On 10/28/06, Les Howell <hlhowell@pacbell.net> wrote:

If you Firefox or Gimp would go into both, to provide the software on the server for the U install disk, and on the system disk for the R version standalone. A diskless client actually has a disk, and will provide itself with swap space on that (minimal sized) disk. But all the software comes from the server. The diskless client will only be a network bootable. On startup, it will put out its address, the server will respond with a host id and boot location, the system will then boot from the network assigned location. Thus the diskless client needs virtually no software other than a boot image in rom that will boot from the network.

When the network is setup, the server gets loaded with a "U" version, and the clients all boot from the server with their own boot mount point. Their swap can be on the server, but the network overhead would be extreme, and the system would be slow. Therefore the local bootable version for the clients will have a setup file that sets the swap partition on the local disk by default, so that only software loads and data take place from the server, All swaps and temps can be set local, and the system performance will nearly be equal to a standalone system. As you move more of the data locally over time, the swap will eventually have all the data (if the local disk is large enough) and the only time files are exchanged over the network is at boot, at new program starts, and at data retrieval times. There are some other things that the server will manage, such as synchronization of databases, or locks in the case of shared development, but essentially the services are only provided on demand if things are setup correctly.

The local partition for each bootable client is the mount point for the client on the server. It makes all the network traffic look like normal file accesses to both machines the the TCP/IP protocol is in a separate level, removing network overhead from the transfer processes that the server and client have to deal with.

The primary differences between the clients and the server deal with things like local description files that are now part of the client partition on the server, so the U disk is designed not to overwrite these bits. Whereas a standalone server will have these files inside the /etc directory locally, and those will be edited directly. On the server the /etc directory is in the client partition, and each client has its own copy, so that actions on one client will not affect the others. You can get more information on diskless clients from most unix sites. It was quite easy for a while, but as security issues began to crop up, some of the interactions had to be managed better, and I quit following all the details. There may be someone here with more current information who can help you set this up.

This is about the extent of my direct knowledge. I suggest that you read Sun's or BSD's website stuff on diskless clients for more information, or take a class somewhere. I'm sorry, I'm not current on all the low level details.

OK, but if my FireFox and GIMP will go into both, then it will use double space from my HDD. It will be duplication instead of devision. I have asked about how-to devide, not how to duplicate... --------------------------------------------------------------------- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org