two U320 small form factor non hot-swap HDDs- 36 or 73GB ► Better performance, better reliability, and choice of capacity ► Support for two HDDs + a new SFF Daughter Card ► Improved I/O: no longer need to sacrifice an HDD to get Fibre or Ethernet connectivity SCSI RAID with BSE-2 Option RAID controller in BSE delivers RAID1 and RAID1E ► Four additional I/O ports available when adding Eth expansion cards ► Two hot swap U320 drives at capacities up to 144GB currently ► Smart power management Processor can adapt to changes in utilization that allow reduced power consumption during non peak hours ► Smarter power management methods help customers reduce power infrastructure requirements ► Bottom Line: choice and flexibility to further the leadership position of IBM BladeCenter © 2005 IBM Corporation IBM Systems Group New HS20 Options BladeCenter SCSI Expansion Unit 2 ► ► ► New SCSI Expansion for Xeon EM64T only Two hot swap hard drives, increased RAID function, and more I/O capabilities Up to 8 ports per blade for network connectivity Set of processor options ► 2.8 through 3.6MHz 800FSB Xeon EM64T Two new SFF U320 HDDs ► 36GB and 73GB non hot-swap offerings New 2000W power supply and DVD in 3XX chassis already shipping Gigabit Ethernet Expansion Card Fibre Channel Expansion Card © 2005 IBM Corporation IBM Systems Group Scale Up Meets Scale Out: HS40 Uses existing infrastructure ►Same chassis as HS20 and JS20 ►Same options, 4Gb Ethernet standard ►Seven 4-way systems in 7U ►Four with Local SCSI option Intel Xeon MP 2.8 and new “double cache” processors Application targets ►Back end workloads (SAP, PeopleSoft, JD Edwards) ►Larger Mid-Tier Applications (Exchange, Notes) Supports Microsoft Windows, Linux, and VMware © 2005 IBM Corporation HS40 4-way Xeon IBM Systems Group JS20 BladeCenter based on POWER4 64-bit POWER at 32-bit Price Two 2.2 GHz PowerPC 970 processors, derived from the POWER4 architecture VMX capabilities provide enhanced compute-intense performance AIX 5L v5.2 supported today SuSE SLES8 y SLES9, Red Hat Enterprise Level 3 U2 support today IBM Director and Cluster Systems Management support Heterogeneous platforms integrated into single chassis PowerPC 970 performance features 90-nanometer silicon-on-insulator 8-way superscaler design, issues up to 8 instructions per clock cycle Vector-processing unit with more than 160 specialized vector instructions © 2005 IBM Corporation JS20 POWER-based IBM Systems Group Feature Comparison vs first model GAed IBM ~ BladeCenter JS20 Dual IBM PowerPC 970 1.6 GHz processors with 800 MHz Front Side Bus SIMD VMX extensions providing exceptional performance for compute- intensive floatingpoint applications 14 blades per Chassis (30mm blade width) 2 Gbps Ethernet Ports standard 4 DIMM slots Up to (2) 40GB IDE with IDE RAID 1 standard Internal switches (Ethernet/Fibre Channel) Redundant/hot-swap fans standard Hot-swap power optional Redundant/hot-swap mgmt optional Support for IBM Director/CSM Support for Linux SLES 8 © 2005 IBM Corporation Since october 2004 IBM ~ JS20 PowerPC 970 2.2 GHz processors Dual IBM BladeCenter with 1.1 GHz Front Side Bus SIMD VMX extensions providing exceptional performance for compute- intensive floating-point applications 14 blades per chassis (30mm blade width) 2 Gbps Ethernet Ports standard 4 DIMM slots Up to (2) 40GB IDE with IDE RAID 1 standard Internal Switches (Ethernet/Fibre Channel) Redundant/hot-swap fans standard Hot-swap power optional Redundant/hot-swap mgmt optional Support for IBM Director/CSM Support for SLES 8, SLES 9 Support for Red Hat Enterprise Linux AS 3 Support for AIX 5L V5.2 Opciones de conectividad © 2003 IBM Corporation IBM Systems Group Expanding BladeCenter Ecosystem Wide range of companies convinced that BladeCenter architecture will add value to their customers’ solutions Industry-leading technology companies delivering innovative business solutions running on Windows, Linux, Novell More choices for customers © 2005 IBM Corporation IBM Systems Group Expanding BladeCenter Capabilities Software IBM Director RDM Virtualization Engine Deployment / Provisioning Partitioning/VMware IBM