Flux: The State of the Cluster

Similar documents
High-Performance Computing at The University of Michigan College of Engineering

Introduction to High Performance Computing and an Statistical Genetics Application on the Janus Supercomputer. Purpose

LBRN - HPC systems : CCT, LSU

Brutus. Above and beyond Hreidar and Gonzales

BRC HPC Services/Savio

OpenAFS A HPC filesystem? Rich Sudlow Center for Research Computing University of Notre Dame

HPC File Systems and Storage. Irena Johnson University of Notre Dame Center for Research Computing

The RAMDISK Storage Accelerator

Installation guide for DataCove Archiver Virtual Machines Hypervisor type: Hyper-V 2012 R2 December 20, 2015 Revision 1.0

The University of Michigan Center for Advanced Computing

sciebo - die Campus Cloud

Service Description. IBM DB2 on Cloud. 1. Cloud Service. 1.1 IBM DB2 on Cloud Standard Small. 1.2 IBM DB2 on Cloud Standard Medium

INTEGRATING HPFS IN A CLOUD COMPUTING ENVIRONMENT

IBM IBM Tivoli Monitoring Express V6.1 Specialist. Download Full Version :

Web Hosting: Mason Home Page Server (Jiju) Service Level Agreement 2012

IBM Case Manager on Cloud

SCS Distributed File System Service Proposal

AN INTRODUCTION TO CLUSTER COMPUTING

IBM Terms of Use SaaS Specific Offering Terms. IBM DB2 on Cloud. 1. IBM SaaS. 2. Charge Metrics

HPC and IT Issues Session Agenda. Deployment of Simulation (Trends and Issues Impacting IT) Mapping HPC to Performance (Scaling, Technology Advances)

SERVERS TO SERVICES HOW MICROSOFT AZURE CAN MODERNISE YOUR IT INFRASTRUCTURE. Joey Lau 9 November 2017

Ekran System System Requirements and Performance Numbers

The BioHPC Nucleus Cluster & Future Developments

Intel Enterprise Edition Lustre (IEEL-2.3) [DNE-1 enabled] on Dell MD Storage

Extremely Fast Distributed Storage for Cloud Service Providers

BIG DATA STRATEGY FOR TODAY AND TOMORROW RYAN SAYRE, EMC ISILON CTO-AT-LARGE, EMEA

IBM System p5 550 and 550Q Express servers

Service Level Agreement Research Computing Environment and Managed Server Hosting

Pushing the Limits. ADSM Symposium Sheelagh Treweek September 1999 Oxford University Computing Services 1

Introduction: UNCC COE MOSAIC Page 1 of 9

3ROX Spring Meeting. Thursday, 20 April 2017 Pittsburgh, PA

Outline. March 5, 2012 CIRMMT - McGill University 2

IBM Content Manager OnDemand on Cloud

IOmark- VM. HP HP ConvergedSystem 242- HC StoreVirtual Test Report: VM- HC b Test Report Date: 27, April

Comet Virtualization Code & Design Sprint

Report Exec Dispatch System Specifications

Habanero Operating Committee. January

Sun Lustre Storage System Simplifying and Accelerating Lustre Deployments

2017 Resource Allocations Competition Results

Introduction. Kevin Miles. Paul Henderson. Rick Stillings. Essex Scales. Director of Research Support. Systems Engineer.

Datacenter Networking

PRICE LIST JULY 2018 SOFTWARE LICENSES

PRICE LIST FEBRUARY 2017 SOFTWARE LICENSES

IBM Terms of Use SaaS Specific Offering Terms. IBM DB2 on Cloud. 1. IBM SaaS. 2. Charge Metrics. 3. Charges and Billing

Parallel File Systems for HPC

Tutorial on Memory Management, Deadlock and Operating System Types

COMPARING COST MODELS - DETAILS

Arista 7320X: Q&A. Product Overview. 7320X: Q&A Document What are the 7320X series?

Organizational Update: December 2015

Red Hat enterprise virtualization 3.0

p5 520 server Robust entry system designed for the on demand world Highlights

VM Migration Acceleration over 40GigE Meet SLA & Maximize ROI

Isilon Performance. Name

1 Copyright 2011, Oracle and/or its affiliates. All rights reserved. reserved. Insert Information Protection Policy Classification from Slide 8

Office 365. It s Past, Present and Future at UW

PART-I (B) (TECHNICAL SPECIFICATIONS & COMPLIANCE SHEET) Supply and installation of High Performance Computing System

Mellanox Virtual Modular Switch

1. ALMA Pipeline Cluster specification. 2. Compute processing node specification: $26K

SvSAN Data Sheet - StorMagic

RESEARCH DATA DEPOT AT PURDUE UNIVERSITY

Lustre overview and roadmap to Exascale computing

IBM dashdb for Analytics

SWAN Technology Users Group. Thursday March 30 th, 1pm Tinley Park Public Library Steven Schlewitt, SWAN IT Manager

Service Description. IBM DB2 on Cloud. 1. Cloud Service. 1.1 IBM DB2 on Cloud Standard Small. 1.2 IBM DB2 on Cloud Standard Medium

NCAR Globally Accessible Data Environment (GLADE) Updated: 15 Feb 2017

LO2 Be able to design virtualisation deployments.

LONI Update. ULS CIO Meeting. Lonnie Leger, LONI / LSU

Introducing SUSE Enterprise Storage 5

SMB Direct Update. Tom Talpey and Greg Kramer Microsoft Storage Developer Conference. Microsoft Corporation. All Rights Reserved.

