Gengliang Weng, et al, explain that even a single Spark node can be useful:
It’s been a few years since Intel was able to push CPU clock rate higher. Rather than making a single core more powerful with higher frequency, the latest chips are scaling in terms of core count. Hence, it is not uncommon for laptops or workstations to have 16 cores, and servers to have 64 or even 128 cores. In this manner, these multi-core single-node machines’ work resemble a distributed system more than a traditional single core machine.
We often hear that distributed systems are slower than single-node systems when data fits in a single machine’s memory. By comparing memory usage and performance between Spark and Pandas using common SQL queries, we observed that is not always the case. We used three common SQL queries to show single-node comparison of Spark and Pandas:
SELECT max(ss_list_price) FROM store_sales
SELECT count(distinct ss_customer_sk) FROM store_sales
SELECT sum(ss_net_profit) FROM store_sales GROUP BY ss_store_sk
To demonstrate the above, we measure the maximum data size (both Parquet and CSV) Pandas can load on a single node with 244 GB of memory, and compare the performance of three queries.
Click through for the results.