Cluster Sys Mgmt 2-way PowerPC Blade Blade Solutions 2-way Xeon Blade EM64T Blade Telco Chassis Hosted Clients 4-way Xeon MP Blade Networking L2 Ethernet Switch L4/7 Ethernet Switch Cluster Switch Storage InfiniBand Switch Cisco Switch Local IDE & SCSI iSCSI Fibre Channel HBA NAS © 2005 IBM Corporation FC Switch Brocade Fibre Switch IBM Systems Group Fibre Channel Expansion Cards High Performance Host Bus Adapter supporting both 1 & 2 Gbps devices Provides TWO 2Gbps fibre channel port connections for the HS20, HS40 and JS20 “Boot from SAN” support in a variety of storage environments Extensive certification from major storage and SAN manufacturers Equivalent function to QLA2342 and uses ISP2312 chip technology Low-cost design for high-density BladeCenter servers © 2005 IBM Corporation IBM Systems Group Myrinet Cluster Expansion Card Overview Announce & General Availability: 9/9/2003 Myrinet is a high-speed/low-latency cluster interconnect for High Performance Computing (HPC) applications. It provides high performance system-to-system connection for distribution of computing over several blades. The Myrinet Cluster Expansion card on IBM eServer BladeCenter provides the same functions and performance, and uses the same software as the standard Myrinet PCI-X card on highly successful server clusters using xSeries rack-optimized servers. Myrinet is an ANSI standard, with open-source software, and is the market leader in high-performance, high-availability, cluster interconnect. Approximately one third of the TOP500™ supercomputer sites use Myrinet technology. Single port card ► BladeCenter Myrinet HCA is designed to fit the BladeCenter form factor. It is analogous to the 2MB version of the PCI-XD interface ► High performance by distributing demanding computations across an array of cost-effective servers ► High availability by allowing a computation to proceed with a subset of the hosts. The interconnect is capable of detecting and isolating faults and using alternative communication paths ► http://www.myricom.com/ © 2005 IBM Corporation IBM Systems Group Myricom Enables High Speed Network Connectivity High-speed/low-latency cluster interconnect for HPC applications Provides high performance system-to-system connection for distribution of computing over several blades • Single port card • BladeCenter Myrinet HCA designed to fit the BladeCenter form factor • High performance by distributing demanding computations across an array of cost-effective servers • High availability by allowing a computation to proceed with a subset of the hosts. • Interconnect capable of detecting and isolating faults and using alternative communication paths High performance connectivity for high performance clusters © 2005 IBM Corporation IBM Systems Group BladeCenter Copper/Ethernet Switch Portfolio NEW IBM eServer BladeCenter 4-port Gb Ethernet Switch Module NEW IBM eServer BladeCenter Copper Pass-thru Module Supplier: DLink Layer 2 Switching Trunking and Link Aggregation Supplier: IBM GbE Pass-thru No switching function © 2005 IBM Corporation Nortel Networks ® L2-7 GbE Switch Module Cisco Systems ® Intelligent Gb Ethernet Switch Module Supplier: Cisco Layer 2 Switching Layer 3/4 services Supplier: Nortel Layer 2 - 7 functionality Layer 3/4 services Load balancing Routing / switching Advanced filtering Content intelligence A full suite of integrated offerings to provide additional flexibility and choice! IBM Systems Group IBM eServer BladeCenter Copper Pass-thru Copper Pass-thru Module Cable Copper Pass-thru Module © 2005 IBM Corporation IBM Systems Group BladeCenter Copper Pass-thru Module Server bay – ports – cable assignments © 2005 IBM Corporation IBM Systems Group Expanding BladeCenter Networking Hardware Cisco Systems Intelligent Gigabit Ethernet Switch Module Integrates Cisco networking technology into BladeCenter Helps reduce datacenter complexity and networking complexity Comprehensive set of Layer 2 features with Layer 3/4 services ► Multicast – IGMP Snooping ► QoS features Supports IOS (Cisco Internetworking Operating System) Reduces deployment and configuration time Only blade solution in industry