background image

BladeSymphony

®

 1000 

Architecture

White Paper

1000

Summary of Contents for BladeSymphony 1000

Page 1: ...BladeSymphony 1000 Architecture White Paper 1000 ...

Page 2: ...On Module Storage 25 I O Sub System 26 I O Modules 26 Embedded Gigabit Ethernet Switch 33 SCSI Hard Drive Modules 34 Chassis Power and Cooling 36 Module Connections 37 Redundant Power Modules 37 Redundant Cooling Fan Modules 38 Reliability and Serviceability Features 39 Reliability Features 39 Serviceability Features 40 Management Software 45 BladeSymphony Management Suite 45 Operations Management...

Page 3: ...he edge of the network the application tier the database tier or all three BladeSymphony 1000 is a system that CIOs can rely on Introducing BladeSymphony 1000 BladeSymphony 1000 provides enterprise class service levels and unprecedented configuration flexibility using open industry standard technologies BladeSymphony 1000 overcomes the constraints of previous generation blade systems to deliver ne...

Page 4: ...but with far greater reliability and scalability than competitive systems BladeSymphony 1000 can be deployed at the application tier similar to quad socket blade server offerings from HP and IBM but with greater reliability and scalability BladeSymphony 1000 ideal for the database tier similar to the IBM p Series or HP rack mount servers but with a mainframe class virtualization solution Designed ...

Page 5: ...eal platform for a wide range of data center scenarios including Consolidation BladeSymphony 1000 is an excellent platform for server and application consolidation because it is capable of running 32 bit and 64 bit applications on Windows or Linux with enterprise class performance reliability and scalability Workload Optimization BladeSymphony 1000 runs a wide range of compute intensive workloads ...

Page 6: ...bility Redundant Cooling Fan Modules four hot swap 3 1 per chassis standard configuration for high reliability and availability Switch Management Modules hot pluggable system management board up to two modules per system for high reliability and availability Figure 2 Key BladeSymphony 1000 components The server blades and I O modules are joined together through a high speed backplane Two types of ...

Page 7: ...nnel Switch modules Chassis Power and Cooling on page 36 provides details on the two chassis models as well as Power and Cooling Fan Modules Reliability and Serviceability Features on page 39 discusses the various reliability availability and serviceability features of the BladeSymphony 1000 Management Software on page 45 discuss software management features Virtage on page 48 provides technical d...

Page 8: ...or large in memory databases and very large data sets Each server blade also includes two gigabit Ethernet ports which connect to the internal gigabit Ethernet switch in the chassis as well as two front side accessible USB 1 1 ports for local media connectivity and one RS 232 port for debugging purposes Figure 4 Intel Itanium Server Blade Intel Itanium Server Blades include the features listed in ...

Page 9: ...bE SerDes 1 25 Gb sec 2 ports Wake on LAN supported USB Two ports per partition Fast Ethernet LAN manage ment Two100Base 10Base ports I2C One port Interface on the front of module USB Two ports per physical partition Compatible with USB 1 1 Serial One RS 232C port for debugging only I O function SCSI or RAID None I O module required for this function VGA None I O module required for this function ...

Page 10: ...on and containment across all major data pathways and the cache subsystem They also feature integrated standards based error handling across hardware firmware and the operating system Bridge Intel 1 PCIe to PCI X bridge South Bridge Intel 1 South bridge connects legacy devices SIO SMSC 1 Super I O chip contains the COM port and other legacy devices FW ROM ATMEL STMicro 8 MB A flash ROM storing the...

Page 11: ...he processor supports up to 24 MB 12 MB per core of low latency on die L3 cache 14 cycles providing 102 GB sec aggregate bandwidth to the processor cores It also include separate 16 KB Instruction L1 and 16 KB Data L1 cache per core as well as separate 1 MB Instruction L2 and 256 KB Data L2 cache per core for higher speed and lower latency memory access Hyper Threading Technology Hyper Threading T...

Page 12: ...ur server blades The server blades connect to each other through the node link maintain cache coherence collectively and can be combined to form a ccNUMA type multiprocessor configuration The Hitachi Node Controller is connected to memory modules through memory controllers The Hitachi Node Controller provides the interconnection between the two processors two memory controllers three PCI bus inter...

Page 13: ...2 MB 1 GB 2 GB and 4 GB DDR2 533 for a total of up to 64 GB per server blade or 16 GB per core The memory system is designed to control a set of four DIMMs for the ECC and the memory device replacing function Accordingly if DIMMs are added they must be arranged in four DIMM units The different DIMMs in each row can be used logically as shown in Figure 5 Figure 5 Memory configuration The memory sys...

