[OpenAFS] New to OpenAFS
Wed, 19 Mar 2008 19:09:20 -0400
> I'm new to OpenAFS and was hoping if the community could help me determine
> if it would be a good fit for my company. We are approx 150 people, with 50
> home users and the rest in small offices of about 10-15 people. I would like
> to have a main file server that everyone can access, but also departmental
> servers in offices that would allow people to save files quickly (without
> going over the WAN).
> In my dream scenario, I would have one main server in our data center that
> stores everything and is backed up. Each office would have a server that
> acts as a caching server. These servers would cache specific folders that
> are often used by the people in that specific office...any times changes are
> made, it is saved on that server and then transferred to the main server in
> the background. This would allow users in branch offices to have fast access
> to files and not have to wait for WAN transfers. And would prevent the admin
> team from backing up every department's server. Is this possible with
> OpenAFS? I read something about a Samba/OpenAFS gateway, would this be
> required? Or are there other products that you think might be a better fit?
> Also, is the windows desktop client used a lot of most deployments? How
> reliable is this?
OpenAFS may be a good fit for your company, but it doesn't fit your
ideal scenario. It does provide a unified way to manage fileserver in
different locations. He is one possible way to use OpenAFS with
Each office has a fileserver. all employees have a home directory and
access to shared directories.
Each employee's home directory would be located on the fileserver for
Here is an sample file tree:
In this example, alice & bob are in new york and charlie & david are in
chicago. Anyone at any office can login to AFS and see any of the file
from any site. Alice & Bob's files are located on a server in new york
and charlie & david's files are on the server in chicago. In addition,
the shared folders for new_york & chicago on servers in their respective
cities as well. The procedures directory is on a server in the HQ in
Atlanta with read-only replicas in all offices. The replicas are only
updated when an IT admin runs a command to refresh the Read-only copies.
Anyone modifying the files can read the read-write copies in Atlanta by
looking in /afs/.example.com/shared/procedures, but the changes are only
copied to all sites when the the IT admin says so.
Each night/week, copies of the data from the new_york and chicago
offices are copied to the central servers in Atlanta. (a scripted vos copy)
Access to files is fast because location-specified data is at each
office. company-wide read-only data has a copy in each office and the
local cache on each client caches copies of any fetched data.
The best part is that no one needs to be notified if data is shuffled
around. For example, the Albany office man not be big enough to need a
server and just uses the new_york server, but they still have an
"albany" shared folder. When the albany office grows, you can set up a
server there and move the volumes from new_york to albany without
changing the paths to the files because it's all transparent to the user.
An added benefit is that alice's data can be moved to Atlanta when she
get relocated but the path won't change.
You can get a feel for the speed of AFS by installing an AFS client and
browsing around some cells like openafs.org.
There is no need for the samba/AFS gateway. Just install OpenAFS clients
on all of the machines.
The windows client is very reliable. At my work, we have over 1000
windows XP desktops running the OpenAFS clients with 10+ servers, 100
solaris desktops, 20 Linux desktops, and 3TB of data which we plan to
expand to 10TB this year. I'm writing this on my Mac laptop with the
OpenAFS client which I use to access my files from home.
Did I answer your questions? Do you have any more questions?