with embedded Cisco switching © 2005 IBM Corporation IBM Systems Group Cisco Systems IGESM Description Equivalent software feature set to Cisco Systems® Catalyst 2950 providing Layer 2+ functionality High Availability: Enhanced Spanning Tree Protocol, IGMP snooping Enhanced Security: 802.1x, Port Security, MAC address notification, RADIUS/TACACS+ Advanced QoS: 802.1p, WRR, Strict Priority Queuing Interfaces Wire-speed switching ► 4 - 1GB External Ethernet (Copper) interfaces ► 14 -1GB Internal interfaces to blades ► Management / Monitoring Cisco IOS Command Line Interface ► Cluster Management Suite ► SNMP - Management Information Base (MIB) based applications such as CiscoWorks ► Management and Power through Management Module ► Console Port on faceplate ► Enhanced Default Configuration ► Multiple VLANs configured as default at power-up © 2005 IBM Corporation IBM Systems Group Nortel Networks Layer 2-7 GbE Switch Module Availability ► Reduce unplanned application down-time in the event of a switch module, server blade, or chassis failure ► Reduce need for planned application downtime Performance ► Enable on demand computing ► Better serve the processing demands of bandwidthintensive applications ► Enhance application performance Manageability ► Reduce time/effort required to deploy new datacenter infrastructure ► Simplify datacenter administration Greater infrastructure scalability Enhanced server security Integrating L2-7 switch into blade chassis reduces datacenter infrastructure TCO by as much as 65% © 2005 IBM Corporation Layer 5-7 Layer 4 Layer 2/3 IBM Systems Group BladeCenter Optical Module/SAN Switch Portfolio Brocade® Enterprise NEW QLogic™ Enterprise IBM eServer BladeCenter Brocade® Entry SAN Switch Module 6-port Fibre Channel Switch Module Optical Pass-thru Module Supplier: IBM Provides unswitched / unblocked optical connection Up to 14-optical connections to SAN Switch Module Supplier: QLogic Equivalent to SANbox 5200 6-1/2Gb Auto sensing ext ports Cascades to (239) Switches Supplier: Brocade Equivalent to Silkworm 3900 2-1/2Gb Auto sensing ext ports Cascades to (2) Switches Supports Brocade Advanced Feature Key options Supplier: Brocade Equivalent to Silkworm 3900 2-1/2Gb Auto sending external ports Cascades to (239) switches Supports Brocade Advanced Feature Key options Supports performance monitoring and advanced zoning FC-SW-2 Compliant external SAN (requires breakout cable option) A full suite of integrated offerings to provide additional flexibility and choice © 2005 IBM Corporation IBM Systems Group Fibre Channel Switch Module The Fibre Channel Switch Module supports the following: Two small form factor pluggable (SFP) ports Long Wave option Short Wave option Self-Configuring according to type of device attached Full Interoperability with Brocade Fabric SFP Transceiver (not included) © 2005 IBM Corporation IBM Systems Group Brocade® Switch Modules Delivers datacenter standards for Brocade customers Seamless connectivity to over 3.5M existing Brocade switch ports Compatibility with Brocade Fabric OS features ► Trunking, Advanced Performance Monitoring, Advanced Security, Zoning, Extended Fabric and Remote Switch Simplifies SAN Management ► IBM Director integration – Q304 ► Tivoli SAN Manager ► Fabric Manager and W ebTools are also supported Modular scalability ► Available as a 2 domain or full fabric switch module Flexible deployment ► High availability and ease of serviceability Fabric switch delivering Brocade functions including performance, manageability, scalability and security to support demanding SANs © 2005 IBM Corporation IBM Systems Group Optical Pass-Thru Module Provide direct connectivity between server blade and external devices Ethernet network devices Fibre Channel network devices Myrinet cluster switches TIP OPM cables do not come with the OPM (must be ordered separately) The 4 SC or LC duplex optical connectors are keyed to the processor blade bays via the ports on the OPM © 2005 IBM Corporation IBM Systems Group Optical Pass-thru Module Overview Announce: 9/9/2003 General Availability: 9/9/2003 The Optical Pass-Thru (OPM) is a module that