Haute disponibilité et stockage pour Exchange Server 2010

Présentation au sujet: "Haute disponibilité et stockage pour Exchange Server 2010"— Transcription de la présentation:

1 Haute disponibilité et stockage pour Exchange Server 2010
3/31/2017 2:48 AM Haute disponibilité et stockage pour Exchange Server 2010 Guillaume Bordier Philippe Maurent Microsoft Services France © 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

2 Introduction Nouvelle approche du stockage
Une toute nouvelle approche de la haute disponibilité : plus de SPOF ni SAN, ni RAID, ni Backup ?

3 Stockage Exchange 2010

4 Le stockage avant Exchange 2010
Premier facteur de coût de l’infrastructure Le besoin en performances (IOPS) croit avec la taille de la boite aux lettres Impact de la mise en tolérance à la panne du stockage Les solutions d’archivage sont souvent vues comme le moyen de limiter la taille de la boîte.

5 Exchange Server 2010 (Beta) Storage Vision
3/31/2017 2:48 AM Exchange Server 2010 (Beta) Storage Vision Optimisation pour les disques peu couteux (SATA) Moins d’IO (random) Grosses boites aux lettres peu couteuses Flexibilité du stockage Haute dispo logicielle S’adapte aux configurations jusqu’à plusieurs millions de boites. Une boite de 10 GB avec Exchange 2010 consomme autant qu’une boite de 200MB © 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

6 Nouveautés pour le stockage
Contigüité logique, physique et temporelle (Schéma) Adaptation du moteur ESE Taille des IOs -> 32 KB OLD2, Checkpoint depth = 100MB Algorithmique du cache lecture en avance Groupement des écritures .. Version Courte Version Longue

7 Réduction d’IO : Schéma
3/31/2017 2:48 AM Réduction d’IO : Schéma Passer d’Ios random à séquentielles Elément Exchange Server 2007 Exchange Server 2010 (Beta) Physical Contiguity (ESE) Pas de gestion de la contigüité physique des pages : 1 IO = 1 page Bonne contigüité : 1 I/O = environ 100 pages. Logical Contiguity (Store) Les en-têtes des messages sont stockées dans une table par dossier : beaucoup de petites I/O réparties sur beaucoup de tables Les en-têtes de toutes les boites sont dans la même table : de plus grosses IO . Temporal Contiguity (View) Toutes les vues et index sont mis à jour à chaque fois qu’un message est délivré : beaucoup d’IO random en permanence Les vues et index ne sont mis à jour que quand l’utilisateur y accède: de grosses I/O sur les mêmes zones du disque moins souvent Schéma © 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

8 IO Reduction: Store Table Architecture
3/31/2017 2:48 AM IO Reduction: Store Table Architecture Per Database Per Folder Mailbox Table Jeff’s Mbx Ann’s Mbx Joe’s Mbx Folders Table Jeff:Inbox Ann:Drafts Joe:Unread Message Table (Msg) Joe:Msg10 Jeff:Msg32 Ann:Msg180 Attachments Table Jeff:Excel.xls Ann:Pic.bmp Joe:Help.doc Message/Folder Table (MFT) Joe:Inbox:H1 Joe:Inbox:H2 Joe:Inbox:H3 Exchange Server 2007 Secondary Indexes used for Views Per Database Per Mailbox Per View Mailbox Table Jeff’s Mbx Ann’s Mbx Joe’s Mbx Folders Table Joe:Inbox Joe:Drafts Joe:Unread Message Header Table Joe:H10 Joe:H302 Joe:H920 Body Table Joe:Msg10 Joe:Help.doc Joe:Msg302 View Tables (e.g. From) Joe:H920 Joe:H302 Joe:H10 Exchange Server 2010 (Beta) Schéma Nouveau Schéma : plus de Single Instance du tout © 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

9 Modifications ESE (moteur)
Adaptation au nouveau schéma Compression de page Défragmentation permanente Alloue l’espace en fonction de la contigüité Cache plus intelligent 100MB checkpoint depth (40% moins de writes) Compression du cache Espacer les écritures pour en faire moins Taille de page : 32K ESE

