nfs vs samba

[orange]

which one is better? is it even possible to use nfs on Amiga?

[MrZammler]

Sure. Genesis (Amitcp) comes with it's own nfs mount app. Not sure whether there is an nfs server for the miggy though.

Pretty stable too, never used samba, but never had any problems with an nfs mount from my linux pc.

[Noster]

Hi,

using the command "netmount" (there is a documentation for this command in the docs directory of AmiTCP resp. Genesis) you could mount any Amiga drive of other Amigas that are in the same network and also running AmiTCP/Genesis. This works really good -- even works with the AmiTCP 4.0 demo from Aminet.
I also used nfs to connect to my Linux server and at my opinion this works better than samba.

Noster

[adolescent]

It really depends on what you want to connect/connect to.  NFS is superior in speed, overhead, etc. but if you need to connect to a Windows box, you'll probably be limited to Samba (unless the PC has additional software).  You can use both at the same time though.

[patrik]

Be sure not to mix samba with smbfs on the Amiga when talking about speed - samba is sloooow and smbfs is fast, much faster, the biggest difference I have seen is 9 times on an A3000. The NFS-client shipped with AmiTCP is equally fast compared to smbfs.


/Patrik

[amigagr]

just to understand, as if all goes well in the near future i'll boot to the same troubles: amitcp/genesis can make the connection to the dsl modem for internet purpose and nfs running on/through this, the share of the drives between amiga and pc?

[patrik]

Btw, there actually exist a NFS server for the Amiga, check this page out.


/Patrik

[amigagr]

ye indeed i found it earlyer on aminet. thanks

[patrik]

Did some tests to compare smbfs and ch_nfsc (the NFS-client shipped with AmiTCP). I ran the test on my A2000 with a 030/25MHz, to be sure the cpu is the limiting factor and not the network card to enable ch_nfsd/smbfs to make as big difference as possible.

Some info about the A2000:
GVP GForce030 - 68030+68882/25MHz
Commodore A2065
Kick/WB 3.1
AmiTCP 4.3
smbfs 1.66
Kickstart + VBR in fastram
CMQ030 CopyMem()/CopyMemQuick() patch


ch_nfsc:

Code: [Select]

MKSoft DiskSpeed 4.2  Copyright © 1989-92 MKSoft Development
------------------------------------------------------------
CPU: 68030  AmigaOS Version: 40.63  Normal Video DMA
Device: IBMen-NFS:    Buffers: <information unavailable>
Comments: ch_nfsc test

CPU Speed Rating: 1363

Testing directory manipulation speed.
File Create:            9 files/sec  |  CPU Available: 1%
File Open:             78 files/sec  |  CPU Available: 0%
Directory Scan:       108 files/sec  |  CPU Available: 0%
File Delete:           68 files/sec  |  CPU Available: 0%

Seek/Read:            564 seeks/sec  |  CPU Available: 0%

Testing with a 512 byte, MEMF_FAST, LONG-aligned buffer.
Create file:       197696 bytes/sec  |  CPU Available: 0%
Write to file:      86208 bytes/sec  |  CPU Available: 0%
Read from file:    206144 bytes/sec  |  CPU Available: 2%

Testing with a 4096 byte, MEMF_FAST, LONG-aligned buffer.
Create file:       321024 bytes/sec  |  CPU Available: 0%
Write to file:     307712 bytes/sec  |  CPU Available: 0%
Read from file:    290816 bytes/sec  |  CPU Available: 3%

Testing with a 32768 byte, MEMF_FAST, LONG-aligned buffer.
Create file:       367720 bytes/sec  |  CPU Available: 0%
Write to file:     388149 bytes/sec  |  CPU Available: 0%
Read from file:    367720 bytes/sec  |  CPU Available: 3%

Testing with a 262144 byte, MEMF_FAST, LONG-aligned buffer.
Create file:       372716 bytes/sec  |  CPU Available: 0%
Write to file:     393216 bytes/sec  |  CPU Available: 0%
Read from file:    372716 bytes/sec  |  CPU Available: 3%
smbfs:

Code: [Select]

MKSoft DiskSpeed 4.2  Copyright © 1989-92 MKSoft Development
------------------------------------------------------------
CPU: 68030  AmigaOS Version: 40.63  Normal Video DMA
Device:  IBMen-SMB:    Buffers: 170
Comments: smbfs test

CPU Speed Rating: 1364

Testing directory manipulation speed.
File Create:           16 files/sec  |  CPU Available: 63%
File Open:            134 files/sec  |  CPU Available: 0%
Directory Scan:       345 files/sec  |  CPU Available: 20%
File Delete:           98 files/sec  |  CPU Available: 2%

Seek/Read:            108 seeks/sec  |  CPU Available: 0%

Testing with a 512 byte, MEMF_FAST, LONG-aligned buffer.
Create file:        92096 bytes/sec  |  CPU Available: 7%
Write to file:      97024 bytes/sec  |  CPU Available: 3%
Read from file:     99456 bytes/sec  |  CPU Available: 2%

Testing with a 4096 byte, MEMF_FAST, LONG-aligned buffer.
Create file:       128382 bytes/sec  |  CPU Available: 46%
Write to file:     141312 bytes/sec  |  CPU Available: 43%
Read from file:    342528 bytes/sec  |  CPU Available: 7%

Testing with a 32768 byte, MEMF_FAST, LONG-aligned buffer.
Create file:       242725 bytes/sec  |  CPU Available: 30%
Write to file:     269661 bytes/sec  |  CPU Available: 27%
Read from file:    490294 bytes/sec  |  CPU Available: 5%

Testing with a 262144 byte, MEMF_FAST, LONG-aligned buffer.
Create file:       280201 bytes/sec  |  CPU Available: 18%
Write to file:     331446 bytes/sec  |  CPU Available: 8%
Read from file:    510255 bytes/sec  |  CPU Available: 5%
Quite even I must say, both have their strengths/weaknesses - ch_nfsc seems to have better overall performance, while smbfs has higher tops given good conditions (large chunk of data to transfer).


/Patrik

[adolescent]

But wow!  Look at that CPU%.  Is that correct, is the NFS client is taking 100% CPU!?  How was the system response while diskspeed was running?

Also, what's the discrepancy with the buffers.  Can you add buffers to the NFS device?

[patrik]

@adolescent:

The system was fine, the nfs-client is running at priority 0, so if something else needs cpu, it will share it fairly with the nfs client.

It is like using the ide-interface of an A600/A1200/A4000 or any other non-dma device. It will chew as much cpu as it can if it is not waiting for something, but it wont kill the system, only lessen the performance of the other applications (more apps sharing the cpu).

The reason smbfs isnt taking 100% cpu is because it partly is waiting for something to happen which in this case is the network.

Also, dont worry about the buffers with ch_nfsc, it is just the way the nfs-client is implemented. It most likely has an internal cache of some kind, even if it doesnt show it nor let you change its size, else small reads/writes would kill its performance completely, which doesnt seem to be the case.


/Patrik