Publications by Christian Plattner
2008
VLDB J., January 2008
@inproceedings{abc, author = {Christian Plattner and Gustavo Alonso and M. Tamer {\"O}zsu}, booktitle = {VLDB J.}, title = {Extending DBMSs with satellite databases.}, url = {http://dx.doi.org/10.1007/s00778-006-0026-x}, year = {2008} }
2006
Proceedings of the ACM SIGMOD International Conference on Management of Data, Chicago, Illinois, USA, January 2006
@inproceedings{abc, author = {Christian Plattner and Andreas Wapf and Gustavo Alonso}, booktitle = {Proceedings of the ACM SIGMOD International Conference on Management of Data, Chicago, Illinois, USA}, title = {Searching in time.}, url = {http://doi.acm.org/10.1145/1142473.1142578}, year = {2006} }
ETH Zürich, Diss. Nr. ETH Zürich, January 2006
This thesis deals with the construction of Ganymed, a platform for cluster based database
replication. Database replication is typically used to either improve fault tolerance or to
improve the performance of computational operations on the replicated data (or both).
Ganymed, however, goes beyond those goals. So called satellite databases are not only
used to achieve the goals of traditional replication, rather they can be used to extend
existing databases in various ways.
In the thesis we first discuss RSI-PC, a new, efficient algorithm for database replication.
The algorithm schedules transactions between a master database and a set of attached
satellite databases. The idea is to focus on transactional web applications: in such
environments, the commonly encountered scenario is a workload that consists merely
of complex read-only transactions that are accompanied by few short and simple update
transactions. Hence, while ensuring consistency and correctness at all times, the main
focus of RSI-PC is to use read-only satellite databases to maximize the speed up for the
execution of read-only transactions. Due to the fact that consistency and correctness are
never abandoned, the resulting system acts like a single image system and therefore no
changes are needed on the client side.
Based on the core system the thesis then shows how it can be extended to cope with
various other issues: the flexibility and independency of a concrete backend database
engine provider is shown by constructing heterogeneous setups where the master database
and the satellites are different engines from different vendors. The ability to extend the
functionality of the master database is shown by using Ganymed as way to implement
new functionality on the satellites. By using Ganymed, clients are not aware that the
added features do not belong to the master but rather are being offered by the satellites.
An example of such a feature is time-travel, where clients can inspect older states of the
database. By using the Ganymed approach, space requirements or query execution times
on the master are not affected in any way, at the same time satellites can be used to offer
arbitrary time-travel functionality.
The thesis furthermore shows that Ganymed is not limited to static setups. By using
dynamic replication the system can vary the amount of attached satellites to a master
as demand requires. Furthermore, the system can be used to implement multi-tenant
support, i.e., many different customer databases can be handled in parallel. For each
tenant the system then dynamically manages a tailored set of satellites. Last but not least
it is important to note that the system and the proposed ideas in the thesis have been fully
implemented.
@phdthesis{abc, abstract = {This thesis deals with the construction of Ganymed, a platform for cluster based database replication. Database replication is typically used to either improve fault tolerance or to improve the performance of computational operations on the replicated data (or both). Ganymed, however, goes beyond those goals. So called satellite databases are not only used to achieve the goals of traditional replication, rather they can be used to extend existing databases in various ways. In the thesis we first discuss RSI-PC, a new, efficient algorithm for database replication. The algorithm schedules transactions between a master database and a set of attached satellite databases. The idea is to focus on transactional web applications: in such environments, the commonly encountered scenario is a workload that consists merely of complex read-only transactions that are accompanied by few short and simple update transactions. Hence, while ensuring consistency and correctness at all times, the main focus of RSI-PC is to use read-only satellite databases to maximize the speed up for the execution of read-only transactions. Due to the fact that consistency and correctness are never abandoned, the resulting system acts like a single image system and therefore no changes are needed on the client side. Based on the core system the thesis then shows how it can be extended to cope with various other issues: the flexibility and independency of a concrete backend database engine provider is shown by constructing heterogeneous setups where the master database and the satellites are different engines from different vendors. The ability to extend the functionality of the master database is shown by using Ganymed as way to implement new functionality on the satellites. By using Ganymed, clients are not aware that the added features do not belong to the master but rather are being offered by the satellites. An example of such a feature is time-travel, where clients can inspect older states of the database. By using the Ganymed approach, space requirements or query execution times on the master are not affected in any way, at the same time satellites can be used to offer arbitrary time-travel functionality. The thesis furthermore shows that Ganymed is not limited to static setups. By using dynamic replication the system can vary the amount of attached satellites to a master as demand requires. Furthermore, the system can be used to implement multi-tenant support, i.e., many different customer databases can be handled in parallel. For each tenant the system then dynamically manages a tailored set of satellites. Last but not least it is important to note that the system and the proposed ideas in the thesis have been fully implemented.}, author = {Christian Plattner}, school = {ETH Z{\"u}rich}, title = {Ganymed: A Platform for Database Replication}, year = {2006} }
Middleware 2006, ACM/IFIP/USENIX 7th International Middleware Conference, Melbourne, Australia, January 2006
@inproceedings{abc, author = {Christian Plattner and Gustavo Alonso and M. Tamer {\"O}zsu}, booktitle = {Middleware 2006, ACM/IFIP/USENIX 7th International Middleware Conference, Melbourne, Australia}, title = {DBFarm: A Scalable Cluster for Multiple Databases.}, url = {http://dx.doi.org/10.1007/11925071_10}, year = {2006} }
2004
Middleware 2004, ACM/IFIP/USENIX International Middleware Conference, Toronto, Canada, January 2004
@inproceedings{abc, author = {Christian Plattner and Gustavo Alonso}, booktitle = {Middleware 2004, ACM/IFIP/USENIX International Middleware Conference, Toronto, Canada}, title = {Ganymed: Scalable Replication for Transactional Web Applications.}, year = {2004} }
IEEE Data Eng. Bull., January 2004
@inproceedings{abc, author = {Christian Plattner and Gustavo Alonso}, booktitle = {IEEE Data Eng. Bull.}, title = {Ganymed: Scalable and Flexible Replication.}, url = {ftp://ftp.research.microsoft.com/pub/debull/A04june/cplattne.ps}, year = {2004} }