Attach_3_Technical_Scenario__2_16-223-SOL-00019.pdf

RFP# 16-223-SOL-00019

ATTACHMENT 3

Scientific Computing Technical Scenario #2

HPC Computing Environment Scenario

FDA’s High Performance Computing (HPC) environment has several very powerful clusters located in the DC metro area - each with between one and two thousand compute cores. The clusters use Centos 6 as an operating system (OS) and other support functions are preferred as open source. The clusters use Sun Grid Engine (SGE) or Univa Grid Engine (UGE) as dynamic resource managers (DRM). The scientific computation load on the clusters is dynamically balanced using the batch queuing system provided by the DRM to ensure that all users’ jobs receive the resources needed to complete them in a prompt manner and with the greatest efficiency, including in the ‘greenest’ way possible.

For financial and efficiency reasons, FDA is exploring the use of Cloud HPC resources to allow for burst demand jobs, and to moderate the amount of overall internal HPC resources required.

Define and describe in detail, a dynamic configuration model that accommodates the requirements within the constraints presented. Include in the description all system configuration processes required to effectively maintain the availability, reliability, performance, and security of the HPC resources.

Environment:

Authentication and access control are through Lightweight Directory Application

Protocol (LDAP).

Host configurations must be rigorously defined and enforced since the compute nodes are all diskless, and each must be approved by information assurance authority.

Applications, scientific or administrative, are configuration checked through a RPM package management system which allows clean install in the directory tree and equally clean uninstall.

All aspects of the clusters must be real-time monitored for load and capacity and log files managed and kept for annual reporting of usage by project, user and department. Other monitoring must be done to spot failed units hardware malfunctions and zombie processes along with potentially unauthorized processes.

Kerberos based authentication is required to manage access to all systems.

All operating systems are minimal to enhance performance so when applications are needed to run there must be an analysis of whether the available OS components are sufficient and whether the memory footprint will allow the program to run.

Only command line based administrator access is tolerated. No graphical interfaces of any kind are allowed.

Internet 2 connectivity, is available at up to 100 GB from data centers to Internet 2 partners.

Specific Cloud resources are not defined, use of IAAS is required to accommodate growing requirements.

All cloud HPC resources must be provided by FEDRAMP qualified resources.

RFP# 16-223-SOL-00019

ATTACHMENT 3

Scientific Computing Technical Scenario #2

Failover architecture is necessary to preserve the ongoing functions of the clusters when connection by network is lost.

Cluster functionality must be single point failure proof.

Datacenter skills are necessary for ensuring the correct cabling and positioning of equipment for efficiency and ease of maintenance.

Requirements:

The Offeror is required to respond to the following:

Prepare a Work Breakdown Schedule (WBS) that addresses activities and associated deliverables required to complete the scenario. Resource each activity by a specific labor category, and hours required to complete.

Identify labor categories including skill sets, associated experience and staffing levels

(level of effort) are required to execute each task.

Identify and describe detailed assumptions upon which the response to the scenario is based.

Identify potential technical, programmatic, and/or business issues and risks inherent in the scenario and provide approaches for their resolution or mitigation.

Discuss the leading service delivery and performance management practices, drivers, and metrics that are key to fulfilling performance measurements and monitoring the technical requirements of this scenario.

Identify potential security issues inherent in the scenario and provide approaches for their resolution or mitigation.

Discuss the security approach associated with integrating Internet 2 services and Cloud

Infrastructure as a Cloud (IAAS) services.

Discuss the security approach to monitoring inbound, outbound (data in transit) and in store data (data at rest) to ensure security levels of the system are maintained. Assume use of FISMA low is used.