[Solved] slower reading speed on Raid 1?

[NASoverdose]

I'm curious why reading is slower than writing on my setup.

TS-412 (3.7.1)
RAID1 (2x ST2000DL004/HD204UI)
Gigabit LAN


transfering large files over shared folders:

~30MB/s (reading from NAS)
~40-45MB/s (writing to NAS)


in comparison ftp:

~60MB/s (read)
~15MB/s (write)

[rinthos]

I'm afraid you'll need to clarify a bit more..
Raid 1 is a disk configuration, not file transfer protocol...

So assuming you get ~30MBps reading for Samba/CIFS (file share reading/writing) versus ~60MBps for writing?

If so there are a lot of factors involved, such as file permissions, number of files/folders, how you are moving content, etc.

Or are you saying you had a different RAID configuration for the volume you were using FTP on? (if so, you'll need to provide a comparison..RAID 1 vs..Raid 5? Raid 0?).

Any details can at least help people guess what is going on...
---

[pwilson]

I'm curious why reading is slower than writing on my setup.

TS-412 (3.7.1)
RAID1 (2x ST2000DL004/HD204UI)
Gigabit LAN


transfering large files over shared folders:

~30MB/s (reading from NAS)
~40-45MB/s (writing to NAS)


in comparison ftp:

~60MB/s (read)
~15MB/s (write)

Read speeds should be faster than Write Speeds, which makes the Samba stats you provided look quite suspicious.

Perhaps we should start with the Published Speeds, as published in the [url]4-Bay NAS Comparison[/url] chart. It claims the following for a TS-412:

[=========] Chart [==] Your Stats
Samba Read: 69.3 .......... 30
Samba Write: 32.7 ......... 40-45
FTP Read: ...110.3 ......... 60
FTP Write: ... 32.8 ......... 15

Your stats are indeed quite different than QNAP's numbers.

It is interesting that your SAMBA "write" speeds are actually 50% faster than claimed by QNAP, yet your other stats are approximately 50% smaller.

Please tell us more about your network. Router Make/Model? Was there any other equipment active on your LAN when these stats were taken?
What QPKGS are installed?



I note than your TS-412 NAS only has a 1.2 GHz CPU, and only has 256 MB of RAM, so performance could be severely affected if you are using many QPKG's on your system, or if you have any form of Torrent software running. How did you make your speed measurements to arrive at those values? What services were enabled at the time of the tests? Are you using more than one of the NAS's Ethernet ports?

Perhaps if I had a better understanding of your network perhaps I could provide some suggestions.

Patrick.

[schumaku]

Good read perromance requires a reasonable amount of free (available) DRAM - othrwise caching and the QNAP-defined NAS config might be not optimal.

[NASoverdose]

Win7x64 <- GBit-Switch -> NAS (LAN Port 1 used, LAN2 unused)

TP-Link TL-SG1005D + 2 short CAT5 cables (but same results without switch and direct cable connection)

compared Samba + FTP Transfer (of the same setup)
large files (2GB+ TV recordings)

Samba shared folders mapped as network drives in Win7

transfer numbers provides by both Windows and Total Commander (+ftp plugin)

Desktop PC hard drives more than capable (SATA II, >100MB/s)

only 3 services running, nothing else (CPU quite idle without file transfer):
Microsoft Networking, Web File Manager, FTP Service

153MB free RAM (of 256, idle)

no QPKG add-ons installed

[=========] Chart [==] Your Stats
Samba Read: 69.3 .......... 30
Samba Write: 32.7 ......... 40-45
FTP Read: ...110.3 ......... 60
FTP Write: ... 32.8 ......... 15

That's correct (numbers rounded).
It's my first NAS Setup, so I'm wondering where is the bottle neck or my error.
I read Raid-1 should provide theoretically quite good reading performance.

[NASoverdose]

Solved it, Windows borked. (Updates + TCP Optimizer)

Getting good Samba read performance now:
(although ftp upload is still as low)

[schumaku]

Rule of thum, applications normal users dont' need on Windows (vista, 7, : 3rd party security packages, TCP optimizers, Registy optimizers, ... and also don't touch the WIndows interface and TCP registy entries. Why are always performacne geeks hit by the similar issue? Figure...