Office of Information Technology

Network File System, Lead - Mark Edgar

Members: Roger Siggs, Steve, Cyril, Dale

March 9, 2006

I've finally determined the correct configuration for getting access-based enumeration working on DFS. I know have some idea of how much work is required to automate the creation and maintainance of DFS roots and links.

I also have the beginnings of a "research summary" paper, which purports to summarize the research and experimentation that has been done so far with both Microsoft DFS and AFS.

http://glory.gc.maricopa.edu/~medgar/fs.html

Several areas of research have not yet been started:

Evaluating the Arla AFS client on the Macintosh.
Evaluating quota on NTFS.
Evaluating storage solutions.
Determining hardware requirements.
HTTP and FTP access to the file system.
Web publishing solutions.
Integrating student UNIX accounts with AFS and Active Directory.
Integrating email service with Communigate and AFS.

A backup solution is a big part of the file system solution; I don't yet know which team will now be responsible for this work.

OpenAFS 1.4.1 was installed on the RedHat EL4 box (afs2.paldream.edu) and it is now fully participating in the AFS file system on the test network.

March 2, 2006

Today, Dale received the requested AFS file server box and installed the OS on it. It will be available for me tomorrow when I will install OpenAFS 1.4.0 (or maybe 1.4.1rc8) for testing.

I'm currently stuck on my evaluation of DFS/SMB with respect to access-based enumeration. It seems that this technology is new and has many interesting quirks. NetWare supports it very nicely, AFS does not support it, DFS/SMB purports to support it, but I have yet to see evidence of this. I don't think we will find an exact replacement for the NetWare "upwards filtering" access-based enumeration.

Josh has obtained a Mac workstation and installed OS X Tiger, so that we can begin testing ADmitMac and Arla.

Theresa and I met today to discuss how new Genesis will be creating and managing home directories and course folders.

The file system team and Roger Bontrager have also been working on a namespace proposal document. I'd like every one to take a look at this document and provide any feedback.

http://glory.gc.maricopa.edu/~medgar/fsdesign.html

February 24, 2006

The file system team has gone back to review the requirements document. Some sort of agreement on this document should be forthcoming.

Dale is installing and configuring a RedHat Enterprise Linux box on which we can install and test OpenAFS file services.

Cyril has initiated a request for a Macintosh client to test AD integration; hopefully this workstation can be used to test SMB/CIFS/DFS accessibility. Specifically, we'll want to test whether the SMB support in OS X can use DFS referrals and whether it can use the Active Directory homeDirectory attribute to locate and mount the user's home directory upon logon.

I have requested a Mac workstation on which to test the Arla AFS client.

A network file name space standard has been proposed, and the team will be working on polishing and presenting a proposal for standardizing the utilization of the network file path name space. A draft is available, but I hesitate to publish it here at least until I get some feedback from my team members.

The test network has been configured with 3 DFS roots (\\paldream.edu\home, \\paldream.edu\shared, \\paldream.edu\apps) with root servers PALAPPS3 and PALWSUS and link referrals to shares on those same servers. A sample home directory layout has been created within the home DFS root (\\paldream.edu\home\a\a through \\paldream.edu\z\z) but no actual home directories have been created. Two sample SoftGrid applications have been copied into the application space, but these have not yet been tested with the SoftGrid client or server. We will soon want to enable and test access-based enumeration within this DFS space which requires installing an additional Microsoft tool on the Windows 2003 servers.

We have yet to discuss solution strategies or more specific requirements for any of the following:

Uptime / fault tolerance / failover
Storage solutions
Backup schedules and disaster recover plans

According to the calendar in HT1-115, there is a deadline date of 28 Feb for this team. At this point I claim ignorance about what tasks we need to have finished by this date. We certainly will not be finished evaluating either SMB/DFS or AFS nor will we be prepared to forward any file service solution proposals by this date, despite any claims I may have made last month. ;)

February 17, 2006

I've installed a NetBSD system with OpenAFS 1.4.0 and Arla AFS client. Authenticating to the domain controller gets you access to AFS. I also have a Debian Linux system running with the OpenAFS 1.4 client software.

Next steps are:

install an AFS Windows client
setup the filesystem layout
create home volumes for test users
test roaming profiles
install an AFS Macintosh client

I need someone to help me with configuring a Mac client. I'll need to know what are the requirements for that client also (so that we can test against them).

I also need to see if anyone on my team would like to start testing SMB and/or DFS.

February 9, 2006

I've not added any new requirements to the requirements document this week. I've not heard any feedback on it, and I've a few questions which need answering. These are evident on the http://glory.gc.maricopa.edu/~medgar/fsreqs.html page.

My research this week has uncovered a commercial AFS support solution (http://sinenomine.net/support/afs) and Windows Server 2003 "access based enumeration" just like NetWare has (http://www.microsoft.com/windowsserver2003/techinfo/overview/abe.mspx)

I'm still waiting for a box on which to install AFS for testing.

February 2, 2006

The Network File System team has started on a set of requirements for a network file system. This working document is available online http://glory.gc.maricopa.edu/~medgar/fsreqs.html and the team is open to any comments and suggestions.

I've proposed that we move discussion of printing services into a new team. There is no technical requirement to use the same product for file and print services, and the requirements for each are almost completely independent of the other. I suspect that the only reason that file services and print services have been lumped together is because this is the way we currently offer these services.

There has also been a question of when to discuss data migration strategies. I believe this should be a separate team also. As its needs should be independent of the particular network file system, such a team can propose data migration requirements separately from file system requirements.