The Birth of LTFS

Changing the essence of tape technology

LTFS 1 Creation: Segment 1


Hi, this is Jay Livens from Iron Mountain and welcome to our podcast where today I'm interviewing Michael Richmond who is a lead architect on the development of LTFS technology. This will actually be a three-part podcast series, initially talking about creation and then we'll talk about the technology, we'll talk about the future. So in this section right now, we are going to talk about creation.

So Michael, thanks for joining today. Why don't you tell us a little bit about yourself.

Hi Jay. I appreciate the opportunity to talk with you today.

I moved to California after completing my doctorate and spent several years working at the IBM Almaden Research Center. One of the projects that I was involved with while I was at IBM was the Linear Tape File System, commonly referred to as LTFS. I had multiple roles on the LTFS project, but in summary I was the lead architect. I wrote the OS X implementation for the Apple platform. And I was primary author of the LTFS format specification document.

Fantastic, thanks. That sounds great. You know, let's jump right into it because I think you can bring some really interesting and unique LTFS insights to our listeners. So let me just start with a simple question. Can you talk about what the original vision that you had for LTFS was, way back in the early days?

Sure. I'll talk about the vision that we had as a team because at the beginning of the work for LTFS, the project grew out of a very simple question that was raised by some people in the IBM tape organization. And the IBM tape organization are the people who research the tape media, develop the tape media in working with the LTO's specification body and who develop the drives themselves. And they came to a few people at IBM Almaden with a very simple question which is: Why can't tape look like a USB drive? Why can't our users work with tape the way they work with a drive that they plug into their PC, because they felt that was limiting the sales that they could make with tape. And out of a few discussions that fell from that, the questions came up of how do we make tape easy to use. And the ultimate answer was, well tape doesn't have a file system on it. When you put data onto a data tape, you're only storing what we refer to as the “essence” of the file. This is the image data that is stored in your JPEG file but not any of the other things that go along with making that data an actual file. So the additional things that make it a file are commonly referred to as file metadata. This would be the file name, the time when the file was created, the owner, and things like the size of the file as distinct from the file essence which is the actual image that you see that you see when you open it in Photoshop or another image-editing application. So that was the initial vision. We wanted to address that problem. And very early on we settled on trying to implement a file system on top of data tape. This is an idea that has been done before over many times of the years and various research communities and by a couple of companies, but the previous solutions never really got adopted.

Once we had this idea of trying this, we decided to build up a prototype and show, see ... well, once we got the prototype working see whether we actually could get any customer interest. And that sort of, that prototype demonstration went well. And then we stopped and we reset and we started moving into the questions of how do we really productize this?

The Speakers:

Michael Richmond
Jay Livens