Les serveurs traitement continu

1
Les serveurs à traitement continu

• Stéphane Gerber
• 2005

2
Stratus Technologies
Stratus fournit des plates-formes à
disponibilité continue permettant une
simplicité dexploitation inégalée et offrant un
avantage financier convaincant à tous nos
clients
3
Stratus Technologies
MidOcean Partners
24 ans dexpérience dans la Disponibilité Continue
Availability Made Simple, Affordable

• CA 350 M
• 10000 Sys. installés dans le monde
• 400 Sys. exploités en France
• 7/10 serveurs monétiques en France

4
Besoins de disponibilité dans tous les secteurs
de marché
Forte croissance des applications critiques
5
Availability Environment Classification
0
Conventional
System could interrupted, data integrity is not
essential (no redundancy)
1
Highly Reliable
System could interrupted, data integrity is
necessary (database server with redundant
storage)
2
High Availability
System could interrupted during predefined
working hours for a short time span (stand-by or
cluster)
3
Fault Resilient
System should not interrupted during predefined
working hours (no single point of failure,
maintenance during off time)
4
Fault Tolerant
System should not interrupted (no single point of
failure, maintenance anytime possible)
5
Disaster Tolerant
System should not interrupted - even in case of a
local disaster (redundant application with
mirrored storage)
(Harvard Research Group)
6
Concepts de la disponibilité
A la poursuite des 9
Technologie
Indisponibilité moyenne annuelle
Taux de disponibilité
Stand alone server
03d 15h 39m 36s
99
Stand-by server
00d 08h 45m 57s
99,9
SW fault tolerance
00d 00h 52m 35s
99,99
HW fault tolerance
00d 00h 05m 15s
99,999
Continuous Availability
00d 00h 00m 31s
99,9999
Calcul de lindisponibilité moyenne annuelle en
minutes 365,25 x 24 x 60 x ( 1 taux de
disponibilité) Ex 365,25 x 24 x 60 x ( 1
99,999/100) 5,2596 5 min 15s
7
Stratus The Best Way to Protect Uptime
The Best Way to Protect Uptime is Every Way
Possible
8
Continuous Processing Technology
9
Technologie Lockstep
Continuous Processing Systems Systèmes à
traitement continu
Indisponibilité plate-forme supprimée
10
Continuous Processing Technology
11
Dispositif logiciel renforcé
Fiabilité logicielle

• Logiciels protégés contre les erreurs matérielles
  
• Drivers renforcés
• Toute erreur est capturée, analysée et corrigée
• Toutes les données, en mémoire et sur disques,
  sont protégées de manière absolue
• Tests intenses en laboratoire de la robustesse
  logicielle

Protection du système dexploitation
Prévient, diagnostique, et résout les problèmes
logiciels
12
Continuous Processing Technology
13
ActiveService Architecture
Technology-Enabled Services

• Automatic fault detection and isolation
• Integrated call-home and online remote support
• Collaborative support with Stratus and third
  party
• Out-of-band management

Failure Prevention
ActiveService Architecture
Detect and report problems before they cause
downtime
14
Les Services embarqués Stratus
1. Sur panne, le système isole lélément en
faute. Lexploitation continue sur lélément
redondant
2. Le rapport dincident est remonté via ASN au
CAC Stratus
3. Nouvel élément envoyé
4. Le système se re-synchronise après
remplacement et continue les traitements. Le
modèle série W 2300 est réparé durant une
période darrêt planifié (hors production)
15
Windows sur le ftServer
Daemons Services Applications Management
Replicator
Other
File Server
Logon
SQL Server
Security
Alterer
MS Exchange
Session Manager
Event Logger
Web Server
Windows 2000 Advanced Server / Windows Server
2003 EE
I/O System
Security
Local
Virtual
Process Structure
Object Services
File Server
Object management/executive run time
Device Drivers (Level 1)
Hardware Abstraction Layer (HAL) PCI
Config IOVA Interrupts
Kernel
Level 2
Platform Interface
I/O Devices
DMA Control
BIOS
Cache Control
Privileged Architecture
16
Indisponibilité Supprimée
Cluster
PANNE
ftServeur
17
Disponibilité des applications vs. disponibilité
des systèmes
Source Standish Group, June 2002. Data
collected over a 12 month period from 1,893
enterprise production systems. Downtime numbers
include planned downtime and unplanned downtime
due to operator errors, network and environmental
issues as well as system software and hardware
problems.
La complexité réduit la disponibilité de
lapplication
18
High Availability (AEC 2)
Clusters
Mainframe
FT
Source Standish Group, June 2002 Data
collected from over 2000 Enterprise
systems Downtime numbers include operator errors,
planned, extended planned, network, environment
as well as system software and hardware.
19
Technologies associées
AEC5/5
AEC3/4
99,999
AEC2
99,9
ClusteredFault-tolerant Servers in a
DR architecture
AEC1
99,5
AEC0
Fault-tolerant servers
Stand-by or clustered servers
Stand-alone servers with RAID storage
Stand-alone servers
20
Deux technologies alternatives
Client PCs
Public Network
Private network
AEC2
AEC4
Cluster Servers
Fault Tolerant Server
99,9
99,999
Récupération dun service applicatif, sur panne
dun nud, après basculement
Résilience locale
Continuité du service applicatif quelque soit la
panne HW au sein du système
RAID Disk Sets
21
Technology Partners

