Tummi - Bottleneck Memory Bandwitdh?

I still have no numbers to extrapolate but assuming a whole Wikipedia with millions of articles parsed as meme pool in an RDF structure will need terabytes of RAM then probably memory bandwidth will be the bottleneck for the SPARQL queries. Currently we have up to 64 cores with 8 DDR4 channels with each ~25 GB/s on a single socket system, IBM's enterprise server may have a bit more throughput, but I guess until we have some kind of 1:1 ratio for cores:channels Tummi won't be able to query the whole Wikipedia in reasonable time...still some time to pass.

Comments

SrdjaFriday, 08 May 2020Hmm, an 1:1 core:channel ratio would make only sense if we lower memory latencies too, afaik these are now about 100 ns for RAM, memory wall. And I am not sure what the compute:memory ratio of current SPARQL DBMS is and how my Tummi queries will perform...maybe all is just fine.

SrdjaTuesday, 18 April 2023Did recently read about MLP, memory level parallelism, modern processors offer ~10-way MLP per core nowadays, so it will depend on the concrete SPARQL queries in my implementation...

Pages

Recommended

Posts by year

Tag Cloud

Bottleneck Memory Bandwitdh?

Comments

New comment