On 24 Nov 2010, at 11:59 , Bram de Kruijff wrote: > 2010/11/23 Martijn van Berkum <Martijn.vanBerkum at gxsoftware.com>: >> Hi, >> >> I would oppose this proposal as SimpleStorage is probably not linear >> scalable and automated clustered, which violates the goal of creating a >> cloud fabric framework. Unless of course we make SimpleStorage compliant >> with those goals. > > Just to clearify the current status of this discussion. At this point > we are not implementing some new custom multipurpose > SimpleStorageEngine as was (sort of) proposed AMDATU-131. Note that it > is resolved as Won't fix. What we are doing is decoupling the amdatu > core services from cassandra by implementing pluggable (and for now by > default simple filebased) storage (AMDATU-130 et al). In a sense we > are just reducing the Amdatu core and dependency footprints and > increasing agility, paving the way for (amongst other things) an > intial provisioning model based on ACE in our next scheduled release > 0.1.0.
To address Martijn's scalability concerns we should make sure that the API we choose for these storages allows us to create an implementation that can scale linearly. > It is true that this does not fully address service fabric / > scalability concerns (yet) but, in all fairness, nor does sticking > Cassandra in their as that only provides (eventually consistent) > persistence. It is not suited (at all) for any form of > (non)transactional (a)synchronous configuration management or > communication. In this area we are just starting to think about how > the service fabric should be designed, but IMHO it is likely to be a > (web) service oriented approach rather then depending on any form of > (eventually consistent) shared state. I guess in this sense local disk > storage is the most scalable approach of them all ;) Agreed, and it's quite possible that clouds will provide a similar virtualized disk store with similar qualities. Greetings, Marcel