IBM Db2 Analytics Accelerator Version 7.1

OBTAINING AN ACCOUNT:

HPC at UZH: status and plans

The next step in Software-Defined Storage with Virtual SAN

HPE Datacenter Care for SAP and SAP HANA Datacenter Care Addendum

The Cambridge Bio-Medical-Cloud An OpenStack platform for medical analytics and biomedical research

NCAR s Data-Centric Supercomputing Environment Yellowstone. November 28, 2011 David L. Hart, CISL

IT Town Hall Meeting

Version 1.24 Installation Guide for On-Premise Uila Deployment Hyper-V

SurFS Product Description

ArcGIS Enterprise Performance and Scalability Best Practices. Andrew Sakowicz

IT Briefing. December 18, 2014 North Decatur Building 4 th Floor Auditorium

SuperMike-II Launch Workshop. System Overview and Allocations

Licensing & Pricing FAQ

Remote Endpoint Management and PCoIP Management Console Oct. 19, 2017

An ESS implementation in a Tier 1 HPC Centre

Introduction to High Performance Computing (HPC) Resources at GACRC

Red Hat enterprise virtualization 3.1 feature comparison

The Faculty of Arts and Sciences High Performance Computing Core

Data center: The center of possibility

Rethink Storage: The Next Generation Of Scale- Out NAS

Virtual Private Servers

INFOBrief. Dell-IBRIX Cluster File System Solution. Key Points

CS / Cloud Computing. Recitation 3 September 9 th & 11 th, 2014

Storage Supporting DOE Science

Parallel File Systems. John White Lawrence Berkeley National Lab

Installation Prerequisites

Version 1.26 Installation Guide for On-Premise Uila Deployment

NetApp: Solving I/O Challenges. Jeff Baxter February 2013

Optimizing Server Designs for Speed

We're Not CIS. The Technical Staff & Computing Facilities. User Community. What CIS Does for CS. CIS: Computing and Information Services.

Transcription:

Flux: The State of the Cluster Andrew Caird acaird@umich.edu 7 November 2012 Questions Thank you all for coming. Questions? Andy Caird (acaird@umich.edu, hpc-support@umich.edu)

Flux Since Last November Hardware added 4,000 cores of standard Flux nodes 2,016 cores (12 cores/node), 4GB RAM per core 1,984 cores (16 cores/node), 4GB RAM per core added 200 cores of larger-memory Flux nodes 40 cores/node, 25GB RAM per core added 342TB parallel filesystem: /scratch connected to compute nodes over InfiniBand peak performance of 44Gb/s (5.5GB/s) /scratch hardware Flux Growth Cores 0 2000 4000 6000 8000 10000 Allocated Cores Physical Cores Flux Since Last November Environment upgraded to the latest version of RedHat Linux RedHat Enterprise Linux 6.3 started requiring MTokens to log in two-factor authentication on IIA s advice Business Administration 2010 2011 2012 2013 the rate was increased from $11/core/month to $18/core/month historical Flux usage data is available in MReports https://mreports.umich.edu/mreports/ pages/flux.aspx M-Token

Flux Today: The Hardware 632 nodes providing 8,016 cores and 30TB RAM (4GB RAM/core) 5 nodes providing 200 cores and 5TB RAM (25GB RAM/core) 80GB home directories 324TB of scratch disk space 25Tb of network bandwidth (639 40Gb network connections) 24 Flux nodes: 288 cores, 1 1 TB RAM, 8 960Tbps bandwidth Flux Today: Growth Flux Growth Cores 0 2000 4000 6000 8000 10000 Allocated Cores Physical Cores 2010 2011 2012 2013

Flux Today: Utilization Flux Project Persistence 1 Apr 2010 17 Oct 2012 Active Flux Projects 0 50 100 150 200 Summed Total Summed Total (no classes) Active Renewed New 2011 2012 Flux Until Next November Hardware adding 2, 000 more cores to get to 10, 000 cores expanding /scratch adding 140 3TB disks for an additional 300 TB /scratch will be 300 disks and 640TB usable space performance will increase, as well as capacity Networking 20Gb Ethernet connection to U-M backbone if you have 10GbE storage, let us know we can add more 10GbE links upcoming network backbone upgrades will provide 100GbE

Flux Until Next November Environment no major OS updates minor OS updates, but no huge software library upgrades decreased resilience against loss of power moving Flux to Modular Data Center we expect this to be less expensive than the MACC, and will reflect that in the rate data center is 20% of the rate most of Flux will move at the end of December during the outage Modular Data Center Flux Until Next November Business Administration research software library for use by U-M faculty and students for publishable research there will likely be a user agreement reflecting this rate increase to $22/core/month this is the last big rate increase for planning purposes, expect 2 5% increase annually you should talk to your Research Dean about subsidy planning Paul Killey has two sessions tomorrow on Flux, for Research and other Academic Administrators at 9:15am and 2:30pm

Flux Operating Environment Federated Flux is an extension of Flux comprising hardware purchased by researchers and a subscription to the Flux Operating Environment (FOE). A subscription to the FOE provides all of the infrastructure and services that comprise Flux except the compute nodes The configuration of compute nodes added to the FOE is based on the most current configuration of nodes in Flux Hardware orders are aggregated and placed three times per year The rate for the FOE is $267 per node per month Web content is coming soon Questions Thank you all for coming. Questions? Andy Caird (acaird@umich.edu, hpc-support@umich.edu)