[OpenAFS] chunk behavior and cache size limits?

[Jason R. Banfelder]

Hello,

I am evaluating AFS (currently in the reading stage) and have a couple 
of questions regarding the applicability of AFS to environments with 
large-ish files (some files in our workflows are ca. 250 GB) that I am 
hoping the community may be able to help with...

First, does AFS impose any limits on the size of the client disk cache? 
I am most concerned about x86_64 Linux clients, where I don't see any 
mention of limits in the man pages. I did find mention of limits on some 
(older?) references the Windows client on the web, but it was not clear 
if they were AFS imposed limits or just local policy. I'd be looking at 
a 1 to 2 TB cache on Linux clients -- is that supported/consistent with 
best practices?

Second, is there any description of chunking behavior that I can read up 
on? The asfd man page states that "AFS uses partial file transfer", but 
it is not clear in which direction(s?), and if this only the RPC level. 
If I run the UNIX head command on a 250GB file, will the whole file be 
transferred to the client's cache (this would take multiple minutes in 
our environment) or just the first chunk? If the whole file is 
transferred, would the head command be able to return the first ten 
lines after the first chunk arrived on the client's cache, or will the 
read block until the whole file is transferred?

Thanks in advance for your help and consideration.
--Jason