10 Allouer l’espace en fonction de la contigüité
3/31/2017 2:48 AM Allouer l’espace en fonction de la contigüité Optimisation de l’allocation d’espace Allocation en fonction de la contigüité ou de la fragmentation (en fonction de relevé d’usage) Space Contiguity Page X Msg Header Page Y Msg Header Page Z Event History DB Cache Space Compactness Disk Page 1 Used Page Page 2 Event History Page 3 Used Page Page 4 Msg Header Page 5 Msg Header Contiguity Random/Compact Sequential/Bloat ESE © 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

11 Exchange Server 2007 Service Pack 1 (SP1) Exchange Server 2010 (Beta)
3/31/2017 2:48 AM Maintenance en ligne New Database Maintenance Architecture: ESE Function Exchange Server 2007 Service Pack 1 (SP1) Exchange Server 2010 (Beta) Cleanup (deleted items/mailboxes) Cleanup performed during Online Defrag (OLD) which occurs during Online Maintenance (OLM) time window Cleanup performed at run time (when hard delete occurs)—happens during Store dumpster cleanup (OLM), pages are zeroed by default Space Compaction (deleted items/mailboxes) Database is compacted and space reclaimed during Online Defrag (OLD) Database is compacted and space reclaimed at run-time—auto-throttled Maintain Contiguity (defragmentation) N/A: Contiguity is compromised by space compaction Database is analyzed for contiguity and space at run time and is defragmented in the background (B+Tree Defrag/OLD2)—auto-throttled Database Checksum When configured, ½ of OLD maintenance window reserved for sequential scan (Checksum), manual throttle—active DB copy only Two options (both Active and Passive copies): Run DB Checksum in the background 24x7 (default). Sequential IO Run DB Checksum during OLM window. Sequential IO ESE © 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

12 Contrebalancer le grossissement
3/31/2017 2:48 AM Contrebalancer le grossissement Problème : Grossissement de la base de 20% (Nouveau Schéma, la defragmentation et les pages de 32 KB) Solution : Compression de page (header et body) DB File Size Comparison DB Space Analysis Counts E2K7 SP1 E2010 Mailbox Count 750 Tables 14.754 92.435 Secondary Indexes 85.784 4.557 Pages Used Pages (%) 85.7% 86.7% Available Pages (%) 14.3% 13.3% Msg Table (% space) 84.9% 80.0% Msg Views 32KB Pages 1 Database, 750 x 250MB mailboxes RTF = RTF Compressed, Mix = 77% HTML, 15% RTF, 8% Text Avg. Message size = ~50KB ESE © 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

13 Réduction des Lectures Augmentation de la taille de page à 32 KB
3/31/2017 2:48 AM Réduction des Lectures Augmentation de la taille de page à 32 KB Exchange Server 2007 DB Read 20 KB Message Page 1 Msg Header Page 3 Msg Body Page 5 Msg Body DB Cache Disk 3 Read IO’s 8 KB Pages Page 1 Msg Header Page 2 X Page 3 Msg Body Page 4 X Page 5 Msg Body Exchange Server 2010 (Beta) DB Read 20 KB Message Page 1 (32KB) Msg Header, Msg Body DB Cache 1 Read IO Disk Page 1 (32KB) Msg Header, Msg Body 32 KB Pages Page 2 (32KB) X Cache © 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

14 Réduction des lectures Lecture en avance
3/31/2017 2:48 AM Réduction des lectures Lecture en avance Page 1 Msg Header Page 3 Msg Body Page 5 Msg Body Exchange Server 2007 DB DB Cache Disk 3 Read IO’s Page 1 Msg Header Page 2 X Page 3 Msg Body Page 4 X Page 5 Msg Body Exchange Server 2010 (Beta) DB Page 1 Msg Header Page 2 Temp Buffer Page 3 Msg Body Page 4 Temp Buffer Page 5 Msg Body DB Cache 1 Read IO Disk Page 1 Msg Header Page 2 X Page 3 Msg Body Page 4 X Page 5 Msg Body Cache © 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

15 Espacement des écritures dans le temps
3/31/2017 2:48 AM Réduction des écritures (dans la base) espacer les écritures et les grouper Exchange Server 2007 DB Write Behavior DB Cache Page 1 Dirty Page 2 Clean Page 3 Dirty Page 4 Clean Page 5 Dirty 3 Write IO’s Espacement des écritures dans le temps Disk Exchange Server 2010 (Beta) DB Write Behavior DB Cache Page 1 Dirty Page 2 Clean Page 3 Dirty Page 4 Clean Page 5 Dirty 1 Write IO Disk Cache © 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

