High performance computing partnership
Advancements in computing systems have made it possible to tackle huge obstacles in simulation from product design, oil field exploration, financial risk assessment, genome mapping to weather forecasting and essential academic research.
According to the Top500.org, Lenovo is the number-one supercomputer provider. 1 in 3 supercomputers runs Lenovo. With industry leading HPC experts, OCF in partnership with Lenovo take a customer-centric, consultative approach to provide HPC and AI solutions that help tackle the data intensive challenges of today and tomorrow.
OCF is a Lenovo Platinum Partner. The partnership began in 2005, when Lenovo acquired the IBM PC business. The longstanding collaboration has delivered many transformational HPC projects at leading academic and research institutions.
As a Lenovo partner, OCF is able to provide to its customers:
The New 1U server is much more powerful than the previous SR630 V1 Gen 2 model. It offers a much higher computing performance with new 3rd generation Intel Xeon Scalable processors, faster and increased memory capacity, PCI 4.0 support and high bandwidth network cards. OCF's recommends this server in its solutions for key roles like 1U compute cluster’s, management, and login nodes.
Lenovo ThinkSystem SR650 V2 features many of the same components as the SR630 V2 but in a 2U form factor. The larger chassis allows for more internal drives, but just as important to HPC, the 2U chassis allows for eight PCIe / one OCP cards to be installed. OCF use these as parallel filesystem file servers as the large number of PCIe slots allows ample bandwidth to connect to high speed back-end storage, as well as to cluster interconnects to efficiently serve compute clients and high speed LANs to serve user workstations. The good balance between internal storage and compute resource, also makes this ideal for compute intensive workloads, high performance software defined storage deployments, supporting a range of drive types from high-capacity SATA to low latency NVMe, and even storage class memory.
As a quad-CPU server this is a natural fit for large centralised transactional database workloads however, it is also a fit for high performance computing.
Four CPUs allows for not only a high quantity of cores but also a high capacity of memory to be installed. The SR850 V2 builds upon this but as a meshed design, allowing intel’s 3rd Gen Processors now with six UPI links ensures increased CPU I/O, fully interconnected with no additional hops. This greatly increases the ability to share memory between CPU’s.
The new 3rd Gen CPU’s supports the new bfloat16 instruction set for Enhanced Deep Learning Boost.
With 7 PCIe 3.0 expansion Slots and one OCP Slot there is plenty of bandwidth to add High speed connects and external storage. Four front drive bays support up to 24 SAS, SATA, NVMe Drives all in a 2U Server.
A 4U quad-CPU server with up to 112 cores, 224 threads using Intel’s 3rd Gen Processors. Suitable for deep learning, analytics, big data, virtual machine density.
Support for double wide GPU’s Nvidia A100, V100S and T4.
Plenty of I/O with 14 PCIe 30 Slots plus OCP 3.0 Slot reserved for an ethernet adapter. Large storage capacity, support for up to 48 x 2.5” 24 of which can be NVMe. 2 x Hot Swap 7mm NVMe Drives supported to the rear of the chassis, M2 boot supported internally.
The Lenovo ThinkSystem SR635 is a storage-rich server taking advantage of AMD’s high throughput EPYC processors. Up to 20 high capacity 3.5” drives can be installed making this server a natural fit for software defined storage solutions such as object store, where we must balance keeping data readily available and cost effectiveness. With up to 64 compute cores this server is also a good fit for big data analytics such as Hadoop workloads.
A storage-rich server taking advantage of AMD’s high throughput EPYC processors. Up to 20 high capacity 3.5” drives can be installed making this server a natural fit for software defined storage solutions such as object store, where we must balance keeping data readily available and cost effectiveness. With up to 64 compute cores this server is also a good fit for big data analytics such as Hadoop workloads.
The Lenovo ThinkSystem SD530 is an ultra-dense compute offering packing four dual-CPU nodes into a 2U chassis. This platform stands out amongst competitors due to two key features, excellent thermal efficiency and shared IO. The thermal efficiency provided by shared fans in the chassis allows high TDP processors to be used up to 2x 205W per node. As these high TDP processors tend to offer the highest compute efficiency it allows us to deliver the most compute throughput in a dense form factor. Supported Shared-IO cards allow pairs of nodes to share a high-speed interconnect connection. This helps drive cost efficiency through using fewer switches ideal for high throughput applications not requiring a dedicated highspeed connection per node.
The Lenovo SD630 V2 is a dule socket 0.5u rack Intel Whitley Ultra Dense Server. server up to 4-Node’s are supported in a 2U DA240 rack enclosure.
OCF recommends this server for AI, high performance compute and analytic workloads when a dense form factor is required. It is an ideal scale out platform designed to deliver maximum performance in a small footprint.
Two dedicated network ports connected to an onboard Mellanox ConnectX-4 Lx controller. Supports 25,10,1GB with supported optical or copper Connections. drive cage support for two 2.5” 7mm SATA/NVMe or one 2.5” NVMe Drive’s. Two M.2 drives for boot may be installed with the M.2 enablement kit that connects to a dedicated slot on the system board.
Lenovo ThinkSystem SR670 V2 features a modular design suitable for Virtualization, rendering, Compute intensive HPC and AI. Supports HGX A100 4 x GPU Board or up to eight GPUs using PCIe 4.0 x 16 slots dependant on front shuttle choice. Support for NVlink between adjacent GPU pairs, all in a 3U Server. Full ASHRAE A2 support with internal fans and Lenovo Neptune™ liquid-to-air hybrid cooling with NVIDIA HGX™ A100. Eight drive slots in the front of the server SAS,SATA or NVMe.
OCF, together with Lenovo, has built a solution capable of growing with your organisation’s needs and scaling in line with your deep learning ambitions. By combining Lenovo systems and NVIDIA accelerators together with LiCO software and OCF integration services, you get the perfect building blocks for your AI project.
Lenovo ThinkSystem DE2000H is a low-cost, hybrid entry-level storage system that is designed to provide performance, simplicity, capacity, security, and high availability for small to large businesses. It delivers enterprise-class storage management capabilities with a wide choice of host connectivity options, flexible drive configurations, and enhanced data management features. The ThinkSystem DE2000H is a perfect fit for small offices and remote and branch offices of large enterprises that run a wide range of workloads, including big data and analytics, video surveillance, technical computing, backup and recovery, and other storage I/O-intensive applications.
ThinkSystem DE4000 & DE6000 Series, billed as entry level storage devices, these systems feature full stripe write acceleration meaning that despite an otherwise modest feature set, these systems have blisteringly fast RAID-6 8+2P performance. OCF takes advantage of this in HPC storage solutions to build a single fast parallel filesystem harnessing the performance of multiple arrays.
These systems are available in 2U and dense 4U configurations supporting mixtures of SSDs and HDDs.
The DSS is formed of pairs of x86 servers resiliently connected to JBODs and uses IBM’s distributed Native RAID to minimise the impact of long disk rebuilds associated with traditional hardware RAID.
The servers act as a high availability pair taking on the roles of storage controllers as well as managing a high performance parallel file system. They remove the bottle necks imposed by traditional hardware raid controllers and allow an incredibly high degree of concurrent access to both many files and even a single file.
The DSS-G is a natural fit for high performance computing allowing many servers to access data simultaneously ensuring compute workloads complete effectively.
LiCO software from Lenovo provides a graphical environment to simplify the use of clustered compute resources for AI development. LiCO provides a single interface to deploy, monitor and manage AI workloads, abstracting the complexity of the cluster from users.
Optimized for the Lenovo ThinkSystem portfolio with the latest Intel® Scalable® processors and NVIDIA® datacenter accelerators, LiCO enables the infrastructure to be used more effectively, flexibly and efficiently for AI development.
Data science is a major initiative in both research institutions and enterprise, and with versions to support running AI workloads on either HPC or Kubernetes clusters, LiCO allows the same ease-of-use access and infrastructure efficiency in the environment that best fits your organization’s needs.
LiCO provides clients the flexibility to deploy a dedicated cluster for AI, converge AI workloads onto an HPC cluster, or leverage a Kubernetes cluster to consolidate application development into a single infrastructure environment.