nosqlsummer papers

[Bram]

Hi

As mentioned before, the suggested papers for the summer are here:

<http://nosqlsummer.org/papers>

There are about 30 papers there, so we have to make choices. We could
device a complicated voting system.

Perhaps people with an opinion, should list their top three, then
compile a list, and start with the most popular ones.

Depending on the size of the group in the next meeting(s), we could
also consider to have a theoretical paper group and a practical paper
group...

Bram

[Òscar Vilaplana]

I've set up a poll:

http://www.twiigs.com/poll/Technology/Computers/57284

Salut,

Òscar Vilaplana

[Mathijs Homminga]

Good idea. My votes are in.

[Mathijs Homminga]

Hi guys,

Reminder: post your votes for the next paper here: 

http://www.twiigs.com/poll/Technology/Computers/57284

Poll is closed tomorrow evening, alright? That will give us a week to read and prepare the next paper.

Cheers,

Mathijs

On Jun 24, 2010, at 16:38 , Òscar Vilaplana wrote:

[Mathijs Homminga]

Hi guys,

After 7 votes, the winner is:

"Google’s BigTable"

The runners up are:

"The CAP Theorem" and "PNUTS: Yahoo!’s Hosted Data Serving Platform"

I propose that we do the BigTable paper next Thursday (July 8), then followed by the CAP paper (for July 21)

That way we might compare two different "NOSQL approaches" (Cassandra and BigTable) when discussing CAP.

Cheers,

Mathijs

[Bram Noordzij]

After 7 votes, the winner is:

"Google’s BigTable"

Nice (It could answer some open questions about Cassandra, such as SST/disk tables, column families, etc.)

 

The runners up are:

"The CAP Theorem" and "PNUTS: Yahoo!’s Hosted Data Serving Platform"

I propose that we do the BigTable paper next Thursday (July 8), then followed by the CAP paper (for July 21)

That way we might compare two different "NOSQL approaches" (Cassandra and BigTable) when discussing CAP.

I would like to make a case against the CAP paper. While indeed everyone (at least everyone attending the nosql meetings) should know and understand (and agree with?) the CAP theorem, it is not a very interesting paper per se. Half of it is devoted to proofing the obvious, the other half deals with a 't-connected consistent' algorithm and proof...

However, feel free to disagree!

There are more interesting and also theoretical papers, like: "Time, Clocks, and the Ordering of Events in a Distributed System", or more generic ones like "The End of an Architectural Era", "Eventually Consistent" and "Paxos Made Simple"

My impression is that Cassandra and BigTable are very similar, but maybe I'll change my mind after the meeting ;)

Another interesting discussion is about the consequences of: nosql. What was good about sql, will we miss it? With nosql one gains massive scalability, but what do we lose? Is it a liberation? How to work with unstructured or lightly structured data? Denormalization? Programming instead of declaring? etc.
Not sure what would be the best fitting papers, perhaps "the end of an architectural era"?

Bram

[Mathijs Homminga]

Hi Bram,

Thanks for your excellent thoughts.

I agree that we should go for interesting articles ;) 

I must admit I've not yet read the CAP paper, but if you state that it's boring (not your words, I know ;)) we might indeed pick another one.

Btw, I'm not sure if everyone indeed fully understands the CAP theorem and it might still be an interesting discussion topic.

I also like the idea of discussing the impact of losing some sql-ish features (although we're talking about NOSQL (not only SQL) and not NoSQL here ;)). 

Let's put this it on the agenda for Thursday!

Mathijs

[Berco Beute]

I agree with Bram. From the first meeting I get the impression that
(at the moment) there is quite some interest in the practical
application of nosql and less in the general theoretical foundations.
The latter is more or less assumed to be commonly understood.

I also think we shouldn't limit ourselves to the papers mentioned on
the nosql-summer site. I bet some of the more interesting papers on
what was good about SQL are not on the reading list. :)

Berco

On Jul 2, 11:28 am, Bram Noordzij <bram.noord...@gmail.com> wrote:
> After 7 votes, the winner is:
>

> > *"Google’s BigTable"*

>
> Nice (It could answer some open questions about Cassandra, such as SST/disk
> tables, column families, etc.)
>
> > The runners up are:

> > *"The CAP Theorem"* and *"PNUTS: Yahoo!’s Hosted Data Serving Platform"*

[Elena Lazovik]

I agree with Bram too. The paper of CAP theorem is very theoretical, and the meaning of theorem itself is quite clear.
It would be really more interesting to concentrate on other more practical papers.

