Remote Storage with the Newisys NA-1400 NAS Appliance

Table of Contents

Ever thought about getting rid of your hard drive entirely? We did. Read about our effort to run Windows with the Newisys NA-1400 NAS appliance. Whether or not Network Attached Storage can run your OS, it’s a good remote storage solution that can put the noise where you are not. The Newisys NA-1400 comes in 1.0, 1.6, and 2.0 terabyte models — just stick it in a basement and enjoy.

May 7, 2006 by Devon
Cooke

Product
Newisys NA-1400 1.0 TB
Network Attached Storage Appliance (1.0 TB model)
Manufacturer
Newisys
Market Price
~US$1,200

THE PROBLEM

Hard drives are noisy. Well, not noisy exactly; most are well below
the 30 dBA@1m threshold that is our rough dividing line between noisy
and quiet. However, once a system has been made quiet, the hard
drive often remains as the primary source of noise. This is especially true
for systems with lots of storage, as high capacity drives tend to be noisier
than smaller drives.

There are several approaches to reducing drive noise, but all involve compromises.

  • Use a notebook drive. The most effective solution, but not practical for
    everyone because such drives tend to be smaller, slower, and more expensive
    than regular drives.
  • Soft-mount the drive, either by suspending it or placing it on a bed of
    foam. Effective for reducing vibration-induced noise, but does nothing for
    direct acoustic noise.
  • Use a HDD enclosure. Can work for reducing direct acoustic noise, and can help
    vibration noise. However, not all enclosures are effective, and the ones that
    are tend to cause overheating.
  • Use solid-state storage; both RAM and Flash-based versions exist. Unfortunately,
    these solutions are very expensive, especially when the costs are broken down
    per gigabyte.
  • Some combination of the above.

If these compromises are acceptable, then it is quite possible to build a system
where drive noise is not an issue at all. A notebook drive in an enclosure is
often completely silent for practical purposes.

However, there are a some situations where these compromises are not
acceptable, either because the cost is prohibitive or because the loss in performance
is unacceptable. In general, these are systems that require mass amounts of
storage: Studio systems, A/V workstations, and especially Home Theater PCs.
All of these require more capacity than a notebook drive or two can provide,
and none can rely on using enclosures — the first two because the additional
risk of failure is unacceptable in a commercial environment (downtime costs
money), and the last because of the additional space required. In addition,
all three applications have much to gain by reducing drive noise. Recording
and mastering require as little background noise as possible, and home theater
PCs should be quiet even when not in use.


A terabyte or two of storage is a tough challenge to silence.

AN EXPERIMENT

So, what options do these systems have for reducing noise? With a little creativity,
a new compromise can be worked out. If drive noise cannot be reduced enough
to be satisfactory, why not remove the drive from the system altogether? Most
motherboards can boot from a network, so why have any local storage at all?
There are no space or size restrictions on network storage, and capacity is
virtually unlimited. Best of all, the drives don’t have to be located near the
user, so noise is simply not an issue. The idea is intriguing, and we’ve never
seen it done with Windows before

Of course, there are some very obvious drawbacks associated with network storage,
namely reduced performance, especially in increased seek times when the network
is busy. And, our experience with computer terminals told us that remote hardware
often has serious consequences for system responsiveness.

Nevertheless, we decided to go ahead and experiment with the idea to see if
it was feasible. After all, most systems access the disk very little once they
are up and running. With a
few registry tweaks
, it is possible to minimize disk access and keep the
operating system running entirely in main memory. With enough RAM, it may even
be possible to disable the disk-based page file entirely. The longer seek times
would still affect load times when a program was first started, but with luck
the programs would run at full speed afterwards.

This approach would not work with disk-intensive programs, such as video editing
or encoding software. However, for these programs it is throughput that matters,
not seek times. A dedicated gigabit network can (in theory) provide the same
throughput as the ATA-100 interface — still easily enough to accommodate
a single drive. In addition, the most time-consuming tasks tend to be limited
by the CPU, not the storage system.

All in all, we thought we had a good chance of succeeding in our experiment.
However, we still needed a way of putting the storage on the network. Building
a whole new system just for serving files seemed a bit of a waste, and the additional
cost didn’t seem worth it just to run a few disks remotely. Besides, a much
better solution is available: Network Attached Storage, or NAS for short.