16 Optimisé pour le SATA: Smooth Database Write IO
3/31/2017 2:48 AM Optimisé pour le SATA: Smooth Database Write IO Les pics d’écriture sur la base affectent les lectures de base et augmentent la latence d’écriture sur les logs Plus il y a d’écriture, plus il y a de contention E2K7=96 Maximum write Queue depth (global) Log Write IO Single 7.2k SATA disk, logs/db on same spindle, Loadgen load generating 250 RPC Operations/second, ~50 IOPS ESE © 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

17 Optimize for SATA Smooth DB Write IO
Throttle DB writes based on checkpoint target (QoS) When checkpoint depth equals 1x ->1.24x of checkpoint target, Limit Max Outstanding DB writes/LUN to 1 When checkpoint depth meets or exceeds 1.25x of checkpoint target, ratchet up max outstanding DB writes/LUN The further behind on checkpoint, the more aggressively we raise the max outstanding DB writes/LUN (maximum = 512/LUN) 20 MB Max Checkpoint Example Works for both JBOD SATA and RAID10 SAN! ESE

18 Diminution des couts de stockage
Boites aux lettres 4 x 2 Noeuds CCR 2 copies (RAID) 6 noeuds DAG 3 copies (JBOD) 24,000 profils très lourd, Mode cache, 2Go par boite, Rack/Cablage non inclus

19 Exchange Server 2010 (Beta) : Flexibilité
Boite d’archive “en ligne” dans Exchange 2010 Optimisé pour le DAS, mais le Stockage SAN est supporté Réduction des I/Os et contrôle des burst permet d’utiliser des disques SATA Une architecture de tolérance à la panne moins complexe et donc moins chère Nous supportons le stockage JBOD* ou RAID storage Exchange Server 2010 (Beta) optimisé pour les disques grand public SSD/Flash supportés mais non recommandés - raisons de coût 100 Bases par serveur, plus de storage group Taille de base maximum recommandée 2 TB* Nombre d’éléments par dossier maximum recommandé (Outlook Online ou OWA) : *3 copy High Availability only

20 Haute Disponibilité : une nouvelle approche

21 Haute disponibilité Exchange 2007
SCC LCR CCR SCR Site 1 Site 2 File Share

22 Haute disponibilité Exchange 2010
DAG Database Availability Group

23 Eléments de la Haute Dispo
Accès client : Client Access Array Le client se connecte au nom du Client Access Array (nom LB) Le CAS est la terminaison RPC du client et se connecte à son tour au DAG Transport Redundancy Shadow Queue/Transport Dumpster, les messages sont conservés sur un HUB tant que le next hop n’a pas confirmé la délivrance (commande SMTP XSHADOW) Base de données : Data Availability Group Jusqu’à 16 instances répliquées d’une même base dans un DAG Active Manager : Bascule automatique d’une copie vers l’autre

24 DAG : Adieu xCR, Adieu SCC
Technologie «Single Copy Cluster » abandonnée: => un serveur est toujours le seul à accéder à une base Technologies xCR remplacée par le DAG Fiabilisation Sockets TCP à la place de SMB La source « pousse » les fichiers logs sur les cibles LLR est de nouveau à 1 Un DAG est un ensemble de serveurs (jusqu’à 16) Windows Failover Clustering utilisé de manière transparente Une base peut être répliquée sur chacun des serveurs DAG

25 Ajouter un nœud dans un DAG
Demo Ajouter un nœud dans un DAG DAG

26 Ajouter un membre dans le DAG
Ajouter un membre dans le DAG est facile, ne nécessite pas de réinstallation, tout est automatique Passage d’un modèle de Quorum à un autre Démo PPT Suite du PPT DAG

