Do not ask my colleagues either. A quick survey of a half dozen of them turned stares up at the reference of the Andrew File System, a technology named after Andrew Carnegie and Andrew Mellon and developed from the early 1980s. But since the Association for Computing Machinery’s award will signify, AFS is worth considering as a technology which paved the way for applications and cloud techniques.
Mahadev “Satya” Satyanarayanan, a Carnegie Mellon University Computer Science professor that had been part of the AFS team, answered a handful of my questions via email concerning the origins of this safe and scalable distributed file system, the importance of it, and where it stands today. Satyanarayanan was known by ACM along with John Howard, Michael Leon Kazar, Robert Nasmyth Sidebotham, David Nichols, Sherri Nichols, Alfred Spector and Michael West, who worked as a staff via the Information Technology Center partnership between Carnegie Mellon and IBM (the latter of which incidentally financed this ACM prize).
Is there some way to measure how widespread AFS usage became and that forms of organizations used it most? Any feeling of how far for what, also it is still used?
Within a approximately 25-year timeframe, AFS has been used by several U.S. and non-U.S. schools. National labs, supercomputing facilities and associations have also used AFS. Companies in the financial sector (e.g., ” Goldman Sachs) and also other businesses also have used AFS. A beneficial snapshot of AFS deployment was supplied by the newspaper “An Empirical Evaluation of some Wide-Area Distributed File System” which appeared in ACM Transactions on Computer Systems in 1996. That newspaper states:
“Originally intended as a remedy to the computing needs of the Carnegie Mellon University, AFS has expanded to unite approximately 1000 servers and 20,000 customers in 10 nations. We estimate that more than 100,000 users use this system worldwide. In geographical span as well as in number of users and machines, AFS is your biggest distributed file system which has ever been assembled and put to serious use.”
Figure 1 from that paper indicates that AFS spanned 39 cells outside the USA, 22 commercial cells, 11 tissues, and 59 cells at the time of the snapshot. Within respective organizations, there were many non-federated deployments of AFS in addition to this big federated multi-organization deployment of AFS.
What has been AFS’s biggest impact on today’s cloud and enterprise computing environments?
The version of storing data from your cloud and delivering parts via on-demand caching of it at the border is something everybody. That version was initially conceived and demonstrated by AFS, and that is its impact. It simplifies handling complexity for staff, while preserving performance and scalability . In the perspective of end users, the ability to walk around any machine and use it along with your own supplies huge flexibility and advantage. The data that’s specific to a user is sent on demand within the community. Keeping all of of the machines which you use becomes trivial. Users discovered an ability. Really, it was this ability that motivated the founders of DropBox to begin their firm. They wanted to enable scale this effortless ability to stay all of of the machines used by someone, also had used AFS at MIT as part of the Athena environment. Many of implementation techniques and the principles of AFS have affected many other systems within the previous decades.
Just how can AFS come to get created in the first location?
Back in 1982, CMU and IBM signed a collaborative arrangement to create a “distributed personal computing environment” in the CMU campus, which could later be commercialized by IBM. The collaboration began in January 1983. A fantastic reference for information regarding these early times is the 1986 CACM paper by [James H.] Morris et al entitled “Andrew: A Distributed Personal Computing Environment”. The context of the arrangement was as follows. Back in 1982, IBM had introduced the IBM PC, that proved to be somewhat successful. IBM was fully aware that use of private computing required the capability to share information safely, and with proper access controls. This was possible from the timesharing systems which were leading in the early 1980s. The way to achieve this at a enterprise’s world wasn’t evident in 1982. A portion of the collaborative arrangement was supposed to develop a remedy for this problem. More than half of the first year of the Information Technology Center (1983) has been spent in brainstorming on how best to achieve this goal. Through this process, a distributed file system emerged by about August 1983 as the mechanism to get enterprise-scale information sharing. The way to apply such a distributed file system became the focus of the efforts.
What will the AFS founders have done differently in building AFS when they had to do it over again?
I can think of at least two items: one big and a small.
The small issue is that the design and early evolution of AFS happened before the emergence of [network address translation (NAT)]-based firewalls in networking. These are in widespread use today in homes, small businesses, etc… Their presence makes it difficult for a host to be able to set a callback channel to initiate contact. If we had grown AFS after the widespread usage of firewalls that are NAT-based, we would have closely rethought how to execute callbacks.
The thing is to do with the World Wide Web. The Mosaic browser arose from the early 1990s, and Netscape Navigator a little. By then AFS had been for several decades in existence, and has been in use at several places. Had we realized just how precious the browser would eventually behave as a tool, we would have paid attention for it. For instance, a browser may be utilized in AFS by using “file://” instead of “http://” in speeches. Each of consistence-maintenance machines that’s constructed to AFS and the caching would subsequently have been available through a tool which has eventually been shown to be hugely valuable. It is possible that the browser and AFS might have had a more cyclical evolution, since HTTP and browsers eventually did.
Appears like perhaps you will find remnants of AFS alive from the open source world?
Really. OpenAFS is still a busy open source project. Many associations (including CMU) continue to use AFS for manufacturing usage, which code is currently according to OpenAFS.
My job onto the Coda File System forked away from the November 1986 model of AFS. Coda has been open-sourced from the mid-1990s. This code base has been functional and alive today. Buried in Coda are code from AFS that is ancient and thoughts.
Do some of you have some spectacular strategies for what they will do with the prize money?
Nothing concrete yet. We’ve discussed donating the funds to some charitable cause.
Connect the Network World communities on Facebook and LinkedIn to comment on matters which are top of mind.