No SQL Summer Session #2

[cclark]

Thanks to everyone who came out last night. We had 14 people in
attendance at the Pulse Energy offices for a meeting which started at
6:05 and ended around 7:30. The conversation continued afterwards as
some of us grabbed dinner and a beer.

I'll respond to Dethe's message with my notes for the meeting but I
wanted to throw out a few questions to get the planning for the next
session under way.

- The next session is currently scheduled for Monday, June 28. Are
there any strong feelings to move to a different night of the week or
does anyone know of other local user group sessions this might
conflict with?

- Are there any motions to start earlier or later? If not 6:00 it is.

- Suggestions for the next paper and/or blog articles to read? The
few comments I captured last night were:
* The Dynamo paper is a must read at some point in the summer
* It would be good to establish a baseline understanding of all the
terms and different types of technologies
* A survey NoSQL landscape would be great and there seemed to be
particular interest in Cassandra.
Take a look at the list of papers on the NoSQL Summer website:
http://nosqlsummer.org/papers

thanks,
chuck

[Dethe Elza]

On 2010-06-15, at 7:29 PM, cclark wrote:

> - Suggestions for the next paper and/or blog articles to read? The
> few comments I captured last night were:
> * The Dynamo paper is a must read at some point in the summer
> * It would be good to establish a baseline understanding of all the
> terms and different types of technologies
> * A survey NoSQL landscape would be great and there seemed to be
> particular interest in Cassandra.
> Take a look at the list of papers on the NoSQL Summer website:
> http://nosqlsummer.org/papers

There are a few things that didn't make the official list that I thought I'd throw out there:

Relational Databases Considered Harmful, H. Baker (1991 letter to the ACM Forum)
http://home.pipeline.com/~hbaker1/letters/CACM-RelationalDatabases.html

Some good documents on tuplespaces came out of IBM's TSpaces project, but the best one, "A Universal Information Appliance" has disappeared behind a paywall. This one, is shorter and more abstract, but introduces the topic which I think predates and anticipates many features of the current crop of NoSQL systems.

Shirt Pocket Transactions
http://www.almaden.ibm.com/cs/TSpaces/papers/shirt.html

Before relational databases there were object databases, and some are still around (Gemstone, ZODB). This paper argues their cause:

The Object-Oriented Database System Manifesto
http://www-2.cs.cmu.edu/afs/cs.cmu.edu/user/clamen/OODBMS/Manifesto/htManifesto/Manifesto.html

Prevaylent Persistence is not a database as such, but covers some of the same ground as the NoSQL databases do. Again, hard to find one good article, but their quick intro provides a summary:

http://www.prevayler.org/wiki/

These are not about the current (rapidly changing) state of the art, but more about how we got here. I don't know how much others are interested in the historical context behind the current popularization of NoSQL, but thought I'd put these out for consideration.

--Dethe

[Howard Yeh]

> Relational Databases Considered Harmful, H. Baker (1991 letter to the ACM Forum)
> http://home.pipeline.com/~hbaker1/letters/CACM-RelationalDatabases.html

Oh ya, this is a great rant.

[cclark]

Thanks for the additional suggestions Dethe. There haven't many other
suggestions but I had thought about throwing out these from the NoSQL
Summer site:

The End of an Architectural Era (http://nosqlsummer.org/paper/end-of-
architectural-era) - Seems to be a paper to explore our sentiments
from last week that the problem isn't SQL it is the relational aspects
and the rigid ACID properties of these data stores.

Eventually Consistent (http://delivery.acm.org/10.1145/1470000/1466448/
p14-vogels.pdf) - Builds on Pritchett's BASE paper and the discussion
around CAP and the ability to relax consistency requirements

Dynamo (http://s3.amazonaws.com/AllThingsDistributed/sosp/amazon-
dynamo-sosp2007.pdf) - More focused on a specific technology
implementation and a bit more meaty.

