Talk:File server

From Wikipedia, the free encyclopedia

I cut quite a large portion of the original article because I think it is fair to say that the term file server is a specific term that relates to a computer that performs a specific function: viz: the ability to map/mount a drive and treat that drive as if it were part of the computer at which the user is sitting.

Furthermore, SAMBA is not a protocol, it is an implementation of the SMB/CIFS protocol.

In all practical applications, this excludes ftp servers, http servers and the like.Ros Power 21:55, 12 June 2006 (UTC)

[edit] Disk Speed

It is not the case that file servers need to access the file system faster than desktop systems per se. Arguably, desktop systems have heavier disk and performance requirements than servers, due to application loading, swap, caching and the graphical environment. File servers, because they are serving multiple clients, need to be able to retrieve data from different parts of the disk quickly, however. The speed of SCSI disks is incidental - it is a function of their higher quality and greater cost.

Furthermore, it is inaccurate to say that disk speed can eliminate bottlenecks in other parts of the system. The whole point of a bottleneck is that you can't fix it by changing anything else! Ros Power 14:03, 18 June 2006 (UTC)

[Excuse me if I'm not adding to this discussion properly; I'm learning] If this is the case, which it logically seems to be so, I think you should remove the Hard Drive Speed section and add that blurb in to a new section, possibly called 'Theory Behind File Servers'? I would write it myself, but since you have a more instrictic knowledge, it's only proper if you get the credit.

Please remember to sign your contribs with four tilde's (~). I always forget though.

I don't think there's any specific theory behind file servers as such, certainly not from a hardware level. It might be interesting to look at some general principles of the software that serves files, such as maintaining synchronicity with clients, data integrity etc. Unfortunately most protocols and software are closed source, though NFS isn't. It might be worth thinking about.Ros Power

[edit] Disputed tag

I find the current state of the article rather inaccurate/awkward:

• It claims that file servers have been "instrumental" to "pushing the demand for RAM", however, I don't think file servers are any different from any other kind of servers in this respect. Computers not dedicated to a single task, in general, benefit a lot from more RAM for cache (often disk cache), and servers are the area where companies typically don't cut corners. And as far as I can tell, it's the database servers that keep pushing the RAM requirements – it is not unusual to see database servers with 32GiB of RAM these days.
• I fail to see the connection between memory requirements and UPSes.
• "Traditionally, file and print services have been combined on the same computers due to similar computing requirements for both functions." I fail to see any similarities, besides the fact that both require a network. Print servers are very often just embedded devices, or even integrated into printers these days.
• "a user can map or mount a drive or directory so that the directory appears to be on the machine at which the user is sitting." Doesn't make a lot of sense to me. It doesn't appear to be "on the machine", it just appears to be part of the same file system. This might sound like nitpicking, but I really find that claim strange – file systems very often work over the network, even though normal users might not know or realize that.

I personally would remove the "Memory requirements" and "File and print" sections, and rephrase the last bullet. Does anyone disagree? -- intgr 17:13, 28 December 2006 (UTC)

Also, "The Novell NetWarefile server operating system, which dominated the market for file servers at the time of greatest growth in demand, read the entire FAT table into RAM on boot" – I am doubtful that any of Novell's operating systems have ever used the FAT file system. If they have, this probably needs a citation -- intgr 17:21, 28 December 2006 (UTC)