river-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dan Creswell <dan.cresw...@gmail.com>
Subject Re: transient tuple space
Date Fri, 22 Feb 2013 09:38:01 GMT
Blitz has users across both modes (persistent and transient) so I
wouldn't say there's a clear divide.

Where a tuplespace is not a system of record and contents can be
re-generated or are time sensitive (trading data) and will be updated
shortly, transient tends to win.

Those treating the tuplespace as a system of record (data storage
scenarios) or needing persistence of a workflow (e.g. for complex
calculations) etc use persistent mode.

Note that in both modes, transactions are used as much as they aren't,
because it's really about a level of atomicity as much as or instead
of durability.

On 22 February 2013 09:16,  <fritsch@joerg.cc> wrote:
> Thanks Dan,
> how is it used in practice? Do transient deployments prevail?
> --Joerg
> -----Original Message-----
> From: "Dan Creswell" <dan.creswell@gmail.com>
> Sent: Thursday, February 21, 2013 6:50pm
> To: dev@river.apache.org
> Subject: Re: transient tuple space
> A bit of history:
> Jini (now River of course) went to great lengths to separate
> implementation details from service abstractions. That is the "how"
> was not nearly as important as the "what". "Do one thing well" was a
> phrase oft-quoted in respect of services.
> And so to JavaSpaces:
> The spec basically says that if the JavaSpace is a transient
> implementation, it will forget its contents at the time of exit (crash
> or normal). Persistent is the opposite, contents are remembered across
> exits (crash or normal subject to transactional constraints of
> course).
> Given the above, a transient space could actually make use of disk
> perhaps as an overflow for memory. Thus (b), below is to my mind the
> "correct" interpretation of the spec.
> On 21 February 2013 20:28,  <fritsch@joerg.cc> wrote:
>> All,
>> what is a transient mode/tuple space by definition. To me the documentation is not
quite clear whether it is
>> a. an in memory tuple space that does not use slow disk space
>> b. a tuple space that may or may not use disk space for storage but that is in any
case not preserved after a restart
>> --Joerg

View raw message