Yes, I agree, but the performance improvement could be one of the values Mnemonic could deliver, as you mentioned the durable object model (DOM) could establishes a new infrastructure for large set of object graph (e.g. IoT data, business models, machine learning/deep learning parameters).
Regarding the durable structures, it would be great useful for user to take fully advantages of Mnemonic for their business and research. as Johnu also mentioned, Mnemonic might provide the data process/analytics capabilities without need developer to be aware of the computing engines and distribute system existing like Hadoop/Spark does. For the performance, by using the existed frameworks to integrated with Mnemonic might not fully exhibit Mnemonic's capabilities, but It could still show some extent of benefits if using Mnemonic in some places (e.g. cache, input/output, checkpoint, internal data persist) In addition, I'd like to add one item to the list you provided. that is the durable native computing, because it breaks the barriers to access object graphs in the same viewpoint from native code and managed code avoid introducing marshal/un-marshal operations. Very truly yours. +Gary. On 3/7/2017 10:10 PM, Debojyoti Dutta wrote: > Johnu: like the durable data structure collection idea. Maybe do solid > benchmarking along the way too. > > Gary? > > thx > debo > > On Tue, Mar 7, 2017 at 9:24 PM, Johnu George (johnugeo) <johnu...@cisco.com> > wrote: > >> I would like to see the durable form of major datastructures. Once >> durable Arrays, Lists, Maps, Sets, Trees(->Graphs) are completely >> performance tested, we should build the distributed layer (which would be >> transparent to the user). That way, mnemonic can handle the growing >> processing/storage needs of the big data applications. >> >> Thanks, >> Johnu >> >> On 3/7/17, 9:06 PM, "Debojyoti Dutta" <ddu...@gmail.com> wrote: >> >> Hi >> >> Mnemonic has a lot of directions for potential impact .... here are >> some >> 1) distributed durable key-val >> 2) durable graph >> 3) integration with other apache projects >> >> Any thoughts on the priority (based on participation)? >> -- >> -Debo~ >> >> >> >
signature.asc
Description: OpenPGP digital signature
