Entry-Level Laptop Surprises in Heavy Database Test
An unexpected competitor has emerged in the realm of heavy database workloads. A 512GB entry-level laptop recently went head-to-head with a variety of high-capacity cloud servers. The showdown was orchestrated by a tech enthusiast, who was keen to test the laptop's limits. The results were quite surprising.
Challenging Cloud Servers with Quadruple Memory
The tech enthusiast shared his findings in a blog post titled "Big Data on an Affordable Laptop". He used two different benchmarks to push the laptop to its limits: ClickBench and TPC-DS.
ClickBench focuses on aggregation and filtering operations and features 43 queries. These operations are conducted on a substantial table with 100 million rows, which consumes about 14 GB when converted to Parquet and a hefty 75 GB in CSV format.
On the other hand, TPC-DS involves a more complex setup with 24 tables and 99 queries, including intricate features like window functions. While there is skepticism about the value of TPC-DS results due to repeated optimizations, it still provides a worthwhile challenge.
The laptop was pitted against two powerful cloud instances in all tests. The first was equipped with 16 vCPU cores and 32 GB RAM, and the second boasted a staggering 192 vCPU cores and 384 GB RAM.
The Unexpected Results
For the ClickBench benchmark, the tech enthusiast conducted two tests: a cold run, to evaluate performance with empty caches, and a hot run, to assess performance with filled caches.
In the cold run, the laptop surprisingly outperformed both cloud instances by a significant margin, completing all queries in under a minute. This was up to 2.8 times faster than the cloud servers.
However, this victory does come with an asterisk. The cloud instances use network-attached disks to access the database, which significantly affects query runtimes. The laptop, on the other hand, benefits from a local NVMe SSD. Although it isn't top-tier, it still provides relatively speedy access on the first read.
The tables turned during the hot run test. The more powerful cloud instance completed the run in 4.35 seconds, followed by the less powerful one at 47.86 seconds. The laptop came in last at 54.27 seconds, which was still about 10% faster than its cold run time.
Despite this, it's worth mentioning that the laptop still outperformed the less powerful cloud server in terms of median query runtimes. Even though the laptop's total runtime was about 13% slower, it was impressive considering the cloud server had 10 more CPU threads and four times as much RAM.
Strong Performance Continues in TPC-DS Benchmark
The results of the TPC-DS benchmark were equally as impressive. The laptop swiftly completed most queries with a median query runtime of 1.63 seconds and a total runtime of 15.5 minutes at SF100.
At SF300, the laptop's memory limitations began to surface. Although the median query runtime remained decent at 6.90 seconds, it was evident that some queries were going to take a while. Most notably, one query took 51 minutes to complete. Nonetheless, the laptop successfully completed all queries in 79 minutes, showcasing the resilience of its hardware and software.
Interestingly, this wasn't the first time the laptop's chipset was put to the test. Previously, the same chipset was tested at an extreme temperature of -50ºC inside a bucket of dry ice, where it completed the run in 478.2 seconds.
In conclusion, the results of these benchmarks demonstrate that even an entry-level laptop can hold its own against powerful cloud servers in a heavy database workload test. This serves as a testament to the ever-improving efficiency and performance of modern technology.