Regards,
Elena.

[Ivor Bosloper]

I agree; the CAP paper is long-winded (the proofs are an exercise that should be left to the reader) but the CAP theorem itself is interesting.

Interesting discussion points for a CAP meeting would be;

- What is C, A and P and why is it (fundamentally) impossible to get all three at the same time
- What are examples and use cases for CP, AP or CP
- What are the CAP-related trade-offs in the design of existing NoSQL systems (e.g. Cassandra is AP or CP, depending on your quorum)
- How to design your application if you don't have C, A or P

But I don't know if there is a paper about it.

Regards,
Ivor

On 07/05/2010 02:47 PM, Elena Lazovik wrote:

I agree with Bram too. The paper of CAP theorem is very theoretical, and the meaning of theorem itself is quite clear.
It would be really more interesting to concentrate on other more practical papers.

Regards,
Elena.

On Mon, Jul 5, 2010 at 2:38 PM, Berco Beute <cyb...@gmail.com> wrote:

I agree with Bram. From the first meeting I get the impression that
(at the moment) there is quite some interest in the practical
application of nosql and less in the general theoretical foundations.
The latter is more or less assumed to be commonly understood.

I also think we shouldn't limit ourselves to the papers mentioned on
the nosql-summer site. I bet some of the more interesting papers on
what was good about SQL are not on the reading list. :)

Berco

On Jul 2, 11:28�am, Bram Noordzij <bram.noord...@gmail.com> wrote:
> After 7 votes, the winner is:
>

> > *"Google�s BigTable"*

>
> Nice (It could answer some open questions about Cassandra, such as SST/disk
> tables, column families, etc.)
>
> > The runners up are:

> > *"The CAP Theorem"* and *"PNUTS: Yahoo!�s Hosted Data Serving Platform"*

[Bram Noordzij]

On Thu, Jul 8, 2010 at 4:05 PM, Ivor Bosloper <ivor.b...@kalooga.com> wrote:

I agree; the CAP paper is long-winded (the proofs are an exercise that should be left to the reader) but the CAP theorem itself is interesting.

Interesting discussion points for a CAP meeting would be;

- What is C, A and P and why is it (fundamentally) impossible to get all three at the same time

Short: <http://en.wikipedia.org/wiki/CAP_theorem>
Longer: <http://www.julianbrowne.com/article/viewer/brewers-cap-theorem>
 

- What are examples and use cases for CP, AP or CP

 

A good and related paper seems to be:"Harvest, Yield, and Scalable Tolerant Systems (1999)", <http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.33.411>

- What are the CAP-related trade-offs in the design of existing NoSQL systems (e.g. Cassandra is AP or CP, depending on your quorum)
- How to design your application if you don't have C, A or P

 

But I don't know if there is a paper about it.

About 'dropping' C, there is a paper in the nosqlsummer list: eventual consistency (interesting). 

There are also people who argue against applying CAP and dropping C, but instead deal with recovery in case of problems with A or P: e.g. <http://cacm.acm.org/blogs/blog-cacm/83396-errors-in-database-systems-eventual-consistency-and-the-cap-theorem/fulltext>

Perhaps the "Brewer: Harvest, Yield, and Scalable Tolerant Systems (1999)", (5 pages), the "Werner Vogels: eventual consistency" (6 pages) and the blog post make a good read for the next session.

Bram

[Berco Beute]

Good idea, and maybe add this blogpost as well:
http://dbmsmusings.blogspot.com/2010/04/problems-with-cap-and-yahoos-little.html

Berco

On Jul 8, 5:12 pm, Bram Noordzij <bram.noord...@gmail.com> wrote:
<snip>

[Mathijs Homminga]

Hi,

I think those short papers and the two blog posts will make excellent food for discussion.

We can have the third meeting (Wednesday 21st) at the Kalooga office (Helperpark 288, Mediacentrale 3rd floor).
I'm on vacation, but Ivor will be your host that day.

Cheers,
Mathijs

[Bram Noordzij]

The dbmsmusing blog was indeed an interesting read!

I won't be able to attend the next meeting, unfortunately (or not, vacation). Hopefully, someone will share his or her meeting notes somehow.

cu
Bram