27 Ajouter un membre au DAG

28 Exchange Server 2010 (Beta) Les databases sont des objets déclarés au niveau Org.
DAG

29 Mailbox Changements pour la haute disponibilité
Single-copy cluster (cluster “classique”) Cluster Continuous Replication Exchange Server 2010 High Availability Granularity de basculement Serveur Base de donnée Nombre de copie de bases 1 2 2 à 16 Temps de bascule ~2 min ~30 sec (POR) Contrôle de bascule Windows Cluster Exchange Server Data replication Technologie non MS (SRDF, VVM ..) Continuous replication Outil d’administration Séparé intégré Compatible avec d’autres rôles ? Non Oui Autres avantages Passer d’un MBX “normal” à un membre de DAG sans réinstallation Configuration de réplication base par base à la demande Pas de contraintes d’architecture réseau DAG

30 Mailbox Better Client Experience
Exchange 2007 Outlook Clients Exchange Server 2010 (Beta) Outlook Clients Failover: Client disconnected for 0-TTL minutes Exchange CAS NLB MBX MBX CAS Failure: Client just reconnects MBX1 MBX2 Failover: Connected client disconnected for 30 seconds (POR) DAG

31 JBOD/RAID-less Storage: Single Page Restore (Active)
3/31/2017 2:48 AM Page corruption detected on Active Copy (e.g ) Database Availability Group (DAG) Mailbox Server Node 1 Mailbox Server Node 2 Mailbox Server Node 3 Active DB places marker in log stream to notify passive copies to ship up to date page DB1-Active DB1-CopyA DB1-CopyB Passive receives log and replays up to marker, retrieves good page, invokes Replay Service callback and ships page Log Log Log Page1 Page1 Page1 Page2 Page2 Page2 Active receives good page, writes page to log, DB page is patched Page3 Page3 Page3 Database Database Database Subsequent page repair from additional copies ignored DAG © 2005 Microsoft Corporation. All rights reserved. This presentation is for informational purposes only. Microsoft makes no warranties, express or implied, in this summary.

32 JBOD/RAID-less Storage: Single Page Restore (Passive)
3/31/2017 2:48 AM JBOD/RAID-less Storage: Single Page Restore (Passive) Database Availability Group (DAG) Page corruption detected on DB Passive Copy (e.g ) Mailbox Server Node 1 Mailbox Server Node 2 Mailbox Server Node 3 Passive copy pauses log replay (log copying continues) DB1-Active DB1-CopyA DB1-CopyB Passive retrieves the corrupted page # from the active using DB seeding infrastructure Log Log Log Page1 Page1 Page1 Passive copy waits till log file which meets max required generation requirement is copied/inspected, then patches page Page2 Page2 Page2 Page3 Page33 Page3 Database Database Database Passive resumes log replay DAG © 2005 Microsoft Corporation. All rights reserved. This presentation is for informational purposes only. Microsoft makes no warranties, express or implied, in this summary.

33 Exchange Server 2010 (Beta) HA Fundamentals Mailbox Database Copy
Defines properties applicable to an individual database copy Copy status: Healthy, Initializing, Failed, Mounted, Dismounted, Disconnected, Suspended, FailedandSuspended, Resynchronizing, Seeding CopyQueueLength ReplayQueueLength ActiveCopy ActivationSuspended DAG

34 Active Manager Primary Active Manager (PAM)
S’exécute sur le noeud du cluster qui possède le “Default cluster Group” Détecte les changements de topologie Réagit aux défaillance de serveurs Sélectionne la meilleure base à activer Standby Active Manager (SAM) S’exécute sur tous les membres du DAG Réponds aux requêtes des autres composants Exchange sur l’état de la réplication et de l’activation des bases DAG

35 Mailbox Points d’architecture
Plus de streaming backup Plus de SCC (Single Copy Cluster) Préconisations : Stockage DAS pour des raisons de coût DAG pour la haute disponibilité Noms unique pour les bases dans toute l’organisation (obligatoire) Combiner base et logs sur les même disques n’est plus un problème A partir de 3 copies de base Le RAID devient optionnel On peut combiner Support de très grosses boîtes 1 à 2 ans de données dans la boîte active Le reste dans la boîte d’archive Office 2007 Service Pack 2 (SP2) minimum (Taille .OST !) DAG