inserts into any switch module bay and provides connectivity to each blade bay. The OPM provides an unswitched/unblocked network connection to each blade server. The BladeCenter Optical Pass-thru module can be used in conjunction with the Myrinet® Cluster Expansion card to deliver a high-performance, high-availability interconnect for High Performance Technical Computing and other cluster-computing applications. The BladeCenter Optical Pass-thru module can be used in conjunction with the Fibre Channel Expansion card to allow an alternative connection to Storage Area Networks. ► The Optical Pass-thru Module can be an alternative to the IBM eServer BladeCenter 2-port Fibre Channel Switch Module. As with the integrated Fibre channel switch module, the pass-thru module provides connectivity to the IBM TotalStorage™ family of products including FAStT, Enterprise Storage Server, SAN switches and tape storage. OPM includes (4) Optical Transceivers and (0) cables ► 02R9080 - IBM eServer BladeCenter Optical Pass-thru Module ► 73P5992 - IBM eServer BladeCenter Optical Pass-thru Module SC Cable ► 73P6033 - IBM eServer BladeCenter Optical Pass-thru Module LC Cable © 2005 IBM Corporation IBM Systems Group OPM Functions The OPM provides the ability to transmit and receive network data traffic between all (14) blade bays and the networking environments below ► Gb Ethernet ● To enable Ethernet, the OPM can be inserted into switch module bays 1, 2, 3 and 4 ● If the OPM is inserted into switch module bays 1 or 2, the OPM interfaces with the integrated dual Gb Ethernet controllers on the blade server ● In order for the OPM to function in bays 3 and 4, the Gb Ethernet Expansion Card is required on the blade server ► Fibre Channel ● To enable Fibre channel, the OPM can be inserted into switch module bays 3 and 4 ● In order for the OPM to function in bays 3 and 4, the Fibre Channel Expansion Card is required on the blade server ► Myrinet ● The Myrinet® Cluster Expansion Card is a SINGLE port card and is hardwired to switch module bay 4 ● To enable Myrinet, the OPM must be inserted into switch module bay 4 ● The Myrinet Cluster Expansion Card is required on the blade server © 2005 IBM Corporation IBM Systems Group Switch Module Bay Configurations Supported BladeCenter Chassis Rear View SM BAY 3 & 4 L2-7 Gb Ethernet (Nortel ESM) SM SM PS 3 1 3 BLOWER 1 PS 1 M M 1 Fiber Channel (IBM FCSM) OPM SM BAY 1 & 2 SM SM PS 4 2 4 BLOWER 2 PS 2 M L2 Gb Ethernet (IBM ESM) L2-7 Gb Ethernet (Nortel ESM) M 2 OPM NOTE: 2. Any combination of modules listed under SM BAY 1 & 2 can be inserted into BAYS 1 & 2 3. The switch module in BAY 4 must match the switch module in BAY 3 and the corresponding I/O Expansion card is required on the blade server to enable BAYS 3 and 4 © 2005 IBM Corporation IBM Systems Group 2 Mo l e x M o d u l e Co n n e c to r 1 Vi te s s e VSC3 1 39 3 Pro g . Cl o c k 4 Se ri a l Pro m 8055 Po we r Re g . Transceivers: 4 standard No cables will be included in base OPM option. Fan out cables must be ordered separately. © 2005 IBM Corporation To BladeCenter MidPlane Purchase cable (s) to connect to External Device High level schematic of standard configuration IBM Systems Group Fanout Cable Connection A Connection B Connection C Connection D The OPM supports up to (4) fanout cables (one shown above) Each cable breakout will be labeled to identify the transceiver port One end of the fanout cable plugs into the transceiver, the other end fans out to four cables Depending on which cable is purchased, the connection at A, B, C, or D will be SC or LC Total cable length 1.5 meters © 2005 IBM Corporation IBM Systems Group Blade Bay - Transceiver Configurations Each OPM transceiver and port has a dedicated connection to a blade bay Blade Bay Num ber 1 Tr ansceiver Number 1 Tr ansceiver Por t A 2 1 B 3 1 C 4 1 D 5 2 A 6 2 B 7 2 C 8 2 D 9 3 A 10 3 B 11 3 C 12 3 D 13 4 A 14 4 B © 2005 IBM Corporation OPM Rear View: 1 2 3 BladeCenter Chassis Front View: LED PANEL 1 2 3 4 USB CD 5 6 7 8 USB FDD 9 10 11 12 13 14 4 IBM Systems Group OPM offers additional configuration flexibility The OPM allows the chassis to support