THE EQUIPMENT

A company called Newisys
was happy to provide one of their brand new NA-1400 NAS appliances for the experiment.
Our test sample came with four 250 GB drives, for a total of 1.0 TB of storage.
Larger capacity versions using 400 GB and 500 GB drives are also available,
and a new version using Seagate’s new 750 GB drives is also in the works, allowing
for a maximum of 3.0 terabytes. The drives came pre-configured in a RAID 5 configuration,
reducing the effective capacity to 750 GB, but adding a layer of data protection
that is necessary in a professional context. Home users who do not need the
extra protection may want to reconfigure the box to create a larger RAID 0 partition.

Newisys is targeting the NA-1400 at the small business and home theater markets.
These markets have different requirements, but both can benefit from
an integrated NAS box. The biggest advantage is that an integrated system does
not require steady maintenance— something that a small business cannot
afford and a home user doesn’t want to be bothered with. In addition, a small
business gets a source of centralized storage, and a home user gets far more
capacity than could fit into a small HTPC case.

The NA-1400 is built around a 600
MHz Intel XScale 80219 processor
and a customized version of Linux called
ApplianceWare.
In addition, it features dual gigabit ethernet ports and dual USB 2.0 ports.
Why would you want these extra ports? The extra ethernet port could be used
to provide a direct data connection to one system that bypasses network traffic
without sacrificing network connectivity for other systems. The USB ports can
be used for just about anything. For example, ApplianceWare allows the NA-1400
to double as a print server. Other capabilities can be added by installing additional
software.


A little black box with a 120W power brick.

Feature Highlights of Newisys NA-1400 (from Newisys’
web site
)
FEATURE & BRIEF COMMENT
1.0TB, 1.6TB and 2.0TB models
More storage than you’d
ever want to keep quiet in your main system.
4 hot-swap SATA,
high-capacity disk drive
The hot-swap
feature is genuinely useful, allowing failed drives to be replaced without
interrupting network access. It will even re-build RAID partitions automatically.
18 – 22 MB/s performance
Hardly on par with a
local drive, but as expected for network storage.
Intel® XScale® processor with Linux-based operating
system for reliability and compatibility
Not Windows based —
it should be stable without requiring constant updating.
Two USB ports on the front panel for easy connection with digital
media devices; both can be configured for separate networks or direct-connect
Multi-purpose, but documentation
on exactly how to use them is scarce.
Two GigE RJ45 Ethernet
ports
on the rear panel for direct-attach or Ethernet-based host connection
A direct connection will
not be subject to performance losses from network traffic.
Locking front and rear panels for added security against unauthorized
removal of hot-swap drives
Important for use in
public places. A slot for a Kensington lock makes it easy to secure.


A lockable door prevents the drives from being removed by quick or inquisitive
fingers.

The device itself is straightforward: It’s a black box with a few ports, some
flashing lights (power, network access, and individual read & write access
lights for each drive), and a power button. All of the lights are bright (very
bright) blue LEDs. The drives are mounted in individual cartridges that can
be easily removed if necessary.


Drives are easily accessible from the front panel.

The few ports are logically located, with the ethernet ports and the power
jack located at the bottom of the back panel, and the USB ports and the power
button accessible from the front. One minor issue was the position of the USB
ports, which are recessed below the door, making them difficult to see when
looking down on the box. However, it’s hard to complain about this: It’s only
a problem when the box is placed near the ground. Besides, the USB ports are
unlikely to see heavy use under ordinary circumstances.


The power button and a pair of USB ports are accessible even when the door
is locked.


Twin gigabit ethernet ports, a 19V power connector, and a hole for a Kensington
lock.

Cooling in the NA-1400 is provided by a single bottom-mounted fan that blows
upwards, forcing air through the system. The majority of the air should flow
up the sides of the case, which are hollow, then force its way between the drives
and eventually out the front panel. Unfortunately, there is very little space
between the drives, so the airflow is very restricted. It probably takes a lot of
pressure to force enough air through the system, which virtually dictates a high speed, noisy fan.

The bottom-mounted fan requires adequate clearance
underneath the system. It is also likely to collect dust quite easily.
We learned both of these lessons the hard way; suffice to say that putting the
NA-1400 on a carpeted surface is a bad idea. As long as the unit is placed on
a hard surface, there should be no issues with cooling.


ECC RAM guards against data corruption.
Cooling is provided by air that flows around the side of the circuit board.


The drive cartridges allow air to flow out the front.
Two blue LEDs on the right light up when the drive is reading or writing.


The drives plug into this PCB at the back of the box.