Page 14: ...dded to a system the amount of contention for memory access quickly increases to the point where the intended throughput improvement of more processors is significantly diminished The processors spend more time waiting for data to be supplied from memory than performing useful computing tasks Conventional uniform memory systems are not capable of scaling to larger numbers of processors due to memo...

Page 15: ...operating systems take into account where data is located when scheduling tasks to run on CPUs using the closest CPU where possible Some operating systems are able to rearrange the location of data in memory to move it closer to the processors where its needed For operating systems that are not NUMA aware the BladeSymphony 1000 offers a number of memory interleaving options that can improve perfor...

Page 16: ...lly suited to online and other front end applications that can divide processing requirements across multiple servers Scaling out can also provide load balancing capabilities and higher availability through redundancy Figure 7 Scale up capabilities Scaling up is accomplished through SMP shown in Figure 7 This approach is better suited to enterprise class applications requiring 64 bit processing hi...

Page 17: ...NUMA also allows applications running on it to perform optimization taking advantage of node localized memory accesses and enabling higher system performance Figure 8 Full interleave mode and non interleave mode Mixture mode This mode specifies the ratio of a local memory at a constant rate There can be some restrictions on the ratio of local memory according to the NUMA function support level of ...

Page 18: ...ters the request and does not forward the request to the local processors This process is illustrated in Figure 10 Figure 10 L3 cache copy tag process Intel Itanium I O Expansion Module Some applications require more PCI slots than the two that are available per server blade The Intel Itanium I O Expansion Module provides more ports without the expense of additional server blades Using the Itanium...

Page 19: ... 5 4 3 2 1 0 CPUslot 7 Backplane IPFI O expan sion module CPU module 0 PCI X slot CPUslot 6 CPUslot 5 CPUslot 0 CPUslot 1 CPUslot 4 CPUslot 2 CPUslot 3 Mem NDC Bridge Bridge Bridge Bridge Bridge Bridge Bridge Bridge EBS Chassis PCI Express x4 Link IO Module 0 Type1 GB Switch SVP 0 GB Switch SVP PXH IO Module 1 Type 2 NDC CPU CPU Mem 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 PXH PXH PXH PXH PXH PXH PXH...

Page 20: ...er Blades provide SMP and the raw performance that the Itanium processors bring to number crunching jobs the Intel Xeon Server Blades which require less power and cooling and are less expensive are ideal for supporting infrastructure and application workloads as well as 32 bit applications The components of the Intel Xeon Server Blade are listed in Table 4 Table 4 Intel Xeon Server Blade component...

Page 21: ...leration Technology I OAT hardware and software supported I O acceleration that improves data throughput Unlike NIC centric solutions such as TCP Offload Engine I OAT is a platform level solution that addresses packet and payload processing bottlenecks by implementing parallel processing of header and payload It increases CPU efficiency and delivers data to and from applications faster with improv...

Page 22: ...ng applications Other features include Fully Buffered DIMM FBDIMM technology that increases memory speed to 800 MHz and significantly improves data throughput Memory mirroring and sparing designed to predict a failing DIMM and copy the data to a spare memory DIMM increasing server availability and uptime Support for up to 128 GB memory Enhanced Intel SpeedStep technology allows the system to dynam...

Page 23: ... to prevent system downtime caused by a memory fault BladeSymphony 1000 supports the online spare memory function in the ten patterns of memory configurations listed in Table 5 The shaded sections represent spare banks Online spare memory excludes the use of the memory mirroring function Table 5 Online spare memory supported configurations Bank Bank1 Bank2 Bank3 Bank4 Slot Slot 1 Slot 2 Slot 3 Slo...

Page 24: ... without the going down in case of a memory fault including a plural bits error Figure 15 Memory mirroring When operating in normal conditions data first writes in the primary slots 1 2 5 and 6 then in the mirror slots 3 4 7 and 8 The arrows in Figure 15 show the relationship between the mirroring source and the destination When data is read out it is read out of either the primary or mirror No me...

Page 25: ...on error correction is ideal for business critical applications running on BladeSymphony 1000 systems Traditional SCSI devices share a common bus At higher signaling rates parallel SCSI introduces clock skew and signal degradation Serial Attached SCSI SAS solves these problems with a point to point architecture where all storage devices connect directly to a SAS port Point to point links increase ...

Page 26: ...CI X I O Module PCIe I O Module and an Embedded Fibre Channel Switch Module Two I O modules are supported per chassis PCI X I O Module The PCI X I O Module supports eight PCI X cards in total with a maximum of two PCI X cards assigned to a single server blade for Chassis A and four for Chassis B In Chassis A eight PCI cards can be attached to four server blades at a two to one ratio In Chassis B f...

