Re: Distributed Transactions
Date: Mon, 31 Mar 2008 18:27:15 -0700 (PDT)
Message-ID: <2afe3754-8712-40ee-9a87-02b49de6cacf_at_i29g2000prf.googlegroups.com>
On 31 mar, 11:36, "Sherif Fadel" <fa..._at_vt.edu> wrote:
> Hi,
>
> I am new to the field so I apologize if this question seems trivial. I am
> currently working on distributed software transactional memory (STM). Since
> database transactions are the semantic inspiration for STM, I was wondering
> about the level of concurrency of distributed database transactions.
>
> Specifically, are subtransactions concurrent or are they sequential? If they
> are sequential, doesn't this prolong the duration of the transaction and
> hence increase the chances of conflicts and hence retries? Has this problem,
> possible "retry thrashing", if you will, been considered in the literature?
> Again I apologize if this question seems trivial, but I am relatively new to
> the field.
How sweet...Rediscovery that concurrency is better handled at database
level and.
talking about reinventing square wheels...
Concurrency handling (serialization) is purely platform dependent. You will not have the same rules if you are using ORACLE SQL Server or DB2. Most transactional concurrency mechanisms are much more complex than a simplystic serialized/concurrent dichotomy. Received on Tue Apr 01 2008 - 03:27:15 CEST