Virtual Performance

In our pursuit of determining the best performance among the leading cloud providers, we conducted a comprehensive comparison of DigitalOcean, Vultr, Linode, and myDEAC. Our comparison was based on their respective virtual machine (VM) offerings in Europe, with VMs sharing common specifications of 1 vCPU, 1 GB of RAM, SSD storage between 25GB - 30GB, shared CPU, local storage space, and operating on Ubuntu 22.04 LTS.

Here are the specifications of each provider’s VM:

To ensure accurate cost comparison, we’ve standardized the pricing by converting the cost of myDEAC, which is listed in Euros, to USD, thus matching the currency used by DigitalOcean, Linode, and Vultr. Please note, the prices we’re comparing are exclusive of VAT.

For our benchmarks, we utilized the Phoronix Test Suite v10.8.4 and executed each test three times for accuracy. Our evaluations focused on the following aspects: the system as a whole, single-core CPU performance, RAM memory capacity, storage speed, and network latency. The tests were conducted on the Ubuntu installations provided by the service providers. While we ensured all updates were installed, no further modifications were made, preserving the default settings.

1. pts/build-linux-kernel - This benchmark measures the time taken to build the Linux kernel using the default configuration.
2. pts/compress-7zip - This benchmark evaluates the compression and decompression performance using 7-Zip’s integrated feature.
3. pts/postmark - Using NetApp’s PostMark benchmark, this test simulates small-file tasks typical of web and mail servers. It sets PostMark to carry out 25,000 transactions with 500 files simultaneously, where the file sizes range between 5 and 512 kilobytes.
4. pts/speedtest-cli - This benchmark uses the open-source speedtest-cli client to assess your internet connection’s upload/download performance and latency against the Speedtest.net servers.
5. pts/pgbench - A straightforward benchmark of PostgreSQL, using pgbench.
6. pts/apache - This benchmark tests the Apache HTTPD web server. It uses the wrk program to manage HTTP requests over a fixed period with a configurable number of concurrent clients.
7. pts/phpbench - PHPBench serves as a benchmark suite for PHP, performing numerous simple tests to benchmark various aspects of the PHP interpreter. PHPBench can be used to compare hardware, operating systems, PHP versions, PHP accelerators and caches, compiler options, and more.
8. pts/stress-ng - This is a stress test for the CPU.

Our comparison focused solely on the technical performance of equivalent virtual machines across the different providers. We did not factor in additional services such as firewall settings, monitoring capabilities, backup policies, alert triggering, etc., offered by these providers. Each of these service providers brings unique additional services to the table. Don’t hesitate to explore and try them out yourself!

We utilized the Python Performance Benchmark Suite (https://pyperformance.readthedocs.io) to conduct performance tests and compare the results across each service provider’s VM. The PyPerformance project aims to be an authoritative benchmark source for all Python implementations, emphasizing real-world rather than synthetic benchmarks and favoring whole applications whenever possible. You can find a detailed description of these benchmarks at https://pyperformance.readthedocs.io/benchmarks.html.

PyTorch, a machine learning framework originating from the Torch library, is widely used in applications such as computer vision and natural language processing. Initially developed by Meta AI, it is now part of the Linux Foundation’s broad project portfolio.