SAS Grid at HC and PHAC June 12, 2014

SAS Grid at HC and PHACJune 12, 2014

Agenda

To Grid or Not To GridThe ApproachThe MetricsLessons LearnedLooking Forward

To Grid or Not to Grid: The Problem

Multiple versions of SAS Multiple servers with multiple OS versions Multiple data stores and multiple types Shrinking resource pools Shrinking budgets A new service provider

The Heterogeneous SAS Environment

This is not unique to SAS… it is everywhere!

A Technology Labyrinth

Levelling The Playing Field

Why SAS Grid?

One standard for HC and PHAC. Local data storage. Limited data traffic across WAN/LAN. Dramatic increase in performance. Reduced licensing cost (Single contract). Reduced Single points of failure. Complete solution for Analytics/Reporting. Scalability, Reproducibility and Flexibility. Security, Auditing and Monitoring

Levelling The Playing Field

What Really Happens

Just Kidding !

Current SAS Grid Architecture at HC and PHAC (Simple view)

The Approach

• Leverage Existing Hardware Investments HP BLc7000 Blade Enclosures HP Proliant BL460c G6 Blades HP EVA 8000 SAN

• Promote to Production Environment HP BLc7000 Blade EnclosuresHP Proliant BL460c G7 Blades3PAR T800 Storage System

Infrastructure

The Approach

The Security Flow

SAMBA Server

METADATA server

Grid Nodes

LDAP Server

SAS User

DesktopMid-Tier

The Approach

The Security LayersSAMBA Server METADATA server NodesCorporate LDAPSAS User

Desktop

Mid-Tier

Unique UID Membership in groups Corporate Security Policies

Department Wide

Centrally Managed

Linux file system ACLs LDAP and Metadata are “synced” UID and GID are unique Linux ALCs use exact GIDs and UIDs

The Metrics: SAS EG 5.1 Write

The Metrics: SAS EG 5.1 Read

The Metrics: The Best of Both Thanks to Tom Kari for providing his test scripts

The Metrics: Teky Stuff

• HP BLc7000 Enclosure Easily scalable – 16 device bays In house expertise Server/Storage/Back-up in same data centre Keep data traffic off Corporate Ethernet Back-bone

• HP Proliant BL460c G7 Blade 2x Intel Xeon 2.66 GHz (8 cores) 100 Gb DDR3 Memory (Max 384 Gb) 2x 8 Gb Fiber Chanel HBAs (Host Bus Adapter) 2x 10Gb Ethernet Adapters 2x 300Gb Internal Drives

Technology Choices: Hardware

The Metrics: Teky Stuff

• 3PAR T800 40 TB of disk space (Expandable to 800 TB Raw) Capable of offering 3-tiered storage

• SSD• FC• SATA

Has been recently discontinued by HP• CommVault Back-Up solution

Existing solution

Technology Choices: Hardware Continued

The Metrics: Teky Stuff

The View from Behind

Courtesy of HP: www.hp.com

The Metrics: Teky Stuff

• OS - SuSe Linux Version 11 Base image identical for all servers

• Simple to image new server Additional packages required

• Metadata server• Application server• Mid-Tier server• SAMBA server (Data access)

Grid nodes are all identical

Technology Choices: Software

The Metrics: Teky Stuff

• Software Specifics OCFS2 clustering software

• All Servers can mount SAS binary and configuration User home directories Data directories SAS GRIDWORK

Technology Choices: Software

The Metrics: Teky Stuff

• Software Specifics• Multipath technology

Provides 4 paths for each server to 3PAR storage

Technology Choices: Software

The Metrics: Teky Stuff

• A few specifics Grid Master is “Closed”

2 Queues created “Normal” and “Priority”

Technology Choices: Configuration

Lesson’s Learned

• The core issue Bandwidth, Bandwidth, Bandwidth…

• Isolate Data Traffic Avoid data movement across corporate Ethernet backbone Encourage “local” data access

• User engagement Critical to bring the end user into the process from the start Understand how SAS is being used. The end state Change is not easily adopted.

Lesson’s Learned

• The Linux vs Windows issue Legacy issues continue to be a concern Use of SAS PC File Server foreign to some

• Phased approach to user migration Enable small groups vs entire organization at once Provide a mechanism for user feedback/problems Provide detailed documents –That nobody reads

• Monitor the SAS Grid proactively Identify key difficulties for clients

Login, data access, data sharing, desktop configurations, library and coding standards

Contact clients before the issues drives them nuts.

Forward Looking

• Collaboration• Several Federal Department and Agencies are

engaged and committed to the Grid technologies• We have the ability to define the road map ahead

• What works best• Talk and exchange successes (and issues) and map

them to your own business needs• Resist the silo• Facilitate change• We no longer live in isolation,

Forward Looking

The BIG DATA vision

A Visitor From The Future

Thank You!

Credit for Our Success

Health Canada Shared Service Canada

Camille Madeira Satnam Girn Jeffrey Sarazin Spawn Gusdal Mario Romeo Charles Lapansee Ian Colby Blair DeanJames Moore Roz MathewsJeffrey Menzies Kurt Combden(CBSA)

SAS Canada Inc. (David Eskenazi, Matej Pivac)

All of our SAS Users at Health Canada and Public Health Agency of Canada