NASLite Network Attached Storage

My NasLite2 box is PIII with 512MB of RAM and Giga Ethernet connected to a XP machine through a D-link Gigabit switch and I get:
27MBs average (write or read) for 1GB file.
I'm happy with that.

Here are the results of my test using Passmark v6.0

Sequential read 37MB/s
Sequential Write 18.9MB/S

Diskmark 263.1

- BIOSTAR P4M80-M4-COMBO35 Intel Celeron D 315 Socket 478 VIA P4M800 Micro ATX Motherboard/CPU Set - Retail
- Netgear Gbit NIC
- 512MB RAM
- ATA100 160GB drive

Again I put to you to, do a real world test and use the same hardware and follow proper testing procedure. You have put forth results based on poor and erroneous methodology. You are asking them to post their results, do your testing in a scientific way and then they may do as you ask and post some of theirs.

I would be supprised if the NASLite server bested the NT2000 in the performance arena for aforementioned reasons of code access and literal man hours spent tunning the OS, drivers, apps, and utils but all things being equil I would expect that it was damm close.

Mike

PS, the spelling nit-pic was crap as well.

I'll put $ on that one, hardware driver interaction.

Here's MY benchmarks
NASlite clearly bested win2000 in this round. 19MBps compared to 13MBps.
There's an 'elephant in the room' if you notice the last graph.

All on the same Zyxel 5port switch.
Maybe if I feel like blowing AN ENTIRE MORNING AGAIN , I'll run NASlite on one of two XP boxes to complete the circle.

thanks the test results are just as I expected they would be but NL does a HELL of a job considering the size of the OS, the cost, and the very small development team involved. I also would more than put my money on the NASLite box for the reliability side of things.

Thanks for the great post,

Mike

Thanks for that! Seriously, hours, nothing is EVER simple nowadays.

Of Note:
I edited the bitmaps laterally to shift them over to put them in the same ballpark in the TaskMan Window.
The transfers are of course untouched.

Oh, ignore all those 'Froms' in the bottom (what I get for copying that string all over)

The caching of NL needs to be noted and can skew benches.
The NL one (3rd down) shows the cache in action. In the top three I started the file copy, Cancelled, then restarted. Win2000 didn't really care, maybe since the operation didn't complete, but NL has a write-back I guess which helped it for the first little bit, although I let the initial copy go a longer on the two win2k's , so maybe if I did it _exactly_ the same, the caching would be the same...

I was wondering if a null device could be supported/created to test through the NIC and bus, excluding drive controller, Drive, and file system. Kinda like a RAM disk, or an OS loopback, but would just be for measuring network throughput.
This might help find slowdowns!

NetCPS is a freeware utility that does just this, pumps data from one machine to another through the TCP/IP connection without any disk access. The executable is a command line utility for DOS/Windows, but it includes source code so an enterprising Unix guy could probably port it over.

http://www.netchain.com/netcps/

.

lmoseley,

I posted my benchmarks and explained my precedure, why don't you re-read this entire thread carefully again and qualify your claims that your so adamant about.

http://www.serverelements.com/phpBB2/viewtopic.php?t=1224&sid=3d292e068da95b9507e4943bfc80b798

I had to chuckle... NASLite is positioned as a product that runs fine on old hardware, but your benchmarks are from a dual-processor machine with 2 GB ram and a RAID. Not exactly represenative...

OK, I am rereading. You have indicated that you think my machine is not working properly. What next? How do I determine from the logs, or something else, what isn't working properly?

OK,

Again, you have managed to confuse me with the above comment. You asked Ralph for benchmarks of the capabilities that NASLite-2 has. NASLite-2 will run happily on a 486 with sufficient resources to accommodate the minimum requirements of the OS. None of the other operating system would even attempt to run on a 486, let alone perform.

What you must not realize is that by doing the test on capable hardware we are in fact giving Windows and FreeNAS the advantage. Perhaps we should repeat the test on a p200 with 64M of RAM and see what the results look like. I bet NASLite-2 will look even better compared to Windows and FreeNAS.

Hi boys,

reading the discussion about benchmarks and which soft is better and so on starts getting boring. The first reason for me using (and paying) Tony`s NAS was the support of german Umlaute (like Ä Ö Ü) what Free-NAS definitly doesn`t and as I got out after a few mails not will in future. My mate and me spent more than 4 weeks of patching and baking new kernels and configs of samba (if we would have got 5$ an hour we could have payed Tony including a big donation). And second I know that both soft do their best to get the best out of bought hardware.
So please keep beeing nice and use this forum to help and to get NAS better.

Roland Berbalk (germany)

Tony, all I want to do is get MY NASLite2 box running fast enough to be usable as an archive machine for sizable filesets. If, in fact, the slow speed of my box is due to hardware problems, as someone has suggested, how do I determine from the log files I have posted just what is causing the problem?

Reasonable request

The following suggestions are purely based on my observations of the syslog you posted, so if something changed, please let me know...

First, I'll reset all BIOS settings to defaults using the jumper. Then, I'll disable ACPI as well as any and all power management options. Follow that by disabling all components such as sound, com, lpt, etc. Also disable VGA IRQ if an option. Then set the PnP OS option to "NO". Anything else you can see in the BIOS to "dumb" the machine down.

You may have to reboot a few times after going in the BIOS and tweaking things until you get a clean IRQs for the NIC and the drive controllers. Shared IRQs are a very bad thing in some situations and can totally flood the CPU with requests that will result in little work done given the effort.

When that is done, boot NASLite-2 and disable SMART on all drives. Make sure that you save and reboot after completing configs.

Try the bench again and ensure that the client/server are both in full duplex. Keep an eye on the syslog and make sure that there are no disk errors. Also take a look at the system page and ensure there are no exessive transfer errors and dropped packets.

Give it a try and see what you get. We'll figure it out

A most informative post!

Hi to everybody,

just to publish my own results: NAS: ASUS TR-DLS, 2x1.4 GHz P3, 1G ram, Intel 1000 tx(64bitpci), Maxtor 250GB 7l250ro on serverworks ide running ata100, aic7892: Ultra160 Wide Channel A, SCSI (64bitpci), ic35l036uwdy10-0. Until now it takes 40 sec for 800MB of received and sent data file over a level one gig switch at any harddisk. Tomorrow I will post the same with a crosslink-cable.
Good night,
Roland

Hi,
the connection with a crosslink-cable cat5e takes 32 sec for the 800MB file(a harddisk-image).
Hope that helps,

Roland