1. Applications Web et bases de données en grappe emmanuel.cecchet@emicnetworks.com Séminaire InTech 3 Février 2005 – Grenoble

2. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 2 - 03/02/2005 Modern Web applications Internet Web server Application server è Dynamic content Web sites è Multi-tier architecture è Replication  High availability  Performance scalability Database server Apache IIS J2EE.Net Oracle, DB2 MySQL, PostgreSQL

3. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 3 - 03/02/2005 Clustered Web applications Internet Web servers App. servers Database è Cluster for replication of each tier è Problems  Transparency  Load balancing  Failover  Management

4. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 4 - 03/02/2005 Web server clustering è Hardware  L4-switch è Software  RR-DNS  LVS (http://www.linuxvirtualserver.org/)  One-IP technique: Emic Lamp/cluster Internet

5. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 5 - 03/02/2005 Application server clustering è Web server/App server  mod-jk, mod-proxy/jk2  Enhydra director è Application server replication  HTTP in-memory session replication  JNDI (naming registry)  Stateful bean replication Internet RR-DNSjk2 session, JNDI, EJB

6. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 6 - 03/02/2005 Scaling the database tier – Alternative 1 (SMP) Internet Web frontend App. server Well-known database vendor here Database Well-known hardware + database vendors here è Cons  Cost  Scalability limit

7. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 7 - 03/02/2005 Scaling the database tier – Alternative 2 (shared disks) Internet Web frontend App. server Database Disks Another well-known database vendor è Cons  still expensive hardware  availability

8. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 8 - 03/02/2005 Internet è scalability è high availability è without modifying the client application è database vendor independent è on commodity hardware Database clustering middleware JDBC

9. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 9 - 03/02/2005 RAIDb concept è Redundant Array of Inexpensive Databases è RAIDb controller  gives the view of a single database to the client  balance the load on the database backends è RAIDb levels offers various tradeoff of performance and fault tolerance

10. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 10 - 03/02/2005 RAIDb levels è RAIDb-0  partitioning  no duplication and no fault tolerance  at least 2 nodes

11. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 11 - 03/02/2005 RAIDb levels è RAIDb-1  mirroring  performance bounded by write broadcast  at least 2 nodes

12. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 12 - 03/02/2005 RAIDb levels è RAIDb-2  partial replication  at least 2 copies of each table for fault tolerance  at least 3 nodes

13. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 13 - 03/02/2005 C-JDBC overview è Middleware implementing RAIDb  100% Java implementation  open source (LGPL) è Two components  generic JDBC driver (C-JDBC driver)  C-JDBC Controller è Read-one, Write all approach  provides eager (strong) consistency è Supports heterogeneous databases

14. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 14 - 03/02/2005 architectural overview JVM Application server C-JDBC JDBC driver C-JDBC controller JVM MySQL JDBC driver MySQL database

15. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 15 - 03/02/2005 Inside the C-JDBC Controller Sockets JMX

16. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 16 - 03/02/2005 Scaling existing databases è add open source databases for more fault tolerance and better performances è rules for on-the-fly query rewriting to handle heterogeneity

17. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 17 - 03/02/2005 TPC-W benchmark (Amazon.com) è Nearly linear speedups with the shopping mix

18. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 18 - 03/02/2005 Horizontal & Vertical scalability è adapt and evolve according to user needs è advanced caching functionalities è dynamic adding of database backends è fault tolerant transaction log è integrated backup and checkpointing è grid & edge-side servers support

19. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 19 - 03/02/2005 Administration è JMX Graphical Console  cluster administration  backup/checkpoint mgt  cluster monitoring  SQL profiling  integrated SQL console (iSQL) è Administration API  integration in any existing administration infrastructure  JMX notifications for all internal events  JMX MBeans for monitoring

20. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 20 - 03/02/2005 J2EE end-to-end reliability è ObjectWeb exclusivity è On-demand availability  no SPOF, transparent failover è On-demand scalability  scalability, flexibility Interne t Mission critical open source software

21. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 21 - 03/02/2005 What about Grids? Internet

22. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 22 - 03/02/2005 Q&A _________ Thanks to all users and contributors... http://www.emicnetworks.com http://www.objectweb.org http://www.apache.org

23. Bonus slides

24. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 24 - 03/02/2005 C-JDBC today è Web site  ~200.000 hits/month  >28.000 downloads  EU (18 countries) 36%, US 28%, Japan 12%, China 5%, Canada 4%, Australia 4%, India 3%, Brazil 2%, … è Community  27 committers both industrial & academics  c-jdbc@objectweb.org: >200 subscribers, 200-300 msgs/month  translation in japanese, italian, chinese, turkish, french, german è RPM on JPackage.org è Professional support by Emic Networks

25. emmanuel.cecchet@emicnetworks.com Séminaire InTech - 25 - 03/02/2005 Current limitations è JDBC only è Distributed joins è Updatable ResultSets è XA support through XAPool only è transparent controller failover not supported when using horizontal scalability with JGroups è network partition/reconciliation not supported