Thursday 27 Aug at the VLDB 2009 conference
Good day,
I wanted to blog everyday while attending the VLDB (Very Large Databases) conference, but I did not have time and my power adapter did not work at first.
So I will start from today and work back to Monday.
There are 4 time slots during the day. Each contains a few sessions. There are 700 researchers here.
Time slot 1:
10 year award keynote:
Database Architecture Evolution: Mammals Flourished long before Dinosaurs became Extinct
Peter A. Boncz (CWI, Amsterdam), Stefan Manegold (CWI, Amsterdam), Martin L. Kersten (CWI, Amsterdam)
A few notes/keywords:
Time slot 2: 11:00-12:30
Data Processing on FPGAs
Rene Mueller (ETH Zurich), Jens Teubner (ETH Zurich), Gustavo Alonso (ETH Zurich)
Interesting to see what they do wiht FPGS.
HadoopDB: An Architectural Hybrid of MapReduce and DBMS Technologies for Analytical Workloads
Azza Abouzeid (Yale Univ.), Kamil Bajda-Pawlikowski (Yale Univ.), Daniel Abadi (Yale Univ.), Alexander Rasin (Brown Univ.), Avi Silberschatz (Yale Univ.)
I liked this talk on HadoopDB. Azza (she presents very well) and Kamil presented.
They have taken Hadoop and have used/integrated PostgreSQL into it. They are also using Hive to provide SQL like functionality.
They have a few benchmark results as well.
HadoopDB is the new baby elephant of Hadoop and PostgreSQL and it needs nurturing.
Parallel datbases vs MapReduce
Read: MapReduce: A major step backwards Time slot 3+4:
Column oriented Database Systems I+II
Daniel J. Abadi (Yale), Peter A. Boncz (CWI, Amsterdam), Stavros Harizopoulos (HP Labs)
This was a great session. They provided a nice history about column oriented databases.
Stavros: the history
Daniel: The compression algorithms used in these databases to compress the colums
Peter: How column databases fit/work with the CPU architecture
My thoughts: What to take a away from the session is that colum databases are a good option, they have grow a lot in the last few years, it is important to think about cpu caches.
Read up more on NSM, DSM and PAX. PAX has been mentioned a few times.
SIMD's are also mentioned at a few talks.
I wanted to blog everyday while attending the VLDB (Very Large Databases) conference, but I did not have time and my power adapter did not work at first.
So I will start from today and work back to Monday.
There are 4 time slots during the day. Each contains a few sessions. There are 700 researchers here.
Time slot 1:
10 year award keynote:
Database Architecture Evolution: Mammals Flourished long before Dinosaurs became Extinct
Peter A. Boncz (CWI, Amsterdam), Stefan Manegold (CWI, Amsterdam), Martin L. Kersten (CWI, Amsterdam)
A few notes/keywords:
- Very good presentation. Funny
- These are the MonetDB guys
- The Started this research in 1985/1994?
- Wrote a few papers
- Databases hit the memory wall paper in .... influenced their research
- They are a COLUMN storage db
- Partitioned hash joins
- multi pass clustering
- RISC relational algebra
- Materialized vs pipeline
- They created Vectorwise that do pipeline
- Read paper "One size fits all: A concept whose time has come and gone" m stonebraker
- INGRES is also invoved with Vectorise
- They said db2, Oracle are dinosaurs. Also in 2005 it was found that there was a mammal that feeds on dinosaurs and that maybe MonetDb is a mammal......
- MonetDB has grow and adapted and for that reason they are still going strong. They are like a fox and Vectorise(new product from them) is a fast shark :)
- Read paper: Breaking the memory wall in MonetDB
Time slot 2: 11:00-12:30
Data Processing on FPGAs
HadoopDB: An Architectural Hybrid of MapReduce and DBMS Technologies for Analytical Workloads
I liked this talk on HadoopDB. Azza (she presents very well) and Kamil presented.
They have taken Hadoop and have used/integrated PostgreSQL into it. They are also using Hive to provide SQL like functionality.
They have a few benchmark results as well.
HadoopDB is the new baby elephant of Hadoop and PostgreSQL and it needs nurturing.
Parallel datbases vs MapReduce
Read: MapReduce: A major step backwards Time slot 3+4:
Column oriented Database Systems I
Daniel J. Abadi (Yale), Peter A. Boncz (CWI, Amsterdam), Stavros Harizopoulos (HP Labs)
This was a great session. They provided a nice history about column oriented databases.
Stavros: the history
Daniel: The compression algorithms used in these databases to compress the colums
Peter: How column databases fit/work with the CPU architecture
My thoughts: What to take a away from the session is that colum databases are a good option, they have grow a lot in the last few years, it is important to think about cpu caches.
Read up more on NSM, DSM and PAX. PAX has been mentioned a few times.
SIMD's are also mentioned at a few talks.
Comments
Post a Comment