multiple Expansion cards given the direct link from blade bay to OPM transceiver port Blade 3 FC Expansion Card Blade 6 Gb Ethernet Card Card Port 1 Chassis SM BAY 3 OPM 1 Card Port 1 Blade Bays 1-14 Transceiver 1, Port C To SAN Fabric Transceiver 2, Port B To Ethernet Network Blade 10 Gb Ethernet Card Card Port 1 Card Port 2 SM BAY 4 OPM 2 Card Port 2 Blade Bays 1-14 Transceiver 3, Port B To Ethernet Network Example 1: Blade #3 has Fibre Channel connection on Port 1 to OPM 1, Transceiver 1, Port C Example 2: Blade #6 has Gb Ethernet connection on Port 1 to OPM 1, Transceiver 2, Port B Example 3: Blade #10 has Gb Ethernet connection on Port 2 to OPM 2, Transceiver 3, Port B © 2005 IBM Corporation Gestión © 2005 IBM Corporation IBM Systems Group BLADE CENTER: POSIBILIDADES DE GESTIÓN Gestión Nativa (HW) Basada en Navegador Web Control total via Módulo de Gestión Gestión total de switches ethernet y fibra via Módulo de Gestión Actualización remota de firmware IBM Director 4.12 Soluciones de gestión inteligente simplificada Optimización Asistentes para despliegue Herramientas para configuración masiva Planificación de tareas para eventos Remote Deployment Manager Herramientas de despliegue Rapida restauración y despliegue de imágenes Integración con IBM Director 4.12 (Blade en espera) Retirada/Baja segura del servidor (datos) Remo te Targ ets Deployment Image Capture RDM Imag e L ib rary Preco n fig u red "Do n o r" system © 2005 IBM Corporation IBM Systems Group GESTIÓN BASADA EN HW © 2005 IBM Corporation IBM Systems Group IBM DIRECTOR: HERRAMIENTAS ESPECÍFICAS PARA BLADES Detección Automática Usa SLP (Service Location Protocol) para automaticamente identificar nuevos chasis y blades Nuevos Datos de Inventario Datos vitales del chasis Slots Ocupados chassis Management Module out-of-band IP Address Nuevos Eventos Alertas de entorno para temperaturas, velocidad de los ventiladores, etc... Alertas para cambios de la configuración como inserción y retirada de blades Habilita automaticamente blade en espera via RDM Nueva Gestión de Objetos para Chasis y Blades Dos Nuevos Grupos Dinámicos BladeCenter Chasis BladeCenter Chasis y Miembros del Chasis © 2005 IBM Corporation IBM eServer™ GESTIÓN DEL BLADECENTER EN IBM DIRECTOR 4.12 Rack Manager Interfaz gráfica incluyendo soporte para chasis y blades Agregar / Retirar chasis y blades del rack Estado de la "salud" del sistema para chasis y blades Asistente para el Módulo de Gestión Soporte para Módulo de Gestión del BladeCenter y configuración de los switches Otras Extensiones IBM Director (Capacity Manager, Software Rejuvenation, System Availability) Funcionan en los blades individualmente © 2005 IBM Corporation © 2003 IBM Corporation IBM Systems Group Pueden ejecutarse Tareas Drag and Drop en cada Blade o en el grupo BladeCenter © 2005 IBM Corporation El grupo BladeCenter contiene Blades con Notes para cada grupo de trabajo IBM Systems Group Vista detallada del Blade seleccionado BladeCenter con los resaltado blades instalados Vista gráfica del BladeCenter instalado en el rack © 2005 IBM Corporation IBM Systems Group Visión general de la Gestión de Sistemas * Tivoli Enterprise Remote Deployment Manager Server Pack Plus Workload Management Software Distribution IBM Director: Entorno Base Gestión de Sistemas Integrado en el Servidor(Service Processor)/ Módulo de Gestión © 2005 IBM Corporation Remote Supervisor Adapter (RSA II) IBM Systems Group IBM Director Portfolio - Advanced, predictive tools with selfmanaging technologies - Deliver optimal server performance and high availability Save time and money by remotely replicating the install of multiple systems, including blades - Easily distribute application packages remotely from a single console, saving travel & labor costs - Higher server utilization by protecting availability and performance of workloads on that server - RDM Plus Pack Server Plus Pack Capacity Manager Software Rejuvenation Rack Manager System Availability Active PCI Manager Remote Deployment Manager Remote, unattended system deployment Updates system and option firmware No limitation on number of systems installs Restores