General Discussion

Using NetApp storage in a Content Delivery Network



First of all, apologies if this is the wrong place to post this.

I have a fairly simple question.

Is a NetApp storage array a good solution for a content delivery network?

I recently joined a company that provides content delivery to lots of different customers.

The current setup is lots of servers in POPs around the world.

Each server has a lot of local storage for caching content.

Some of the storage is shared using GlusterFS.

I'm investigating what benefits a local storage array might provide.

The primary requirement is (very) high I/O throughput.


Any and all advice appreciated.








NetApp has a few products, including ones that can be used as front-end Object Storage solution - StorageGRID, and Object storage support on ONTAP (The unified storage system that runs on Disks, SSD and NVMe). This can maybe co-exists next to your existing front-end solution  or even integrate with your own Object storage system. 


Another that is low-price system that can be direct-attached (like a JBOD you maybe have today) - it called the "E-Series" but might also be overkill for you (see below).


And it has the enterprise unified product - that I believe most people associate when they hear "NetApp" or "Storage" that sit at the back-end as block or file storage. And for that part - I think It's a very hard question to answer. Yes, NetApp unified storage can store data "efficiently" and serve it "fast" to the CDN front-end.


But as hinted the last two might be overkill. As I think that the application being a CDN actually conflict a bit with what the main advantages NetApp and other enterprise storage focus on - and that's the data protection and availability;  starting from the Disk layout (RAID), Local snapshots (fast point-in-time-restore), remote replication, and other resiliencies (communication, power, secondary node).

I assume because the CDN only cache data - you don't need any of these features as currently if a node fail you probably just spread the workload on the other nodes, so you will just be paying overhead on software and hardware.


This is all theoretical, I don't really know your product and software stack, it may fit perfectly and solve a problem you currently have, and it might be a big overhead that is not really needed - for that, you best contact a sales person at NetApp or a partner, and present the problems you wish to solve. 

Gidi Marcus (Linkedin) - Storage and Microsoft technologies consultant - Hydro IT LTD - UK