• Collaborate on OS availability features
• Memory resynchronization
• Hardened driver features
• Support model for mission-critical applications
• ftServer systems in Redmond test labs

• Joint program to insure processor determinism
• Early access to new processors
• ftServer systems in Intel labs

• ftServer systems in EMC Qualification lab
• Reseller relationship for CLARiiON
• Collaborative support model

22
Architecture ActiveService
Automatisation du Service

• Dispositif embarqué
• Détection et isolation automatique des pannes
• Système dappel intégré
• Supervision et support à distance
• Support collaboratif avec Stratus et les
  partenaires
• Contrôle complet de la machine

Le Serviceen continu
Détection proactive des pannes Remontée immédiate
des diagnostics
23
Gamme Stratus ftServer W Series

• Conforme aux standards
• Processeurs Intel
• Serveurs Microsoft Windows
• HCL Microsoft (Hardware Compatibility List)

5600
3300
6600
2300
Bi ou quadri
Mono ou bi-processeur
Mono
Serveurs extensibles de milieu de gamme pour des
besoins supérieurs en environnement critique
Accessible, FTServer dentrée de gamme
Serveurs dentreprisePour les applications les
plus critiques et exigeantes en performance
Leader Prix/performance pour les applications
critiques
24
Famille ftServer W Series
25
ftServer versus ClusterLes critères de décision
Complexité
Coûts
26
ftServer versus Cluster
27
ftServer versus Cluster
28
Stratus la technologie
29
Stratus les économies induites
Pas de coûts cachés Pas de connaissances
spéciales à avoir Pas de coûts dadministration
supplémentaires Pas de coûts de maintenance
supplémentaires Pas de coûts darrêt serveur Pas
de licences logicielles doublées
Coût total de possession 1/3 moins cher quun
cluster UNIX 1/5 du coût dun HP NonStop
Server 1/50 darrêt par rapport à un cluster Intel
Retour sur investissement Équilibre financier
rapide 12-15 mois (comparé à un serveur
simple) 9-12 mois (comparé à un cluster)
30
Stratus Continuous Availability
ftServer Availability Uptime MeterSM www.stratus.
com
31
Ils ont choisi Stratus
Boeing Grupo Elektra Euronext Caisses
dEpargne Lockheed Martin United Airlines Goldman
Sachs Lehman Brothers NASDAQ Thalès BNP-Paribas UP
S Albertsons PMU
EDF Sprint Crédit Lyonnais Merrill
Lynch BancOne France Telecom Bred-Banque
Populaire Bank of America Vivendi Wells
Fargo Ford La Poste Alcatel Chronopost
32
Stratus ftServiceSM Offerings
Assured AvailabilitySM Offering
Assured Availability PlusSM Offering
Standard Warranty
System AvailabilitySM Offering
W Series 2300 system is RTF warranty and
Warranty Uplift (not Warranty extension)
Not available on W Series 2300 system
33
Solutions à disponibilité continue Serveurs ?
Applications ? Services
34
Mise en oeuvre

• Partenaire certifié Microsoft
• Partenaire et revendeur de VMware
• Partenaire certifié et revendeur de Stratus
• Partenaire pour les baies de stockage EMC2
• Votre partenaire pour mettre en uvre ces
  technologies dans votre entreprise ou chez vos
  clients
• www.gerber-systems.ch

35
Questions?
36
ftServer W Series 2300 Systems

• Workloads and Attributes
• Highly affordable fault-tolerance
• Easy to deploy and simple to manage
• Ideal for volume deployment at remote locations

• Entry 1-way fault-tolerant server
• Next generation lights-out remote access
• Ideal for remote replicated applications

Specifications

• 4U rack-optimized form factor
• Pedestal option
• 1-way CPU
• 3.06 GHz Xeon processor
• Dual modular redundancy
• SATA storage supporting6 internal disks

• 240 GB internal storage
• 1GB 4GB memory
• 4 64/33 PCI slots
• Integrated 10/100/1000 Ethernet
• Windows 2003 Standard Editions

Single CRU
37
Integrated 2300 System Design
38
ftServer W Series 2300 Integrated Components
System Packaging / Rear View
W Series 2300 system
VTM Access (ethernet, serial) USB ports (2) VGA
(1) Serial ports (2) 10/100/1000 CU Enet (2)
39
ftServer W Series 2300
System Packaging / Front View
USB Connector System LEDs Power Switch Dump
Switch CD-ROM or CD-RW/DVD Hot-swappable SATA
Disks
W Series 2300 system
40
Service Model Comparison
41
Building Blocks of the ftServerW Series Modular
Family
42
ftServer W SeriesModular Family Components
CPU Module

