Caching is just part of it, to ensure faster download on nearby nodes, the revolutionary thing is that even 100 GB image can be started in a few seconds, and files streamed when accessed.
It's not a new idea, I remember at least one project from a few years ago, which used a modified Docker engine to download accessed files on-the-fly, but this is the first time a cloud provider offers it.
I feel like "a hold my beer" version of this just accepts webhooks from your artifact registry and pre-pulls new versions of running images, or maybe images that node has seen in the last X hours
> The way Kubernetes traditionally works when scaling up your application is that the entire container image must be downloaded onto the node before the application can boot.
The way backups traditionally work .... is that the entire backup must be restored onto the system before the app can start using the data.
But data protection vendors improved on that many years ago (example: Veeam Instant Recovery - https://www.veeam.com/instant-vm-recovery.html), so Google's image streaming is hardly revolutionary. Reads of blocks not yet available go to the repo and as blocks get downloaded new reads directed to local FS. Just like with modern Data Protection software.
Some of you may recall that French startup that did this close to a decade ago (based on a modified Bittorrent client). They got acquired by somebody (Docker?).
5 comments
[ 2.9 ms ] story [ 20.1 ms ] thread...
> caching
There was a time when Google did more engineering than marketing.
It's not a new idea, I remember at least one project from a few years ago, which used a modified Docker engine to download accessed files on-the-fly, but this is the first time a cloud provider offers it.
The way backups traditionally work .... is that the entire backup must be restored onto the system before the app can start using the data.
But data protection vendors improved on that many years ago (example: Veeam Instant Recovery - https://www.veeam.com/instant-vm-recovery.html), so Google's image streaming is hardly revolutionary. Reads of blocks not yet available go to the repo and as blocks get downloaded new reads directed to local FS. Just like with modern Data Protection software.
Some of you may recall that French startup that did this close to a decade ago (based on a modified Bittorrent client). They got acquired by somebody (Docker?).