system hard drives w/ PowerRestore Basic Hardware Management Inventory Monitoring Alerting Group Management RAID Manager Mgmt Processor © 2005 IBM Corporation AWM SWD Software Distribution Premium Edition Can package and distribute software targeted to an end user or group of users IBM Director v4.1 5000 Managed Nodes Upward Integration (Tivoli, CA, HP, MS SMS, BMC , NetIQ) IBM Director Agent Application Workload Manager (AWM) Allows multiple applications to share a server efficiently and reliably Manages resource contention Help Desk & Support Remote Control Remote Session File Transfer Real Time Diagnostics IBM Systems Group Topología de IBM Director IBM Director Server Application Logic Database (Windows or Linux) Management Console(s) Managed Clients (Servers, Desktops, Laptops) Hardware Management Inventory Monitoring Alerting Group Management RAID Manager Mgmt Processor © 2005 IBM Corporation Java GUI (Windows or Linux) IBM Director v4.1 5000 Managed Nodes Upward Integration (Tivoli, CA, HP, MS SMS, BMC , NetIQ) IBM Director Agent Help Desk & Support Remote Control Remote Session File Transfer Real Time Diagnostics IBM Systems Group Consola Centralizada de Gestión ►Interfaz de gestión intuitiva ►Indicadores ►Ejecución de “Salud del Sistema” identifican sistemas con problemas de tareas sobre grupos de sistemas con una simple acción de “arrastrar y soltar” con el ratón © 2005 IBM Corporation IBM Systems Group © 2005 IBM Corporation IBM Systems Group Server Plus Pack para IBM Director v4.1 Capacity Manager ► Monitoriza la utilización de recursos Hw (procesador, memoria, HDD, y tráfico de red ) . Capacidad para detectar y predecir cuellos de botella Rack Manager System Availability ► Proporciona información acerca de la disponibilidad y tiempo de parada de un sistema o grupo de sistemas Software Rejuvenation ► Predicción de posible agotamiento de recursos del SO ActivePCI Manager ► Optimiza el rendimiento del servidor ayudando a diseñar la colocación óptima de los adaptadores PCI © 2005 IBM Corporation IBM Systems Group Server Plus Pack - Capacity Manager Diagnostica problemas de rendimiento y analiza posibles cuellos de botella ► Muestra la información de rendimiento gráficamente , en un formato de fácil análisis ► Los gráficos y datos son exportables para ser utilizados en otras herramientas de generación de informes. © 2005 IBM Corporation IBM Systems Group Remote Deployment Manager Actualización de Microcódigo de Servidores /Actualización de parámetros de CMOS/ Instalación de Sistema Operativo Rápido despliegue de Imágenes Eliminación de datos segura Power Restore , partición oculta para salvaguarda. Configuración automatizada de controladoras RAID © 2005 IBM Corporation IBM Systems Group Distribución de Software Extensión a IBM Director para distribuir remotamente paquetes de software Puede distribuir paquetes de software a usuarios o grupos de usuarios Microsoft Windows Installer InstallShield Linux Red Hat Package Manager © 2005 IBM Corporation IBM Systems Group Application Workload Manager Extensión a IBM Director para controlar la utilización de los recursos del servidor por parte de múltiples aplicaciones Protege contra el agotamiento de los recursos por parte de una aplicación Guarda un registro con la totalidad de los recursos consumidos por los procesos durante su ejecución. Facilita una mejor utilización de los recursos del servidor por parte de varias apliaciones © 2005 IBM Corporation Consumo © 2005 IBM Corporation IBM Systems Group Contents Power trends ► Going to get more difficult before it gets better Where’s the power used in a server? ► You might be surprised BladeCenter was designed to reduce power and heat ► BladeCenter is more than a 1U turned on it’s side Airflow requirements © 2005 IBM Corporation IBM Systems Group The Industry problem Moore's Law continues – greater performance = more power Customers need help to make new kit fit into old data centers 175 150 CPU Power Rack Power 125 100 This is where we are now 75 50 25 0 2001 2003 2005 2007 2009 2002 2004 2006 2008 2010 © 2005 IBM Corporation 35 32.5 30 27.5 25 22.5 20 17.5 15 12.5 10 7.5 5 2.5 