• 1U or 2U Blindmate Enclosure, front removable
• 1-way, 2-way or 4-way capable modules
• Six or twelve DIMM slots
• 200 MHz Memory bus
• 400 or 533 MHz Front side bus
• LED Status panel, front accessible

43
ftServer W SeriesModular Family Components
Core I/O

• 1U Blindmate Enclosure, front removable
• One CD-ROM Blindmate, front removable
• Up to 3 Hot plug, SCSI 3.5 Drives, front
  removable
• Power On button, front accessible protected
• LED Status Panel, Dump Button Speaker, front
  accessible
• 2 Customer Configurable PCI slots, rear
  accessible
• 1 PCI slot, ftSAA or Video, rear accessible
• 1 SCSI port 0.8mm, rear accessible
• 1 10/100 1 10/100/1000 Com port, rear
  accessible
• Status LED rear viewable

44
ftServer W SeriesModular Family Components
Expansion I/O

• Supports 3 x 64bit/66 MHz PCI cards
• 3.3V or Universal Slot PCI cards only
• PCI bus will also operate at 33 MHz
• Extra LVDS channel for 3rd CPU Support (TMR)
• System management connects via Core IO BMC.
• Controls Power On/Off Reset
• Controls CRU Status LEDs
• Monitors Temperature, Voltages, Fan Speeds

45
ftServer W SeriesModular Family Components
Passive Backplane

• Passive backplane for seamless component
  interconnect
• Blind mates power and signal to CPU, Core IO and
  expansion IO CRUs
• Standard 19 EIA compatible
• Adjustable rail fits cabinet depth range 24 to
  30
• Shared connectors located in rear

46
ftServer W SeriesModular Family Components
System Packaging / Rear View

• Ethernet Ports
• 10/100 Enet (2)
• 10/100/1000 Cu Enet (2)
• Serial ports (2)
• USB ports (2)
• VGA (1)
• Redundant power

3300 Example
47
ftServer W Series 3300 Systems

• Workloads and Attributes
• Dedicated business processing application server
• Flexible enough for remote offices ideal for
  space-constrained data centers
• Price / performance leader

• Space efficient
• Entry 2-way fault-tolerant server
• Ideal for business processing applications

Specifications

• 4U rack-optimized form factor
• Pedestal option
• 1-2 way SMP
• 3.06 GHz Xeon processor
• Dual modular redundancy
• Ultra160, storage supporting6 internal disks

• 876 GB Storage, optional external JBOD
• 1 GB 6 GB memory
• 6 PCI slots 4-64/33, 2-32/33
• Integrated 10/100 and
• 10/100/1000 Ethernet
• Windows 2003

Core I/O
Core I/O
1-2 way SMP CPU
1-2 way SMP CPU
48
ftServer W Series 5600 Systems

• Workloads and Attributes
• Rack-optimized for data center locations
• Apps needing higher I/O throughput like Exchange
  and databases
• Ultimate investment protection platform

The no-compromisefault-tolerant server solution
when flexibility, expandability and headroom are
required.
Specifications

• 6U rack-optimized form factor
• 1-2 way SMP
• 3.2 GHz Xeon processor
• Dual modular redundancy
• 36 GB to 146 GB drive capacity
• 1 GB 12 GB min/max memory

• 6 PCI slot I/O expansion standard
• Integrated redundant 10/100 and
• 10/100/1000 Ethernet
• Optional redundant ftServer access adapters
• Optional external JBOD

Core I/O
Core I/O
1-2 way SMP CPU
1-2 way SMP CPU
49
ftServer W Series 6600 Systems

• Workloads and Attributes
• Enterprise-class, 4-way server designed for data
  center environments that require maximum
  computing power for highly complex
  business-critical applications

Specifications
The bet-your-business, enterprise-class
server delivering maximum performance
and availability.

• 2.0 GHz 1MB iL3 cache 2-way or 4-way DMR 6600
• 3.0 GHz 4MB iL3 cache 2-way or 4-way DMR or TMR
  6600
• 2 GB 24 GB Min/Max Memory
• Windows Server 2003, Enterprise Edition
• Integrated redundant 10/100 and 10/100/1000
  Ethernet
• Optional external 2 Gb Fibre Channel JBOD
• Standard ftServer access adapters
• 12/10 PCI slots total/available
• 10U rack rack-optimized form factor

gt 12 GB requires future ftServer software
release
50
End-to-End Storage - EMC

• CLARiiON Hardware Supported
• CX300, CX500, CX700
• Software support
• PowerPath I/O path management
• Navisphere Management Suite
• SnapView array-based local replication
• MirrorView remote data mirroring
• VisualSAN console for SAN management
• VisualSRM centralized storage resource management

CX300
CX500
CX700
51
Additional Hardware Peripheral Options
KVM switch
1U monitor
15 LCD display
Rack-mount tape