THE RESULT

Once the NA-1400 arrived at the SPCR lab, we eagerly set about copying
an installation of Windows over to it. Unfortunately, our assumption that we
could simply use the network boot function of the motherboard to boot from a
remote disk image turned out to be very, very wrong. Although it is possible
to install Windows from an NAS box, running it is out of the question.
Even then, the installation requires setting up a server to boot from. This
was far more complexity than we wanted to deal with, so we were forced to abandon
the idea of using remote storage for the boot volume.

For those who are interested, we did come across two technologies that do
allow diskless installations of Windows:

  • iSCSI: A protocol that allows disk commands to be sent directly to a device
    on a network. The Newisys NA-1400 does not support this protocol, but similar
    devices exist that do. Even then, the device would need to support some kind
    of boot server software, like netBoot
    /i
    .
  • Ardence
    Desktop Edition
    : Server software that allows any kind of network storage,
    including an NAS box like the NA-1400, to hold a bootable disk image. This
    requires a server in addition to the storage box and the client machine.

To make a long story short, the technology to enable non-local installations
of Windows does exist, but it is aimed at the server market and is not practical, affordable or simple enough for most end-users. At this point in time, only eSATA provides
a (more or less) practical way of moving a Windows OS drive out of the system, and even that has a maximum distance
of two meters.

A COMPROMISE

It is now clear our original solution to the problem of silencing large
amounts of storage is unworkable with this NAS appliance. But, the idea of using an NAS box to locate
most of the storage remotely is a sound one. Why not use a small notebook
drive to boot from, but still keep most of the storage away from the main system?
After all, this is exactly what the NA-1400 is designed for. With that in mind,
we set about testing the NA-1400 on a conventional system.

Test System

  • Antec P180 case, with only the rear TriCool 120 fan, soft-mounted with rubber
    grommets, running at the Low speed
  • Soltek SL-K8T939FL Motherboard
  • AMD Athlon 64 3800+ (Newcastle Core, 89W TDP)
  • 2 GB generic RAM
  • NVidia GeForce Ti4600 video card, with cooling fan removed + Nexus 80mm
    fan @ 5V installed over VGA card with a Zalman FB165 bracket
  • BenQ DW1620 DVD burner
  • Enermax Noisetaker 325 power supply with all fans removed + Nexus 120mm
    fan @ 7V installed in PSU channel
  • 2 x Samsung Spinpoint P80 hard drives in lower chamber, one 80 GB (JVC motor),
    one 160 GB (Nidec motor)

The network consisted of a 10/100 switch, a 10/100 router, and several computer
systems. The NA-1400 and the Test System were both connected to the switch,
and both received dynamic IP addresses from the DHCP server in the router. A
gigabit network may have provided better performance, but 100 Mbit connections
are still much more common in home networks.

Ambient conditions at the time of testing were 22°C, 20 dBA, and 122V@60Hz.

Setup and Configuration

Setting up the NA-1400 was as simple as plugging it into the network, pressing
the power button, and waiting for it to boot. After a minute or two it showed
up in Network Places, and could be treated like any other share storage. By
default, all files were accessible via any system on the network, although user
accounts and permissions can be configured manually if needed. A small utility
allowed the IP address of the box to be discovered so that it could be configured,
but it was not necessary to run this utility during ordinary operation.

Setting up a direct connection was not so simple, since no DHCP server was
available to give the box an IP address. Instead, it assumed a default IP address
and relied on the main system to connect to the rest of the network. Another
small utility needed to be run before this would work correctly. Unfortunately,
we were unable to get the direct connection mode working correctly, so we were
unable to test the box using a full gigabit connection.

Configuring the box was a matter of typing the appropriate IP address into
a browser window. The resulting pages worked much like the configuration pages
for a home router, with all interaction with the device taking place through
a browser. The java-based interface was very slow, and it took five or six seconds
for the box to respond to a command. Backing up the operating system onto a
flash drive took more than half an hour. During this time, only a “please
wait” message that refreshed once every few seconds showed that the box
had not frozen entirely. Luckily, things were already configured more or less
correctly, so most users should not need to spend much time in the configuration
menus. Most of the options were the set-once-and-forget variety.

The ApplianceWare operating system allowed for quite a wide range of configurations,
from the default RAID 5 volume to more specialized uses, such as multiple RAID
volumes and partitions and the ability to control read and write permissions
to various partitions. Individual user accounts could be configured on the box
itself, but most users who do this will probably want to offload the authentication
duties to an external server, using WINS or some other authentication system.
Other features include logging tools and a limited amount of hardware monitoring.

Despite the wide range of configurations available, we were quite happy to
stick with the default configuration. Things worked perfectly well in their
plug-and-play defaults, and there was little reason to change them in a personal
computing context.

