"Seagate introduces a new drive interface: Ethernet"

[snowhiker]

The Ars Technica article here.

 

[Mercutio]

So... the drives aren't by themselves going to be smart enough to do anything but store data and there's no underlying concept of a filesystem other than what KOSP applications expose? The data just magically exists and we can totally just trust that without having to have any understanding of where or how even though this is theoretically something we'd be deploying in-house and be responsible for managing? I think I'm going to need more than a two-page article on Ars for this one.

 

[Howell]

For some reason im having trouble understanding the Ars article. From the press release It sounds a lot like NAS on drive with a simplified interface between the two because they can use custom firmware on the drive.

 

[Howell]

It is probably closest in concept to iSCSI but with thinner overhead and storage web instead of point to point.

 

[Mercutio]

I'm thinking there's going to wind up being a drive controller someplace, probably in some little box that looks about like an ethernet switch, but the drives can be distributed to where ports are. So you pay Seagate for a drive controller instead of Synology or EMC or whomever. I imagine that the drives will still go in concentrators of some sort, like 12 or 24 bay chassis with say 2 or 4 ethernet connections per chassis since at some point the cost for connecting those ports is going become a burden to the infrastructure guys and it's not like anyone is going to want the extra headache of managing crap tons of extra switch ports or IP ranges just for each single drive.

 

[ddrueding]

The only advantage I'm envisioning would be distributing your storage array throughout the racks of servers, allowing your existing distribution (cooling, generators, etc) to also give some protection tot he storage system.

 

[Handruin]

This idea looks like the consumer version the ever-growing trend of object-based storage in the cloud. As storage sizes continue to increase and the amount of data we store continues to grow, the more of a pain it becomes to manage. Object based storage removes your need of worrying about the underlying filesystem and architecture and just use it as a hole to put your data. The first thing i thought of before reading the article was Amazon's S3.

What would be great is if they have or eventually have a QoS algorithm built into their KOSP so that we can mix and match SSD and spindles along with data redundancy and eventually not worry about making as many backups, etc. Through in the ability for snapshots and you have your own little datacenter-like storage platform in your house. Next, add in remote capabilities to store the data at a friend's house or a data facility for off-site protection.

Assuming they can manage the performance in a decent way, this could be a great way to manage storage in your PC.

 

[Mercutio]

Even if storage is abstracted for end users, someone, somewhere is going to have to care about its distribution and management. From the end user perspective the only issue with network storage is "Is there enough space for me to save this thing?" and that's been true for my entire professional life.