DSM-G600, DNS-3xx and NSA-220 Hack Forum

What's normal for my network, may not be normal for yours - especially when wireless networking is involved - the throughput on a wireless network can be affected in a major fashion by the surrounding environment, both physical and electrical - it's easier to predict wired speeds, and yes, if you want maximum throughput

For a wired network - if it's gigabit, you can see speeds of upto 22 MB/sec and if it's 10/100 anywhere from 8~10 MB/sec.
For wireless I'd guess 2.2 MB/sec for wireless-g, upto maybe 6 MB/sec or therabouts for wireless-n - when I say upto, I mean it could be less, significantly less.

Also - these are the numbers you'll see transferring large files - if you transfer a single 4GB file it will go much, much quicker than if you transfer 1000 x 4MB files.

#jhorsnel: Why are you using a pc? The boxes can do this on their own.

fordem wrote:

What's normal for my network, may not be normal for yours - especially when wireless networking is involved - the throughput on a wireless network can be affected in a major fashion by the surrounding environment, both physical and electrical - it's easier to predict wired speeds, and yes, if you want maximum throughput

For a wired network - if it's gigabit, you can see speeds of upto 22 MB/sec and if it's 10/100 anywhere from 8~10 MB/sec.
For wireless I'd guess 2.2 MB/sec for wireless-g, upto maybe 6 MB/sec or therabouts for wireless-n - when I say upto, I mean it could be less, significantly less.

Also - these are the numbers you'll see transferring large files - if you transfer a single 4GB file it will go much, much quicker than if you transfer 1000 x 4MB files.

I have a 200 GB mirrored DNS323 setup. I'm going from a windows vista ultimate x86 machine to the DNS on gigabit Ethernet and I'm getting a max of 17MB/sec. Is that an OK or 'normal' speed? Can't seem to get better than that.

Oh, I was getting about 9MB/s on XP on a 100 Base T connection.

Last edited by gazoo (2008-12-23 22:21:26)

I was stating my speeds for the OP's benefit, but also left a little room to make double sure there's nothing wrong by adding the interrogation mark.  [/unshakes head for you]

Uros

I can't tell you how to get to 22, and even if I could, I don't think it would make a significant difference - unless your real world usage of the NAS co-incidentally matches your test conditions.

Even though I can transfer data as high as 28~29MB/sec, my real world usage (backup) averages 0.5 MB/sec - you might not see as dramatic a difference as I do, but you will probably see a significant one.  My testing is done with single files 2GB is size, but real world usage I have no control of the file sizes and there are thousands of small files in the data sets.

When looking at file transfer speeds you need to look at the entire picture - which looks something like this ....

DISK1<->SATA<->Disk controller<->Expansion bus<->NIC1<->CAT5e<->Network Switch<->CAT5e<->NIC2<->Expansion bus<-> Disk controller<->Interface<->DISK2

Your transfer is going to be limited by which ever of those components is the slowest - you cannot look at the DNS-323 and the network and just ignore the computer at the other end.

Drive rpm - 5400, 7200, 10,000, 15,000 - it makes a difference, network cards - I have seen gigabit network cards that cannot transfer faster than 170 mbps, file sizes - transferring a single 2GB file is much much quicker than transferring 1000x2MB files, even though it's the same 2GB of data, disk fragmentation - a fragmented disk will have a noticeable negative impact on throughput.

As an example ...

My desktop - a Dell Optiplex GX280 gives - 9.65MB/sec writes and 16.34 MB/reads, wheras my server - an IBM x206 gives 12.76MB/sec writes and 20.87 MB/sec reads, enable jumbo frame (9000bytes) and the server will give 17.75 MB/sec writes and 27.28 MB/sec reads - the only changes made were to run the test from a different computer and to enable/disable jumbo frame on the server.

As you can see the hardware at the other end can make a major difference - but - I deliberately didn't describe the hardware, you didn't and few people seem to consider it.  To actually make this comparison meaningful you need to know quite a bit more about the hardware.

The Optiplex GX280 has a 3GHz HT P4, a gigabyte or RAM, a Broadcom NetXtreme 57xx ethernet controller (no jumbo frame) and a 7200rpm 80GB Maxtor SATA drive, the IBM x206 has the same processor and memory, but an Intel PRO/1000MT ethernet controller (jumbo frame supported) and a pair of 7200 rpm Seagate Barracuda 250GB drives in a RAID1 array.

Server grade equipment, even low end server is optimized for network transfers, desktops are not - their primary function is the productivity applications that they are used with.

Last edited by fordem (2009-01-29 18:03:12)