36 Database Availability Group (DAG)
Exchange Server 2010 (Beta) HA/JBOD Storage Example: Single Site (or more), 3 Nodes, 3 Copy DAG 10,000 mailboxes heavy Profile: 120 messages/day .11 IOPS/mailbox 8 Cores 32 GB RAM 8 Cores 32 GB RAM 8 Cores 32 GB RAM 2 GB mailbox size Mbx Server 1 Mbx Server 2 Mbx Server 3 3,333 active mailboxes/server DB1 DB2 DB3 DB4 DB5 DB6 DB1 DB2 DB3 DB4 DB5 DB6 DB1 DB2 DB3 DB4 DB5 DB6 DB7 DB8 DB9 DB10 DB11 DB12 DB1 DB7 DB8 DB9 DB10 DB11 DB12 DB1 DB1 DB7 DB8 DB9 DB10 DB11 DB12 3 nodes, 3 copies = secondary failure resiliency DB13 DB14 DB15 DB16 DB17 DB18 DB13 DB14 DB15 DB16 DB17 DB18 DB13 DB14 DB15 DB16 DB17 DB18 DB1 DB1 DB1 DB19 DB20 DB21 DB22 DB23 DB24 DB19 DB20 DB21 DB22 DB23 DB24 DB19 DB20 DB21 DB22 DB23 DB24 1TB 7.2k disks (SAS/SATA) DB25 DB26 DB27 DB28 DB29 DB30 DB25 DB26 DB27 DB28 DB29 DB30 DB25 DB26 DB27 DB28 DB29 DB30 JBOD: 30 disks/node Database Availability Group (DAG) online spares Active copy Passive copy Spare Disk battery backed caching array controller DAG Legend

37 « Une vraie architecture multi-tiers »
Client Access Server « Une vraie architecture multi-tiers » CAS

38 Client Access Consolidation of Store Access Paths
Exchange Server 2007 Exchange Server (Beta) Exchange Components OWA Sync UM Transport Agents Mailbox Agents WS Middle Tier MAPI RPC MAPI.Net Core Objects XSO Mailbox Store Exchange Components OWA Sync UM Transport Agents Mailbox Agents WS Outlook / MAPI clients Entourage Middle Tier XSO MAPI.Net Outlook / MAPI clients NSPI Entourage Mailbox MAPI RPC Store DAV DSPROXY CAS

39 Client Access Points d’architecture
Nouveau ratio CAS:MBX 3:4 !!! Le CAS devient le point de connexion des clients pour tous les protocoles: MAPI (MoMT et DoMT) Outlook Anywhere OWA/ActiveSync/IMAP/POP/EWS/Entourage MoMT détecte le serveur disposant de la copie de base valide et effectue la connexion et gère un pool de 100 connexions avec chaque MBX Exchange Server 2010 (Beta) CAS dans chaque site où un MBX 2010 est présent Haute Disponibilité Hardware Load-Balancer obligatoire sur les serveurs CAS+MBX Car NLB n’est pas compatible avec WFC HLB conseillé au delà de 8 CAS dans la ferme OCS 2007 R2 obligatoire pour l’intégration OWA/OCS (Présence & IM/Chat) OWA 2010 ne permet d’accéder qu’aux Dossiers Publics E2010 CAS

40 « Garantir la livraison du message »
Hub Transport « Garantir la livraison du message »

41 Hub Transport Participation étendue des serveurs Hub Transport à la fourniture de la haute disponibilité: Transport Redundancy (Nouveau) Shadow Queues Mailbox Server resubmission Transport Dumpster (améliorations) Basé sur la santé des bases Transport

42 Transport Shadow Queue
Le Transport Shadow Queue fournit une solution supplémentaire au Transport Dumpster Les messages sont conservés sur le serveur précédent jusqu'à ce qu'il soit délivé plus loin Lorsqu'une erreur est détectée (timeout), le serveur précédent redélivre (à un autre HUB par exemple) Des extensions SMTP sont utilisées (XSHADOW, XDISCARD) Génère un léger overhead dans le dialogue Nécessite une chaine de serveurs SMTP Exchange 2010 (Hub Transport, Edge Transport) Le Submission service sur le MBX resoumets les messages si le HUB n‘a pas émis d‘avis de réception Transport