0 Thousands CPU Power - watts dual core solution in late 2006 Rack Power - watts Processor power expected rise over coming generations until the arrival of smarter IBM Systems Group Trends More powerful processors – Moore’s law at work More being used by HDDs More power being used by memory ► More DIMMs ► Future Fully Buffered DIMMs get even worse More power in = more heat out More air required to cool the servers 2006 ► 1U servers are likely to have 700W power supplies ► 2U servers closer to 800W power supplies ► BladeCenter chassis in 2006 can draw close to 7KW per chassis © 2005 IBM Corporation IBM Systems Group What limits density in the data center? Traditionally real estate or “U space” was the limiting factor Today it is far more likely that other factors drive down your density Power input ► UPS not large enough ► PDUs not capable of handling load ► Municipal power issues Thermal limits ► Old data center ► Not enough cooling capacity ► Limited air flow capabilities of room Weight limits of older floors ► Not designed to handle the 1500-2000lbs of today’s 1U/2U/blade racks. © 2005 IBM Corporation 25% less power 25% less heat 33% less than 1U IBM Systems Group What’s using the power? The processor power growth is the largest single contributor but there are many other areas- the more you pack into a server the more power it needs! BladeCenter helps in this area Low Voltage processors help in this area OTHER? Other 25% •AC to DC Transitions •DC to DC Deliveries Standby 2% •Fans and air movement Other 44% HDD 7% Memory HDD 13% 6% Planar 5% PCI 4%Standby 2% © 2005 IBM Corporation Processor 30% Processor 46% Memory 11% PCI Planar 3% 4% Processor Memory PCI Planar HDD Standby Other IBM Systems Group What IBM can effect . . . We have Power deliver –Super energy efficient power supplies deliver more power to the server – less wasted watts in AC to DC 2.8GHz Low Voltage Nocona 2.8GHz full power Nocona transition 3380 Watts per chassis 4920 Watts per chassis 2.8GHz processors, 4GB memory, Less parts two 36GB U320 SCSI, Two Port 44% Fibre HBA, Dual ethernet –Smarter shared infrastructure design means less **Same performance** components that draw power – less hardware means How less does watts that compare per blade? HS20 2.8GHz full power Xeon EM64T - 351W HS20 2.8GHz low voltage EM64T Smarter thermal 241W solution HS20 2.8GHz Xeon 400MHz –Smarter thermal solution reduces 266W the number of fans Nearly half HP Blade 3.2GHz Xeon EM64T* 469W from 112 down- to just 2 low power use blowers the power 30% Low Voltage Processor –Full performance 2.8GHz Xeon processor at substantial power savings over standard Xeon *HP entry Xeon EM64T is 3.2GHz. Intel guidance is all Xeon EM64T draw 103W per processor © 2005 IBM Corporation IBM Systems Group Air flow requirements BladeCenter blowers are completely redundant and hot swappable ►Note: These two blowers take the place of 112 individual fans found in a comparable 14 server 1U installation (8 fans/server) The MM controls blowers, they respond to external temperature changes Inlet temperature is critical- temperatures at the back of the rack are less critical At 25oC (77F) the airflow required by BladeCenter will be 250CFM (cubic feet per minute) As temperatures rise to 32oC the fans increase in speed to a maximum of 450CFM. Blowers also go to max in the event of a management module failure The increase is linear to the temperature increase For a 3 BladeCenter installation under normal data center conditions we need 750CFM of air flow © 2005 IBM Corporation IBM Systems Group Airflow comparisons • BladeCenter • 265 CFM per chassis at <25C • 28 processors • 9.5 CFM per processor • X336 • 37 CFM per server at <25C • 2 processors • 18 CFM per processor • X346 • 52 CFM per server at <25C • 2 processors • 26 CFM per processor © 2005 IBM Corporation IBM Systems Group BladeCenter Delivers IT Integration Storage ED AT MIC NO EN IBM IBM BladeCenter BladeCenter AUT O OP Servers VIR TU IZED AL INTE GR Applications Networking Integration can help dramatically reduce infrastructure costs © 2005 IBM Corporation IBM Systems Group Gracias © 2005 IBM Corporation