Cloud Blob Storage Performance: A Side-by-Side Comparison
This blog post aims to shed light on the blob storage performance of three leading cloud providers — AWS, Azure, and GCP. We’ll delve into the key metrics of latency and throughput, providing a clear comparison and uncovering the strengths and weaknesses of each provider.
Benchmarking Basics:
Before we dive into the results, let’s clarify how we measured performance:
Throughput: We assessed throughput by uploading and downloading 100MB of data using a single thread. This method provides insights into the raw data transfer speeds of each platform.
Latency: We evaluated latency by measuring the time it takes to upload or download a single byte of data. This metric reveals the responsiveness of each cloud storage service.
We used the following virtual machine instances across the three providers:
AWS (Amazon Web Services): m5.xlarge instance
GCP (Google Cloud Platform): n2-standard-4 instance
Azure: Standard_A8_v2 instance
Now, let’s explore the performance results and uncover the champions of cloud storage!
Latency: How Fast Can You Access Your Data?
Latency measures the time it takes for data to travel from one point to another. Lower latency means faster response times and smoother user experience.
AWS: AWS shines in this area, especially with its Express service. Downloading data is about 2 times faster than Azure and a whopping 6 times faster than GCP. Uploading data is also significantly quicker with Express, about 2 times faster than both Azure and GCP.
Azure: Azure performs well in download speeds, but uploading data is a bit slower compared to AWS (without Express). However, it still beats GCP by a significant margin — 6 times faster uploads and 5 times faster downloads.
GCP: GCP falls behind in terms of latency, with both downloads and uploads being slower compared to AWS and Azure.
In short: If your applications require quick data access, AWS, particularly with Express, appears to be the best choice. Azure is a decent alternative, especially for downloads. GCP needs to improve its latency to be more competitive.
Throughput: Moving Lots of Data Quickly
Throughput measures the amount of data that can be transferred per unit of time. Higher throughput is essential for tasks like media streaming or transferring large datasets.
AWS: AWS takes the lead here, achieving upload speeds 4 times faster and download speeds 3.5 times faster than Azure. Compared to GCP, AWS boasts 2.5 times faster uploads and slightly faster downloads (around 1.2 times).
GCP: GCP performs well in downloads, almost matching AWS and outperforming Azure by 3 times. However, its upload speed is 2.5 times slower than AWS.
Azure: Azure lags behind both AWS and GCP when it comes to moving large amounts of data quickly.
In short: If you need to transfer large amounts of data regularly, AWS, especially with Express, seems to be the best option. GCP is a good alternative for downloads, while Azure might not be the best choice for data-heavy tasks.
Understanding the Latency vs. Throughput Puzzle: The Case of GCP
You might have noticed something interesting: GCP has higher latency but better throughput than Azure. How is this possible?
Several factors could explain this:
Network Optimization for Large Data Transfers: GCP might prioritize its network for handling large data transfers efficiently, even if it means slightly slower speeds for smaller data requests. This can involve techniques like data compression or using special routes for bulk data movement.
Storage Architecture Trade-offs: The way GCP stores data might be optimized for throughput over latency. This could involve using larger chunks of storage, which are great for moving large files but might introduce a small delay for smaller requests.
Demystifying AWS Performance: The Express One Zone Advantage
Curious about how AWS achieves its impressive performance? The secret lies in its Express one zone storage class. Here’s a closer look at AWS storage performance:
Download Latency: Express one zone class significantly outperforms the standard service, with a median latency of 0.00557 seconds compared to 0.01317 seconds for standard. This suggests that Express benefits from enhanced network and storage configurations.
Upload Latency: Similar to downloads, the Express One zone class excels in upload speeds, achieving a mean latency of 0.00858 seconds, significantly lower than the standard service’s 0.01872 seconds.
Consistency is Key: Express one zone class demonstrates less variability in latency, indicating more consistent performance, which is crucial for applications that demand reliability and speed.
It’s clear that AWS Express one zone class, plays a significant role in boosting both latency and throughput performance. This makes it an excellent choice for applications requiring high-speed data access and consistent performance.
Benchmark data
If you’re interested in exploring the full benchmarks and delving deeper into the data, we encourage you to visit the following links:
Throughput Test: https://www.benchops.io/report?id=4c130263-f989-45b8-a2bf-d49955e1fd12
Latency Test: https://www.benchops.io/report?id=5a906b29-1265-4ded-bd6f-d13caf42e5b9
.
