There is already a brilliant reply / rebuttal:

http://yz.mit.edu/wp/infrequently-asked-questions-on-deterministic-distributed-transaction-management/


Have fun,
Paulo Gaspar

On 2010-09-07, at 17:46, Ann W. Harrison wrote:

> paulruizendaal wrote:
>> There is some interesting discussion around a scale-out database paper over at highscalability:
>> http://highscalability.com/blog/2010/9/1/paper-the-case-for-determinism-in-database-systems.html
>>
>
> Actually, what Thomson and Abadi are describing is VoltDB - aka
> H-Store, from Dr. Michael Stonebraker. The documentation is available here:
>
> http://community.voltdb.com/downloads
>
> It's an interesting system (sort of). All transactions are
> precompiled stored procedures and run serially on a partitioned
> database. Yes, that's right, serially. One at a time. As long
> as your procedure runs in one partition, other procedures run
> in parallel on other partitions. If not, not.
>
> The definition of their subset of SQL is in the appendices to the
> UsingVoltDB manual. No subqueries, no outer joins, no union ...
>
> Here, for example, are the restrictions on CREATE VIEW.
>
> The following limitations are important to note when using the
> CREATE VIEW statement with VoltDB:
>
> � Views are allowed on individual tables only. Joins are
> not supported.
> � The SELECT statement must obey the following constraints:
> � There must be a GROUP BY clause in the SELECT statement.
> � The first columns listed in the GROUP BY must match the
> initial columns in the SELECT statement in both name and
> order.
> � SELECT must include a field specified as COUNT(*).
> Other aggregates (SUM or COUNT) are allowed following
> the COUNT(*).
>
>
> ------------------------------------
>
> Yahoo! Groups Links
>
>
>



[Non-text portions of this message have been removed]