Where is the bottleneck in NAS drives?

ghart999

Member
Jun 27, 2001
124
0
0
Hi all. Sorry for the newbie questions but I am trying to make sure I understand where the throughput bottleneck would be in a NAS.

Assume I have a 1TB NAS drive that uses a standard ATA-100 IDE hard drive and 10/100 ethernet. That would mean that drive is capable of 100MB/sec but that my network connection would be somewhere between 10MB/sec and 100MB/sec. Hence the network would be the bottle neck right?

If the internal drive was SATA I would have much higher drive throughput, but I would still be limited by the 10/100 ethernet connection.

If this is the case, I would think that from a pure performance standpoint I would be better off with an external HD with USB2.0 or Firewire 400/800 since all three of these bus protocals would be faster than my 10/100 ethernet?

I hope I have my numbers correct here as I am assuming that ATA-100 is 100 megabits per sec and not megabytes right? And that 10/100 ethernet is based on megabits per second and not megabytes per second. Sorry for the dumb questions, but that you as well.

Gregg
 

MerlinRML

Senior member
Sep 9, 2005
207
0
71
On a 10/100Mb network, your network is likely to always be your bottleneck.

However, you have one point wrong. You're comparing ATA bus speeds, not actual hard drive performance. Not even the fastest SATA drives can hit 100MB/sec without adding additional drives in RAID arrays.

So yes, in your case, pretty much any drive on any modern bus would be fast enough. However, keep in mind that gigabit networking is getting more and more affordable, and you may be in a position to upgrade. So leaving yourself some room to grow wouldn't be a bad idea.

For reference:
disks
ATA-100 = 100Mbytes
SATA 150/ SATA 1 = 150MBytes or 1.5Gbit
SATA 300/ SATA 2 = 300MBytes or 3.0Gbit

networking
10/100/1000 = 10Mbit/100Mbit/1000Mbit or 1Gbit
 

child of wonder

Diamond Member
Aug 31, 2006
8,307
176
106
100Mb Ethernet can transmit at 12.5 MB/s as a theoretical maximum (100,000,000 bits divided by 8)

Your hard drive has a theoretical maximum read speed of 100 MB/s, much faster than your ethernet connection.
 

ghart999

Member
Jun 27, 2001
124
0
0
Thanks all. This helps to understand that no matter which solution I go with the network is my bottleneck until I go with gigabit someday.
 

Madwand1

Diamond Member
Jan 23, 2006
3,309
0
76
Note that even if you had gigabit, it's unlikely that an inexpensive consumer NAS box could give you much better than 100 Mb/s speed. Once the network is theoretically not a bottleneck, everything else is, including the drive, NIC, CPU, server OS, etc.

Inexpensive NAS drives have one advantage -- they don't complicate the system further with RAID 5. RAID 5 would put greater load on the system and could slow it down further.

Bottom line -- a NAS has some complexity, and at the budget end, with inexpensive hardware, this complexity limits performance to be significantly less than theoretical gigabit speeds; check the details of your candidate devices before assuming much about performance.

Firewire / USB2 are simpler and can hit around 30 MB/s read/write more reliably than inexpensive NAS drives. eSATA can do even better. These may be good alternatives at the low end in performance terms.
 

Smilin

Diamond Member
Mar 4, 2002
7,357
0
0
Originally posted by: child of wonder
100Mb Ethernet can transmit at 12.5 MB/s as a theoretical maximum (100,000,000 bits divided by 8)

Your hard drive has a theoretical maximum read speed of 100 MB/s, much faster than your ethernet connection.

Also there will be a great deal of protocol overhead.

Ethernet + IP + TCP + SMB. That's a lot of headers, a lot of acks, and a lot of protocol negotiation. Ethernet will throw a frame before IP starts. TCP will throw three before SMB Starts. SMB will throw like a dozen frames before it even transmits one byte of a file.
 

JackMDS

Elite Member
Super Moderator
Oct 25, 1999
29,539
418
126
While many inexpensive Network Devices can work very well, NAS is Not one of them.

The sub $500 NAS is still an Industry Albatross.
 
Jan 31, 2002
40,819
2
0
Originally posted by: JackMDS
While many inexpensive Network Devices can work very well, NAS is Not one of them.

The sub $500 NAS is still an Industry Albatross.

Not if you build it yourself. ;)

- M4H
 

ghart999

Member
Jun 27, 2001
124
0
0
Thanks all. I think I will just go with internal drives in one PC and share the drives. Then for backup I will use an external USB/firewire drive.
 

zephyrprime

Diamond Member
Feb 18, 2001
7,512
2
81
Originally posted by: Madwand1
Note that even if you had gigabit, it's unlikely that an inexpensive consumer NAS box could give you much better than 100 Mb/s speed.
Is this really true? I don't see how a gigabit interface could be so slow. Even the slowest copper gigabit benchmark for a NIC I've seen can attain ~600mbps.

 

Madwand1

Diamond Member
Jan 23, 2006
3,309
0
76
Originally posted by: zephyrprime
Originally posted by: Madwand1
Note that even if you had gigabit, it's unlikely that an inexpensive consumer NAS box could give you much better than 100 Mb/s speed.
Is this really true? I don't see how a gigabit interface could be so slow. Even the slowest copper gigabit benchmark for a NIC I've seen can attain ~600mbps.

Typically the gigabit NIC and cabling and switch are not the bottlenecks. This is counter-intuitive at the onset -- that the network isn't the bottleneck for network transfers at the gigabit level although it used to be at 100 Mb/s. A number of other factors come into play here, where the rest of the system has a good ways to go before the network becomes the bottleneck again.

For a cheap consumer NAS box, even the CPU will be a bottleneck, let alone the hard drives, the storage controller implementation, the overall system I/O bandwidth, the network drivers, file transfer protocol, OS tuning, etc..

Small Net Builder has a number of consumer NAS measurements which demonstrate this behavior. (Although it looks like they're fiddling with their web configuration atm, and have broken their NAS charts temporarily; their data is usually accessible.) Take their data with a grain of salt -- e.g. ignore all the "small file size" measurements -- these essentially demonstrate cached performance and are fairly useless. I'd look at the "1 GB" file size measurements only. Also don't consider their DIY NAS box measurements representative -- their design is sloppy from a performance standpoint (e.g. PCI NIC and PCI storage controller), and they stated other priorities in their design.