43 Transport Dumpster Améliorations
Diminution des IOPS supplémentaire dues au Dumpster Purge du Dumpster en fonction de l’état de la réplication de base Vérification à intervalle régulier par l'Active Manager du LastLogInspected pour chaque copie de DB Considère la "plus vieille" copie dans le DAG et utilise cette information comme "référentiel" (dumpster watermark) pour chacune des DBs (plus petit dénominateur commun pour une DB donnée) Les éléments plus anciens que le "référentiel" de la DB sont supprimés du dumpster (processus de feeback régulier) La taille du Transport Dumpster est donc adaptée à partir des éléments de latence de réplication et de la fréquence de feedback Les requêtes de resoumissions ne renvoient que les items plus récents que le "référentiel" (dumptser watermark) Transport © 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

44 Transport Points d’architecture
Les serveurs MBX 2010 ne peuvent communiquer qu’avec des HUB 2010 Communication possible entre HUB 2010 et HUB 2007 Un HUB 2010 dans chaque site ou un MBX 2010 est présent Pas besoin de RAID sur les files d’attente car le HUB grâce au Shadow Queue Transport

45 Shadow Queue Demo

46 Shadow Queue Suite de la présentation Démo PPT
Jean-Paul envoie un mail à Roberto Jean-Paul envoie un mail à Clint Relai SMTP tier Roberto Site S1 Site S2 Clint Jean-Paul Hub Transport CASHUB1-S1 Hub Transport CASHUB1-S2 XDISCARD XSHADOW Démo PPT Suite de la présentation

47 Tranport Shadow Demo

48 Synthèse

49 Chaque brique à sa place
Architecture totalement « stateless », on peut perdre définitivement n’importe quelle machine sans jamais perdre un message MBX => DAG + Dumpster HUB => Shadow CAS => stateless par nature 100 users x 10 GB = 3 disques SATA à 100€/disque !

50 High Availability Design Example Double Resiliency
AD: Dublin Upgrade server 1 Server 2 fails Server 1 upgrade is done 2 active copies die Single Site 4 Nodes 3 HA Copies JBOD -> 3 physical Copies CAS NLB Farm X Mailbox Server 1 Mailbox Server 2 Mailbox Server 3 Mailbox Server 4 X DB3 DB5 DB6 DB3 DB4 DB1 DB2 DB7 DB8 DB1 DB7 DB5 DB4 DB5 DB3 DB4 DB6 DB7 DB8 DB6 DB2 DB8 DB1 DB2 Database Availability Group (DAG) DAG

51 Pas de SAN?, pas de RAID , pas de Backup ?
Le SAN était obligatoire pour le SCC Exchange 2010 n’a plus besoin du même niveau de performances Il est désormais possible d’augmenter le service à l’utilisateur (taille de sa boîte) en maîtrisant le coût du stockage « Si ça marche sur un SATA 7200 en DAS, ça marchera aussi sur un SAN » Mécanismes de snaps ? RAID : à partir de 3 copies de bases, Microsoft supporte le RAID-less Réparation de page dynamique, détection d’erreurs Être en mesure de supporter un reseed complet dans certains cas Backup Combien de fois effectuez-vous des restauration de boites aux lettres de plus de 30 jours? Possibilité d’avoir une base en « retard » de 14 jours

52 3/31/2017 2:48 AM Q & A © 2009 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

53 3/31/2017 2:48 AM © 2009 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION. © 2009 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

54 Exchange Server 2010 (Beta) Storage Requirements/Best Practices
3/31/2017 2:48 AM Exchange Server 2010 (Beta) Storage Requirements/Best Practices Storage Guidance Stand Alone Exchange Server (Beta) 2010 HA (2 copies) Exchange Server 2010 (Beta) HA (3+ copies) Storage Type DAS, SAN (Fibre Channel, iSCSI) Disk Type SAS, Fibre Channel, SATA (with battery backed cache), SSD RAID RAID recommended RAID optional RAID Type RAID-1/0, RAID-5, RAID-6 JBOD DB/Log Isolation Best Practice Not required Windows Disk Type Basic (recommended), Dynamic Partition Type GPT (recommended), MBR Partition Alignment Windows 2008 Default (1MB) File System NTFS NTFS Allocation Unit Size 64 KB for both database and log volumes Encryption Support Outlook Protection Rules, Bitlocker © 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.