Page 27: ...lock diagram is shown in Figure 17 The card includes two1 2 4 Gb sec FC ports supporting FC AL and point to point switch fabric Two gigabit Ethernet ports are also included These ports support auto negotiation and VLAN compatible to IEEE 8 2 1Q and a maximum of 4096 TagVLANs Figure 17 PCIe I O Combo Card block diagram Embedded Fibre Channel Switch Module The Embedded Fibre Channel Switch Module co...

Page 28: ...to external storage Figure 19 depicts the back view of the module and a blow up of the Fibre Channel switch The block diagram for the module is shown in Figure 20 Figure 19 Back view of Embedded Fibre Channel Switch Module with blow up of the Fibre Channel switch 2 RJ 45 connector 1 Serial Port 3 Error LED Fiber channel switch close up SFP SFP SFP SFP RJ45 10 Fibre Channel port status LED green or...

Page 29: ...r benefit is reduced latency on the data path This dramatically reduces complexity administration and points of failure in FC environments It also reduces the effort to install and or reconfigure the storage infrastructure 48V 12V Glacier 5V 12V 3 3V 12V main 5V Standby FC SW Processor CPLD else SFP SFP SFP SFP RJ45 For management RS232C Connector I2C Hub I2C Reg I2C UART I2C Local Data Bus Total ...

Page 30: ...port with the function U_port to self detect port type Switch expandability Full fabric architecture configured by up to 239 switches Interoperability SilkWorm II SilkWorm Express and SilkWorm 2000 families Performance 4 250 Gb sec full duplex Server blade 0 Slot 0 Server blade 0 Slot 1 Server blade 1 Slot 2 Server blade 2 Slot 3 Server blade 3 Server blade 1 Server blade 2 Server blade 3 Server b...

Page 31: ...er FC HBA 1 port Fabric delay time Less than 2 microseconds no contention cut through routing Maximum frame size 2112 byte payload Service class Class 2 class 3 class F frame between two switches Data traffic type Unicast multicast broadcast Media type SFP Small Form Factor Pluggable Fabric service SNS Simple Name Server RSCN Registered State Change Notification Alias Server Multicast Brocade Adva...

Page 32: ...ade Fabric Watch SAN monitor for the switches made by Brocade It constantly monitors the SAN fabric to which the switch is connected detects any possible fault and gives the network manager a prior warning automatically Brocade ISL Trunking Groups ISLs between switches automatically to optimize the performance of the SAN fabric The Fibre Channel HBA supports Common HBA API version 1 0 partly 2 0 d...

Page 33: ... The switch provides up to 24 Gb sec total throughput performance and the ability to relay packets at 1 488 000 packets sec Additional features are listed in Table 9 Figure 23 Back view of chassis with blow up of Embedded Gigabit Ethernet Switch The Embedded Gigabit Ethernet Switch can be configured for high availability and fault tolerance when a second redundant switch module is added A single s...

Page 34: ...4 to 7 in Chassis B Disk drives installed in HDD modules can be hot swapped in a RAID configuration with the RAID controller installed on the PCI card The HDD Modules support RAID 1 5 and 0 1 and spare disk Figure 24 HDD Modules Table 9 Embedded Gigabit Ethernet Switch features Item Description Port Backplane side 1 Gb sec x 8 External 10 BASE T 100 BASE T 1000 BASE t auto connection Auto learning...

Page 35: ...lled through the backplane The logical numbers of the SCSI connectors on I O module 0 and 1 are defined as 0 to 1 and 2 to 3 respectively Figure 25 Connection configuration for HDD Modules Bridge Bridge Bridge Bridge SCSI Bus Internally connected Sample Configuration SCSI SCSI SCSI SCSI 6x HDD SCSI i f SCSI 2 SCSI 3 Server chassis 6 x HDD has only one SCSI I F port Note that this port is not conne...

Page 36: ...ovides connections between each server blade slot and two slots in a PCI module or Embedded Fibre Channel Switch Module Chassis B provides four connections from server blade slots 1 to 4 to two slots in a PCI module or Embedded Fibre Channel Switch Module The connections for both chassis types are illustrated in Figure 26 The specifications for each chassis type are listed in Table 10 Figure 26 Ch...

Page 37: ...e system Up to four Power Modules are installable in a chassis They are installed redundantly and support hot swapping The service processor SVP checks the power capacity when it starts up If SVD detects redundant power capacity it boots the system in the normal way If SVP cannot detect Switch Management Module 1 standard 2 maximum 1 standard 2 maximum I O Module PCI X 2 maximum 2 slots maximum pe...

Page 38: ...ed in Figure 27 The fans cool the system by pulling air from the front of the chassis to the back The modules can be hot plugged enabling a failed Cooling Fan Module to be replaced without disrupting system operations Cooling Fan Modules support the following functions Control rotation Detect abnormal rotation Indicate the faulty location with LEDs Built in fuse Figure 27 Top view and cooling fan ...

