That does sound interesting. Can you send the summary to dev@cassandra.apache.org ?
On Thu, Feb 4, 2016 at 7:22 AM, Igors K <igors.km...@gmail.com> wrote: > Hi Jonathan, > > I have the paper, but it is not in English, and I have described the > algorithm in a natural language. Therefore, it would take a significant > amount of time to translate it, but I can write a very short version with > the main points and findings. If there are questions, I will answer them. > > Igors > > 2016-02-03 22:24 GMT+02:00 Jonathan Ellis <jbel...@gmail.com>: > >> Hi Igors, >> >> Sounds interesting! Are you planning a formal writeup? >> >> On Wed, Feb 3, 2016 at 2:02 PM, Igors K <igors.km...@gmail.com> wrote: >> >>> Dear Mr Jonathan Ellis >>> >>> I have developed a load balancing solution for user defined functions >>> (UDF) as part of my study project. The study looks at the problem when UDFs >>> are used for small calculations with different time complexity, and nodes >>> of a Cassandra cluster could be loaded unevenly. >>> >>> The algorithm is designed for the CPU bound UDFs, which are executing on >>> a multi-core CPU. It estimates how long a node of the cluster will be busy >>> with UDFs execution, and it is tested on a client-server architecture but >>> potentially could be used in a P2P architecture. To use the solution, a >>> user provides the function which estimates runtime of the UDF. There are >>> executing many UDFs in parallel, but the algorithm predicts the time when a >>> cluster node will finish executing all UDFs. In addition, the algorithm >>> takes into account the load of read queries and could be extended to >>> evaluate the impact of write queries and other processes. >>> >>> The results are promising, although the implementation of the algorithm >>> does not take into account other CPU intensive processes of the server such >>> as the *garbage collection *process. The solution is verified on the >>> virtual servers of Amazon EC2. It is not production ready, but it could be >>> a good starting point. >>> >>> If there is an interest in the findings, I can extend and contribute the >>> solution. Please let me know if you have any questions. >>> >>> Best wishes >>> >>> Igors Kočins >>> >> >> >> >> -- >> Jonathan Ellis >> Project Chair, Apache Cassandra >> co-founder, http://www.datastax.com >> @spyced >> > > -- Jonathan Ellis Project Chair, Apache Cassandra co-founder, http://www.datastax.com @spyced
