Software Define Storage for Dockers

I was reading a little bit about Dockers and how Dockers works and why it is so popular.

The most interesting is stories I found about Dockers, was how it sharing information between the containers and how one container almost could have almost 0 byte of information accept metadata. When you deploy that container it will automatic pull in data from all other containers where the actual data are located. Almost like BitTorrent but internally in the Dockers application.

My thoughts did then direct start thinking of how that will impact the storage system.

Will it generate a lot of IOPS?

How should I scale my storage system if it does?

Can I use cheap NL-SAS storage for this?

Can I use Network Attach Storage, NAS?

Can I use any Scale-Out solution?

A lot of questions comes up and as more I read about Dockers, I can’t find any information about this from HP, EMC, IBM, Hitachi or any other storage vendor. No one explains how to solve this.

Why?

Is this not an issue?

Or does no one have a solution for it?

Or do we only accept it this way? Like Hadoop. Most people just accept it cost more hardware then necessary because someone said it.

After some research did I finally find one website that sort of explaining the issues that I was expecting, and there is two interesting websites that inspiring me to write this blog, one from StorageSwiss and the other one from a IBM Developer Blog.

When I read does blogs, do I realize that Scale-Out filesystem is the perfect solution for Dockers, and be able to scale your filesystem in the way that fit Dockers and how it's pulling in data from each container to be able to deploy the software fast.

If we think of how a traditional disksystem works will you only get a limited number of disks to work with, and also with traditional RAID system will never perform faster than the slowest disk.

When Dockers will start building up the container for distribution, you disksystem will be the bottleneck for your performance.

 

If we do the same solution but using software define storage solution such Scale-Out NAS or a Scale-Out Filesystem you can then see the containers will then spread out on multiple disk systems and it can help each other much better, and you can then build your container much faster.

Then when get to a discussion, what is the different between a Scale-Out NAS and Scale-Out Filesystem.

If I referring to the other blog, will we then see following issue with a Scale-Out NAS that the problem is not the product itself but the NFS protocol will then be the bottleneck. Then we come to a question if the NFS protocol is good enough, that's another question.

By using a Scale-Out Filesystem such IBM Spectrum Scale (GPFS) will you then also eliminate the NFS limitation and will then get the full speed from your software define solution and not limited to a single protocol.

If you are setting up your Software Define Storage solution correct then will you end up with less hardware, better performance and easier to grow with your Dockers ideas.

I hope you get inspired of this and are ready to build your own Dockers solution based on SDS.

Thanks for reading my blog, but I recommend that you also read following links if you got time.

 Storage Swiss: http://storageswiss.com/2015/08/31/docker-and-storage-understanding-docker-io/?utm_content=buffer178ec&utm_medium=social&utm_source=twitter.com&utm_campaign=buffer

IBM Blog:

https://www.ibm.com/developerworks/community/blogs/storageneers/entry/Is_a_scale_out_NAS_system_the_same_as_a_scale_out_file_system?lang=en