One thing that we did need to configure was to mount the drive under a drive
letter. This was done on the main system itself, and did not require going through
the web-based configuration menus. This made the NA-1400 viewable in My Computer
like any other drive, and significantly speeded the access time when it was
accessed for the first time after booting up.

Noise and Power

As mentioned, the main reason to use an NAS box is to move a major source of
noise away from the main system. This means that there is little point to using
the NA-1400 if it will just be sitting beside the main system. To be effective,
it needs to be elsewhere … a long way from the main system. On its own, the
NA-1400 was far noisier than the four drives would have been on their own; the
cooling fan on the bottom was downright loud and could be heard from more than
a room away. To provide any acoustic benefit, the NA-1400 needed to be locked
away in a basement or a garage. It was simply too loud to be left in a room
that is used with any frequency.

Newisys NA-1400: Noise & Power
Activity
SPL
Power Consumption
Idle
40 dBA@1m
42W
Heavy Seek
42 dBA@1m
44~57W

To be fair, an attempt was made to control the fan speed; the noise level increased
noticeably after the box had been running for a while or during heavy use. Unfortunately,
the baseline noise level was intolerable even at its quietest. Once in a while,
the fan would not turn on at all until a few minutes after the power was turned
on. At this time, only the idle noise from the drives could be heard. Unfortunately,
this happened too infrequently to be measured, but we would estimate it at a
little below 30 dBA@1m.

The noise character was dirty and unpleasant. The main source of noise was
the fan, which made a grinding sound at low levels and produced a loud whine
when it speeded up. A ringing resonance from the drives also faded in and out.
Seek noise was very sharp and sudden, and seemed to be amplified by the metal
casing. It was clearly audible, even above the drone of the fan.

Power consumption was about as low as could be expected for a box with four
hard drives in it. Assuming that each drive consumed ~7 watts at idle, the chipset,
fan, and efficiency losses in the power supply consumed only 14 watts. Peak
power consumption occurred when the drives were under heavy activity, and was
about 15 watts higher than idle. Even under heavy activity, the power consumption
rarely reached its peak amounts, since the drives did not often hit their peak
consumption simultaneously.

MP3 Sound Recordings of the Newisys NA-1400

Newisys
NA-1400 @ Idle (40 dBA@1m)

Newisys
NA-1400 @ Seek (42 dBA@1m)

HOW TO LISTEN & COMPARE

These recordings were made
with a high resolution studio quality digital recording system. The microphone
was 3″ from the edge of the fan frame at a 45° angle, facing the intake
side of the fan to avoid direct wind noise. The ambient noise during all
recordings was 18 dBA or lower.

To set the volume to a realistic level (similar to
the original), try playing the Nexus 92 fan reference recording
and setting the volume so that it is barely audible. Then don’t reset
the volume and play the other sound files. Of course, tone controls or
other effects should all be turned off or set to neutral. For full details
on how to calibrate your sound system to get the most valid listening
comparison, please see the yellow text box entitled Listen to the
Fans
on page four of the article
SPCR’s Test / Sound Lab: A Short Tour.

PERFORMANCE

The most obvious test of performance is simply to test the amount of time it
takes to copy a set amount of data. This tests the data throughput of the device,
and should give some idea of how it compares to a conventional hard drive. Two
tests of this sort were done, one with a single large file (997 MB) and one
with multiple smaller files (42 files totaling 30.3 MB). A performance baseline
was established by copying the same files locally without involving the NAS
box. The local drives were Samsung Spinpoint P80 models. All tests were repeated
three times for accuracy, and none of the results varied by more than a second
or two.

Newisys NA-1400: Large File Performance
Test
Test
Time
Local disk to the same local disk
51s
Local disk to different local disk
29s
Local disk to NA-1400
2m 46s
NA-1400 to NA-1400
4m 41s
NA-1400 to Local disk
2m 24s

 

Newisys NA-1400: Small File Performance
Test
Test
Time
Local disk to the same local disk
4s
Local disk to different local disk
3s
Local disk to NA-1400
6s
NA-1400 to NA-1400
10s
NA-1400 to Local disk
5s

The raw performance numbers do a good job of showing the difference between
the network storage and the local disk, but they do not do justice to our subjective
impressions of the NA-1400. In terms of raw throughput,
the network box was less than half the speed of a local disk. This pattern was
the same for both of the tests, although the performance gap was bigger for
the large file test. The bottleneck was most likely the 100 Mbit network; file
transfers would most likely have been faster on a gigabit network, although
Barry Hutt, VP of Newisys, stated that
the performance bottleneck on a gigabit network is actually the XScale processor,
not the network. Performance was worst when files were being copied internally,
perhaps because data needed to be both sent and received over the network.

