[OpenAFS] Web Browsers

[Rubino Geiß]

Strange: 100000 * 1K == 100 * 1000K ~~ 100M
Which is obviously less than 4MB

Did you mean 4GB or was the cache size never the problem.

In the case you meant 4GB instead of 4MB, then you "correctly" will
experience a freeze of your browser: AFS implements write on close. That
is, all data is guaranteed to be written on the server disk as soon as
the client operating system returns successfully from close(). In the
meantime nothing is guaranteed. On most occasions you don't "feel" this
fact. But if you write a lot of data, say > 100MB, from a fast source to
a not equally fast server / network you will get a significant pause at
the time of closing this file. The problem is, most applications do not
anticipate that close will take some time, so that multithreading is not
always introduced to get the gui responsive.

> Okay, I checked the cache and this is what I had:
> AFS using 73370 of the cache's available 100000 1K byte 
> blocks. The file was about 4MB so it was larger than the 
> cache. So I increased my cache by 10x to: 1000000 1K byte 
> blocks. and re-tried the download. I noticed the cache usage 
> did not increase much. AFS using 77294 of the cache's 
> available 1000000 1K byte blocks. Then I noticed network 
> activity continued when the browsers were hung. So I waited 
> several minutes and the download eventually completed. It is 
> strange that during that 
> time I
> cannot stop the download or the browser without taking 
> aggresive action. Yet it appears after it does finally 
> complete the browser returns to normal.
>