2.5I/O Node Requirements (TR-1)
The following requirements are specific to the I/O Nodes (ION) and augment the general node requirements (Section 2.4) above. As defined in Section 2.1, ION do not contribute to the system peak performance. The CN data path for file IO to the LLNS supplied Lustre file system may be between the CN and the ION over the system interconnect where the IO operations are handled by the Lustre client and then over the Offeror provided SAN interfaces to the LLNS supplied SAN infrastructure to the LLNS supplied Lustre MDS and OSS. The CN IO data path for IP based communications to the LLNS supplied SAN based IP devices may be between the CN to the ION over the system interconnect where the IP packets are routed to Offeror’s provided SAN interfaces to the LLNS supplied IP based devices.
2.5.1ION Count (TR-1)
Offeror will configure the system with sufficient ION to meet the IO file system performance requirements in Section 2.9.1. If the delivered system does not meet these performance requirements, then Offeror will, at no additional expense to LLNS, add additional ION to meet these requirements.
2.5.2ION IO Configuration (TR-2)
The ION may incorporate one or more SAN interfaces with PCIe2 x8 or faster busses each with a single slot filled with single PCIe2 x8 or faster InfiniBand 4x QDR or faster interface with one 40 Gb/s SFP+ or smaller interface to short range (SR) multi-mode fiber (MMF) optics capable of driving optical cables of at least 40m length or PCIe2 x8 40 or 100 Gb/s IEE 802.3ba compliant Ethernet interface with one SFP+ or smaller interface to short range (SR) multi-mode fiber (MMF) optics capable of driving optical cables of at least 40m length. The Sequoia ION is the data path from the CN to the SAN based resources (Lustre MDS and OSS, external 1 and 10 Gb/s Ethernet networks). The Offeror’s proposed ION configuration may carefully balance the delivered ION Interconnect bandwidth with the delivered PCIe2 x8 bus/slot bandwidth and with the delivered SAN network card bandwidth. In addition, the Offeror’s proposed ION configuration may carefully balance these delivered IO rates with the delivered integer processing performance and delivered memory bandwidth of the ION.
2.5.3ION Delivered Performance (TR-2)
The proposed ION PCIe2 x8 slots may deliver 3.6 (3.6+0 or 0+3.6) GB/s uni-directional and 7.2 (3.6+3.6) GB/s bi-directional bandwidth. The proposed InfiniBand or Ethernet interface may deliver at least 90% LNET Self Test with 1MB transfers. These interfaces may deliver an aggregate of 90% of bidirectional peak to TTCP benchmark. ION configurations with more than one PCIe2 x8 bus/slot and populated network interfaces may deliver an aggregate bandwidth of at least 90% of the delivered individual bus/interface bandwidth times the number of populated network interfaces.
Example: An ION with two PCIe2 x8 buses and InfiniBand cards may deliver at least 0.9*2*0.90*7.2 = 11.67 GB/s bidirectional bandwidth with the LNET SelfTest with 1MB transfers. This ION would require at least 13 GB/s delivered interconnect bandwidth and sufficient memory bandwidth to route packets to/from the interconnect from/to the InfiniBand.
2.6Login Node Requirements (TR-1)
The following requirements are specific to the Login Nodes (LN) and augment or supersede the general node requirements (Section 2.4) above. As defined in Section 2.1, LN do not contribute to the system peak performance. The LN are one or more node(s) that provide the hardware necessary for end-users of the system to login to the system and perform normal user activities of code development, job launch, job management and data movement. Local disk on the Login node may be used for OS configuration and high speed local (/tmp and /var/tmp) temporary file systems to accelerate editing, compiling and loading large applications. The LN data path for both file IO to the Lustre file system, local RAID disk and external networking based IP devices (such as NFS servers and user login via ssh) are all over the SAN interface.
2.6.1LN Count (TR-1)
Offeror will configure the system with sufficient LN configured with external networking interfaces to meet the high speed external networking interfaces requirement in Section 2.3. The Offeror will configure the system with sufficient number of LN with appropriate memory and local disk resources to support 100 active users executing 50 simultaneous compiles. The LN may have a similar SAN interface as the ION defined in Section 2.5.
2.6.2LN Locally Mounted Disk and Multiple Boot (TR-1)
The LN may have sufficient disk resources in aggregate to allow the storage of: 1) at least 10 system software images for each type of node (if not managed on the SN); and 2) 50 TB of local disk for /tmp and /var/tmp. These disk resources may be packaged with the node (i.e., physically local) or packaged remotely, but locally mounted. Each system software image of each type of node may have sufficient disk space for operating system, code development tools and other system binaries, swap and local tmp, NFSv4 cache. The LN locally mounted disk may be configured with High Availability, High IOPS RAID 5 (or better) arrays of hard disks as specified in Sections 2.9.2 and 2.9.3.
The LN may have the capability to boot up to 10 different versions of the operating system and all associated software (i.e., ten completely separate and independent software releases or patch levels). Switching to a new boot device will be accomplished by the root user issuing commands at the shell prompt and will not require recabling any hardware.
2.6.3LN IO Configuration (TR-2)
The LN may have one or more interfaces with PCIe2 x8 or faster busses each with a single slot filled with single PCIe2 x8 or faster InfiniBand 4x QDR or faster interface with one 40 Gb/s SFP+ or smaller interface to short range (SR) multi-mode fiber (MMF) optics capable of driving optical cables of at least 40m length or PCIe2 x8 40 or 100 Gb/s IEE 802.3ba compliant Ethernet interface with one SFP+ or smaller interface to short range (SR) multi-mode fiber (MMF) optics capable of driving optical cables of at least 40m length.
Offeror’s proposed LN configuration may carefully balance the delivered LN local iSER RAID file system bandwidth with the delivered PCIe2 x8 bus/slot bandwidth and with the delivered SAN network card bandwidth. In addition, Offeror’s proposed LN configuration may carefully balance these delivered IO rates with the delivered integer processing performance and delivered memory bandwidth of the LN.
2.6.4LN Delivered Performance (TR-2)
The proposed LN PCIe2 x8 slots may deliver 3.6 (3.6+0 or 0+3.6) GB/s uni-directional and 7.2 (3.6+3.6) GB/s bi-directional bandwidth. The proposed InfiniBand or Ethernet interface may deliver at least 90% LNET Self Test with 1MB transfers. These interfaces may deliver an aggregate of 90% of bidirectional peak to TTCP benchmark. LN configurations with more than one PCIe2 x8 bus/slot and populated network interfaces may deliver an aggregate bandwidth of at least 90% of the delivered individual bus/interface bandwidth times the number of populated network interfaces. These interfaces utilizing iSER or iSCSI may deliver an aggregate of at least 1,024 4KiB transactions per second with 50% read and 50% write balance.
Share with your friends: |