Our subjective impressions were actually
much better, perhaps because our usage pattern rarely tested the full throughput.
In general, our usage consisted of reading and writing small amounts of data
at a time: Reading and writing individual files that were rarely more than a
couple of megabytes in size. A small amount of lag was still noticeable, and
a couple of files took a second or two to process. The large discrepancy in
raw performance did not translate into a large discrepancy in our user experience.
The experience was more like using a computer that was slightly out of date
than having an experience that was “half as good” as using a local
disk.

To make this more concrete, we tried playing numerous video files from the
NAS box, including a number of ripped DVDs and several XviD-encoded and WMV-encoded
files. None of these experiments showed the NA-1400 to be any worse than a local
drive. At worst, the files took a second or two longer to open, but there was
absolutely no difference in playback quality.

To illustrate this, we did one last performance test, this time using GordianKnot
to encode a DVD copy of Secret
Things
into an MKV
video file, using XviD as
the codec. As before, several different source / destination combinations were
tested, and each test was repeated three times. Results were rounded appropriately
to account for variances in test result, but they rarely varied by more than
10 seconds. Each test consisted of two encoding passes and one multiplexing
pass. GordianKnot’s log function was used to measure the amount of time that
each stage took, and also to calculate the total encoding time.

Newisys NA-1400: Video Encoding Test
Test
Encoding Stage
Time
Local disk to the same local disk
First Pass
50m 10s
Second Pass
1h 14m 30s
Multiplex
1m 50s
Total
2h 6m 30s
Local disk to different local disk
First Pass
51m 20s
Second Pass
1h 14m 30s
Multiplex
1m 40s
Total
2h 7m 40s
Local disk to NA-1400
First Pass
50m 30s
Second Pass
1h 17m 10s
Multiplex
8m 0s
Total
2h 15m 30s
NA-1400 to NA-1400
First Pass
50m 30s
Second Pass
1h 17m 10s
Multiplex
8m 0s
Total
2h 15m 40s
NA-1400 to local disk
First Pass
51m 20s
Second Pass
1h 16m 30s
Multiplex
2m 0s
Total
2h 10m 0s

The results of the encoding test explain why we noticed so little subjective
difference: With the exception of the Multiplexing stage, all of the steps were
CPU-limited, so the storage system had very little effect on encoding time.
The multiplexing stage itself took about four times as long when the NA-1400
was the destination, but in the grand scheme of things, the speed of the storage
system mattered very little. In fact, the difference in encoding time between
the fastest and slowest times was just 7.5% — hardly noticeable. Few people
are likely to wait around to notice the extra 9 minutes, so there is little
practical difference.

CONCLUSIONS

Although we were disappointed to find that we could not run Windows from the
NA-1400, moving mass storage away from the main system
is a viable way to reduce noise. For many purposes, the loss in performance
that comes with using network storage is insignificant, and our subjective impressions
of the performance did not match the worst-case results from our early tests.

Unfortunately, the noise level of the NA-1400 itself was far too loud to be
placed in a living space, so it should be kept in a basement,
garage, well-insulated closet, or server room where its noise will not bother anyone. Apartment
dwellers may want to think twice.

As a stand-alone unit, the NA-1400 has several advantages over a custom-built
storage server. Most prominently, it is more or less plug-and-play and requires
very little configuration. Home users can easily plug it into a home network
and immediately use it without needing to know anything about networks or user
permissions. At the same time, complex RAID systems and user accounts can be
set up if the circumstances require. It also consumes significantly less power
than a full system.

In the right circumstances, the NA-1400 could be a very worthy investment.
It provides a way to access a huge amount of storage without needing to deal
with the noise that four large drives would add to an otherwise quiet system.
It is an ideal fit for a media server, where it can be plugged in, turned on,
and left in the basement indefinitely.

* * *

Much thanks to Newisys
for the opportunity to examine this product.

*

SPCR Articles of Related Interest:
NoVibes III HDD Decoupling Rack
Rubber Boxes & Carved Foam: More HDD Silencing
IS the Silent PC Future 2.5-inches wide?

Forum Thread:
HDD vibration & noise reducing methods – ranked

* * *

Discuss
this article in the SPCR Forums.

Silent PC Review is reader-supported. When you buy through links on our site, we may earn an affiliate commission. Learn More

Leave a Comment

Your email address will not be published. Required fields are marked *