Page 39: ...r failed components are replaced The BladeSymphony 1000 is designed with features to help ensure the system does not crash due to a failure and to minimize the effects from a failure These features are listed in Table 11 Table 11 Reliability features Function Feature Quickly detect diagnose failed part BIOS self diagnostic function Memory scrubbing function Intel Itanium Server Blade Failure recov...

Page 40: ...le 12 defines the components of the Switch Management Module Two Switch Management Modules can be installed on one chassis In this case the main SVP normally performs the SVP function Health checking works between the main and sub SVP as they monitor each other If the main SVP fails the sub SVP takes over operation The Switch Management Module houses the gigabit Ethernet switch to which the gigabi...

Page 41: ...x 1 SDRAM 128 MB ECC Flash ROM 16 MB Stores the OS image NV SRAM 1 MB Battery backed up SRAM Saves system configuration information and fault logs Compact flash 128 MB 512 MB Backs up the SAL and BMC firmware Ether switch Broadcom 2 Connect the management LAN of each server blade Gigabit Ether switch Broadcom 1 PHY chips for External ports 10BASE T 100BASETx 1000BASE T Auto Negotia tion Auto MDIX ...

Page 42: ...gure 29 Fast Ethernet and I2C connections in SVP management interface Base Management Controller BMC One instance of BMC is installed for each CPU module primarily to take charge of management within the physical partition including a single CPU module Only one instance of SVP is active throughout the system managing the entire system in cooperation with BMC SVP and BMC communicate with each other...

Page 43: ...nium Server Blade the OS console and SVP console can share one communication pathway The OS console and SVP console are bound to local sessions serial communications via MAINT COM connection or remote sessions telnet session on MAINT LAN with the Console Manager which controls binds between the console and the session Console Manager is a software entity that is run on SVP and undertakes binding o...

Page 44: ...ormally using the EFI tool Under study for the IA32 CPU module Setting of the SVP clock Debugging of the system Setting of IP address Remote Console When running Linux or Windows a graphical console is available as the OS or System Firmware console The graphical console consists of VGA a keyboard and a mouse In the Intel Itanium Server Blade a Windows remote desktop is used A keyboard and mouse ca...

Page 45: ...workload requirements shift This provides an unheard of level of flexibility for accommodating spikes in demand for specific application services BladeSymphony Management Suite BladeSymphony 1000 can be configured to operate across multiple chassis and racks and this extended system can be managed centrally with BladeSymphony Management Suite software shown in Figure 30 BladeSymphony Management Su...

Page 46: ...rkload to be shared among the working servers Operations Management Reports real time or historical reports can be generated Report interval and display time intervals hour day week month year can be specified Graphical display drill down is possible for detailed analysis An export function supports output of HTML or CSV files Fault and error detection and notification when a problem occurs on a s...

Page 47: ...ement The BladeSymphony Management Suite Network Management function provides one point management of network switch VLAN configuration information Configuration information related to VLANs is obtained from the switches on the network and then managed from one port The management GUI can be used to set up and manage configuration information regardless of the command specifications for each type ...

Page 48: ...antage over host emulation virtualization offerings because guest operating systems can be simply and directly executed on virtualized environment without host intervention Virtage leverages Intel s Virtualization Technology VT to help ensure that processor performance is optimized for the virtual environment and to provide a stable platform that incorporates virtualization into the hardware layer...

Page 49: ...r advantage is the ability to dynamically change the services ratio for any given partition The system monitors the activity of a partition and if one partition is idle while the other is using 100 percent of its share the system temporarily increases the service rate until CPU resources are required by the other partition High I O Performance When deployed on Itanium processor based server blades...

Page 50: ...ared physical NIC Virtage enables multiple VNICs assigned to LPARs to share a physical NIC This function takes full advantage of the connections between VNICs and external physical networks The physical NIC shared between VNICs is called a shared physical NIC in the virtualization feature Integrated System Management for Virtual Machines Hitachi provides secure and integrated system management cap...

Page 51: ...ver Blades within in the same chassis Capability of scaling up or out to offer up to two16 core SMP servers in a single chassis In addition Virtage embedded virtualization technology brings the performance and reliability of mainframe class virtualization to blade computing enabling Hitachi to offer the first true enterprise class blade server Virtage provides an alternative to third party softwar...

Page 52: ...valds in the United States and other countries Windows is the registered trademark of Microsoft Corporation in the United States and other countries Hitachi is a registered trademark of Hitachi Ltd and or its affiliates Blad eSymphony is a registered trademark of Hitachi Ltd in the United States Other trademarks service marks company names may be trademarks or registered trademarks of their respec...

Reviews: