In Partnership with: Application Optimised Hardware: David Power: Head of HPC MEW25 Scale-Out HPC Openstack Cloud Distributed Storage
In Partnership with:
Application Optimised Hardware:
David Power: Head of HPC
MEW25
Scale-Out HPC Openstack Cloud Distributed Storage
In Partnership with:
Who are Boston?
• Founded in 1992 – 22 years of innovation
• Supermicro’s longest established & largest global partner – for 20 years
• Technology focused, trade-only global distributor & solution manufacturer
• Global HQ in London with branch offices in London, Munich, Mumbai, Bangalore and New York. Further global expansion is planned.
• Experts in solutions and HPC system design using premium
equipment for Supermicro and leading component partners
• Worldwide stockholding
• Route to market via authorised resellers and partners
• Awarded Supermicro ‘Fastest Growing Partner’ 2012/2013
In Partnership with:
The Best Server Products on the Market
FatTwin™
GPU/Xeon Phi-optimized HFT Solutions
Storage Solutions
MicroBlade
DCO Products
Product Families – X10 Haswell Launch on Sep 8th..
Twin Architecture
The original Twin architecture innovator. FatTwin™ & TwinPro™ and more
GPU/Xeon Phi Optimized Solutions
Full line of GPU/Xeon Phi optimized solutions, non-blocking airflow for 8 cards in 4U И
HFT & Workstation
Hyper Speed solutions optimized for High Frequency Trading (HFT) И
Ultra and DCO Products Ultra architecture is easily upgradable with expandable I/O Options, DCO supports higher thermal envelope and more power efficient И
MicroCloud High density , Space saving. 8/12/24 UP Server nodes in 3U. Avoton getting ready И
SuperBlade & MicroBlade™ New GPU/Xeon Phi blade solutions. MicroBlade 112 Atom/28 Xeon nodes in 6U И
IPC and Embedded Long Life Cycle, high efficiency, compact form factor and low power solutions И
Switch Products Top of the Rack - Layer 2, Layer 3, 48 Port 10G SFP+, Software-defined-networking И
Storage Solutions Storage for Enterprise & Big Data: Double-sided 72/90, Hadoop, SBB, Cold Storage И
System Management Software SSM, Remote Management IPMI 2.0, SPM, SUM, SuperDoctor 5 И
TwinPro™
3
NEW!
32”
NEW! NEW!
Ultra Architecture
In Partnership with:
Microblade
Atom Microblade: - 4 servers per blade - 1x C2750 (8 cores, 2.4Ghz) - 2x SODIMM slots (32GB) - 1x 2.5” SATA3 - 1x SATADOM - 2x 2.5Gbps eth ports - Intel Virtualization Technology - Best performance per watt - 112x Atom C2000 nodes in 6U - 4x GbP switches (40G/10G uplinks)
In Partnership with:
Microblade
Atom Microblade (14w): - 4 servers per blade - 1x C2550 (4 cores, 2.4Ghz) - 2x SODIMM slots (32GB) - 1x 2.5” SATA3 - 1x SATADOM - 2x 2.5Gbps eth ports - Intel Virtualization Technology - Best performance per watt - 112x Atom C2000 nodes in 6U - 4x GbP switches (40G/10G uplinks)
In Partnership with:
Microblade
Intel UP Microblade: - 1 server per blade - 1x Intel E3-1200v3 (up to 84w) - 4x VLP DDR3 ECC UDIMM (up to 1600 MT/s) - 2x 3.5” 6Gb/s SATA3 - or 4x 2.5” 6Gb/s SATA3 - Onboard dual-port 1 GbE NIC - Onboard BMC for IPMI 2.0 support - 196x Intel E3 nodes in 42U - 4x GbP switches (40G/10G uplinks)
In Partnership with:
Microblade
Intel DP Microblade: - 1 server per blade - 2x Intel E5-2600 v3 (up to 120w) - 8x VLP DDR4 RDIMM (up to 2133 MT/s) - 2x 3.5” 6Gb/s SATA3 - Onboard dual-port 1 GbE NIC - Or Onboard dual-port 10 GbE NIC - Onboard BMC for IPMI 2.0 support - 196x Intel E5 DP nodes in 42U - Industry leading density for DP!
In Partnership with:
OCP Solutions • Coming soon (Deploying with key customers) • Rack, Compute and Storage solutions • Disruptive approach to designing scalable and
efficient datacenters • Testing in the lab with openstack and CEPH. HPC
Benchmarks to follow soon • Significant power savings on DC
In Partnership with:
Converged Cloud Solutions (vScaler)
• Scalable Software Defined Data Centre • Software defined building block • Converged storage and compute platforms • Scale IO and Compute in one modular • Highly resilient by design • Start small and grow with demand
In Partnership with:
• 1-Click install of CloudX software – Highest cloud performance
• Includes all CloudX ingredients – O/S installation ( RHEL, CentOS) – OpenStack components ( currently based on RDO) – Mellanox technology configuration ( Ethernet or InfiniBand )
• With nodes provisioned and configured, CloudX then initiates OpenStack cloud, ready to run
One-Click Cloud
Limitless Cloud Performance for the Applications of Tomorrow
CloudX One Click Install Platform
In Partnership with:
UltraDIMM – Sandisk & Diablo
200/400GB 19nm MLC NAND Flash
• Implementation of MCS™ technology
• Enables data and control communication for storage
devices over memory bus
• Full SSD is memory form factor
• Same power/thermal envelope as a normal DIMM (or better)
• Negates the need for 3rd party busses, interconnect and protocols
• Compatible with most motherboards with a small BIOS/UEFI
update (initial support on Supermicro X9DRT, X9DRX and X9DRH)
• Software support for Windows, Linux and Vmware.
In Partnership with:
Optimised Platforms
• Partnership signed with Inktank/Redhat to provide commercial support on CEPH
• Validated CEPH Reference Architecture
• Mon Node: SYS-5017R-MTRF
• OSD Node(s): 12 Bay, 36 Bay and 72 Bay
• Huge increase in interest for this technology around Openstack delpoyments
• Validated on ARM/Viridis
12 Bay Ref Arch 12 Bay Ref Arch
12x 3.5 Bays for OSD 12x 3.5 Bays for OSD
2x 2.5 SSD Journals 2x 2.5 SSD Journals
36 Bay 36 Bay 72 Bay 72 Bay
In Partnership with:
dataScaler-SDS Range
Archive: 144TB-288TB
Archive: 144TB-288TB
General: 48TB General: 48TB HPC: 14.4TB HPC: 14.4TB 10GB/40GB/56GB: IB or Eth
10GB/40GB/56GB: IB or Eth
DELIVERING AGILITY AND COST ADVANTAGE
FOUNDATION FOR HYBRID CLOUD AND BIG DATA
DEPLOY DATA ANYWHERE PHYSICAL, VIRTUAL CLOUD
ELASTIC CAPACITY AND PERFORMACE
MODERN, SECURE FILE AND OBJECT STORAGE
CONVERGE COMPUTE AND STORAGE
dataScaler-SDS
MANAGING UNSTRUCTURED DATA AT SCALE
ENSURE DATA PROTECTION AND AVAILABILITY
In Partnership with:
IEEL Appliance • Validated solutions for IEEL • HA Configured (MDS and OSS) • Based on Lustre 2.5 • Infrastructure for policy-driven tiered storage solutions (HSM) • Simple but powerful UI lowers management complexity • Intuitive heat map chart options integrated with Lustre “jobstats” to
easily identify performance constraints • Near real-time monitoring storage health and performance • Generate historical charts and reports • Advanced troubleshooting tools (consolidated log files, intelligent
scanning of logs, configurable event notification) • Cost optimised solutions for smaller deployments (100TB – 1PB)
In Partnership with:
BeeGFS Parallel Storage
• BeeGFS on-demand for cluster FS creation in one command • Built in benchmark tools • BeeGFS sits on top of an existing local filesystem using the normal POSIX
interface • Servers can be added to an existing system without downtime. • Flexible striping on a per file or per directory basis • Distributed Metadata • Very high performance (3.5GB/s write and 4GB/s read on single stream / FDR) • MD Perf: linear scalability through dynamic metadata namespace partitioning
In Partnership with:
GPU K80 Test Drive! Bostonlabs provide remote access for users on K80
Cluster access – test MPI scaling across FDR IB
Remote testing facility for benchmarking K40 and K20 GPU Nodes
Intel Xeon E5-2600v3 Cluster with FDR IB
CUDA 6.5 and fully build toolkit / SDK
HPC Support Engineers Experts from various scientific fields on hand to assist with any problems
Blog opportunity Report findings to the community
http://www.bostonlabs.co.uk