Originally published at: https://developer.nvidia.com/blog/high-performance-remote-io-with-nvidia-kvikio/
Workloads processing large amounts of data, especially those running on the cloud, will often use an object storage service (S3, Google Cloud Storage, Azure Blob Storage, etc.) as the data source. Object storage services can store and serve massive amounts of data, but getting the best performance can require tailoring your workload to how remote…
Hi,
Thank you for this blog. I have a question regarding the achieved bandwidth and the published bandwidth of the instance.
Let’s refer to Figure 1. The achieved bandwidth is approximately 12000 MBps (around 12 Gbps), whereas the g4dn.xlarge EC2 instance has a published bandwidth of up to 25 Gbps.
I’m unsure if this achieved bandwidth is considered good. Specifically, I’m wondering why the achieved bandwidth is less than half of the network bandwidth. Could it be due to an S3 limitation (then not an issue in KvikIO), meaning that the full network bandwidth can never be reached when using S3?
(Figure 4 shows a similar behavior for another instance.)