Hi Chas, if you want to 'spawn' independent processes over different machines, terracotta could be an option. But as soon as these processes are going to share state, it gets a lot more complicated since a scalar clock MVCC based stm is not going to be scalable over different machines.
Afaik the focus of Paul was on distributing the Clojure STM using terracotta. On Jul 20, 5:47 am, Chas Emerick <cemer...@snowtide.com> wrote: > I can't respond to that, but presumably these issues are irrelevant if > one were using terracotta to coordinate asynchronous independent > computation, e.g. using agents heavily? > > - Chas > > On Jul 16, 2010, at 8:00 AM, peter veentjer wrote: > > > To repeat myself again: > > > The big problem with a MVCC based STM, is that there is a central > > clock > > that needs to be touched by independent transactions. That was one of > > the > > reasons for me to get not started on a distributed STM. > > > So you will get something up and running on your laptop, but it will > > not be > > of great value in a scalable system (since it limits scalability). > > > The best way forward imho is to remove the central clock either by: > > - a vector clock (most expensive read validations instead of a simple > > long comparison, more complex infrastructure since you don't want to > > have ever growing timestamps in your system). > > - use a different (d)stm mechanism. The cool thing about mvcc/tl2 is > > that you > > can go back to previous history.. with other stm mechanisms this might > > be > > more tricky. > > > Or you could try to enter the nosql domain, so dealing with various > > levels > > of consistency and perhaps broken failure atomicity (since some of > > them only > > guarantee atomicity over a single 'record' and not over records > > spanning > > multiple machines that are modified in a single transaction. > > > On Jul 16, 4:51 am, Alex Miller <alexdmil...@yahoo.com> wrote: > >> Hi, > > >> I used to be a tech lead at Terracotta but I am now a full-time > >> Clojure dev. I think it would be very interesting to explore the new > >> Terracotta Toolkit product to provide a distributed store for Clojure > >> data structures. I think it actually comes out in GA next week > >> although it's been available to try for a while. If only I had 47 > >> hours per day I would be all over giving it a try. If someone is > >> banging on it and needs help, I'm happy to connect you to the right > >> people or give you some pointers. > > >> Also, Sergio is doing some awesome work with Terrastore and I'd > >> highly > >> recommend giving it a shot if it fits your needs! > > >> I'd really like to hear some detailed use cases for what you want out > >> of Clojure/Terracotta integration. Is it shared Clojure data? > >> Locking? > >> Coordination? Caching? Remote code execution? > > >> Alex > > >> On Jul 15, 7:51 pm, Paul Stadig <p...@stadig.name> wrote: > > >>> If anyone is interested, the latest version of my terracotta TIM > >>> is athttp://github.com/pjstadig/tim-clojure-1.0.0andittries to be > >>> a Clojure > >>> 1.0.0 compatible TIM, which shows how its a bit out-of-date. > > >>> I am very open to collaboration, and I would love pull requests, > >>> or any > >>> patches that anyone sends. > > >>> Paul > >>> ____http://paul.stadig.name/(blog) > >>> 703-634-9339 (mobile) > >>> pjstadig (twitter) > >>> p...@stadig.name (jabber) > > >>> On Thu, Jul 15, 2010 at 12:35 PM, scx <mark_addle...@bigfoot.com> > >>> wrote: > >>>> Hi -- > > >>>> I'm noob to both Clojure and Terracotta but if you're willing to > >>>> tolerate basic questions from me, I'd be very interested in helping > >>>> out. > > >>>> On Jul 12, 3:36 am, peter veentjer <alarmnum...@gmail.com> wrote: > >>>>> I don't think it every is going to scale. > > >>>>> MVCC/TL2 based STM designs rely on a central clock, so if you can > >>>>> update the clock in 0.1 ms on all machines, the maximum > >>>>> throughput is > >>>>> 1/0.0001 = 10.000 transactions/second... no matter how many > >>>>> machines > >>>>> you throw at it. Even on a single machine the central clock can > >>>>> cause > >>>>> scalability problems (10/20M transactions/second and this will > >>>>> degrade > >>>>> when you throw more cores at it). > > >>>>> This is one of the reasons I dropped the TL2 approach for > >>>>> Multiverse > >>>>> and switched over to the SkySTM model (with some magic of my > >>>>> own) that > >>>>> doesn't relied as much on a central mechanism. > > >>>>> On Jul 11, 6:50 pm, scx <mark_addle...@bigfoot.com> wrote: > > >>>>>> hi -- > > >>>>>> i've seen paul standig's work with clojure +terracotta. > >>>>>> wondering if > >>>>>> anyone has continued his work? > > >>>> -- > >>>> You received this message because you are subscribed to the Google > >>>> Groups "Clojure" group. > >>>> To post to this group, send email to clojure@googlegroups.com > >>>> Note that posts from new members are moderated - please be > >>>> patient with > >>>> your first post. > >>>> To unsubscribe from this group, send email to > >>>> clojure+unsubscr...@googlegroups.com<clojure%2bunsubscr...@googlegroups.com > > >>>> For more options, visit this group at > >>>>http://groups.google.com/group/clojure?hl=en > > > -- > > You received this message because you are subscribed to the Google > > Groups "Clojure" group. > > To post to this group, send email to clojure@googlegroups.com > > Note that posts from new members are moderated - please be patient > > with your first post. > > To unsubscribe from this group, send email to > > clojure+unsubscr...@googlegroups.com > > For more options, visit this group at > >http://groups.google.com/group/clojure?hl=en -- You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