If we picked one of the first two then perhaps it makes sense to also
look at one of Dethe's suggestions. My gut tells me the Dynamo paper
is one that would be good unto itself.

Votes? I'd like to send out a schedule for the next paper by this
time tomorrow.

thanks,
chuck

On Jun 15, 8:58 pm, Dethe Elza <de...@livingcode.org> wrote:
> On 2010-06-15, at 7:29 PM, cclark wrote:
>
> > - Suggestions for the next paper and/or blog articles to read?  The
> > few comments I captured last night were:
> >  * The Dynamo paper is a must read at some point in the summer
> >  * It would be good to establish a baseline understanding of all the
> > terms and different types of technologies
> >  * A survey NoSQL landscape would be great and there seemed to be
> > particular interest in Cassandra.
> > Take a look at the list of papers on the NoSQL Summer website:
> >http://nosqlsummer.org/papers
>
> There are a few things that didn't make the official list that I thought I'd throw out there:
>

> Relational Databases Considered Harmful, H. Baker (1991 letter to the ACM Forum)http://home.pipeline.com/~hbaker1/letters/CACM-RelationalDatabases.html

>
> Some good documents on tuplespaces came out of IBM's TSpaces project, but the best one, "A Universal Information Appliance" has disappeared behind a paywall. This one, is shorter and more abstract, but introduces the topic which I think predates and anticipates many features of the current crop of NoSQL systems.
>

> Shirt Pocket Transactionshttp://www.almaden.ibm.com/cs/TSpaces/papers/shirt.html

>
> Before relational databases there were object databases, and some are still around (Gemstone, ZODB). This paper argues their cause:
>

> The Object-Oriented Database System Manifestohttp://www-2.cs.cmu.edu/afs/cs.cmu.edu/user/clamen/OODBMS/Manifesto/h...

[Howard Yeh]

> Dynamo (http://s3.amazonaws.com/AllThingsDistributed/sosp/amazon-
> dynamo-sosp2007.pdf) - More focused on a specific technology
> implementation and a bit more meaty.

+1 for Dynamo.

I personally would want discuss a few context setting papers about how
people are using nosql in the real world.

What are the situations in which nosql solutions become attractive &
sensible. And what are the extreme conditions where they become
necessary. And many of these are based on relational databases used
non-relationally. This is the space i find most fascinating,

Flickr, on sharding and using a ticket server
http://code.flickr.com/blog/2010/02/08/ticket-servers-distributed-unique-primary-keys-on-the-cheap/

FriendsFeed on schema-less use of mysql
http://bret.appspot.com/entry/how-friendfeed-uses-mysql

Reddit on scaling postgres (cache everything. denormalize everything)
http://carsonified.com/blog/dev/steve-huffman-on-lessons-learned-at-reddit/

Or,

Digg's using cassandra,
http://about.digg.com/blog/looking-future-cassandra

[Ken Pratt]

This is a great overview of the current state of concurrency,
including an intro to the CAP theorem.

"A Dismal Guide to Concurrency"
http://www.facebook.com/notes/facebook-engineering/a-dismal-guide-to-concurrency/379717628919

[cclark]

So from the conversations I had at the last session and most of the
comments in the group it seems like there is an interest in some real
world applications and looking at specific NoSQL implementations.

Let's start with the Dynamo paper [0]

There was a suggestion to add in relatively short blog posts as well
so in that spirit I'd like to suggest the extra credit assignment is
Ken's suggestion
"A Dismal Guide to Concurrency" [1]

There were no suggestions for alternative times and days so we'll keep
it Monday at 6:00 at the Pulse Energy office.

I look forward to seeing everyone and having a lively discussion.

thanks,
chuck

[0] - http://s3.amazonaws.com/AllThingsDistributed/sosp/amazon-dynamo-sosp2007.pdf
[1] - http://www.facebook.com/notes/facebook-engineering/a-dismal-guide-to-concurrency/379717628919