Selecting an appropriate drive technology for a dedicated
server can be somewhat challenging, particularly considering
the many available options in terms of capacity and drive
type. On this page, we explore various significant factors,
and also provide some information on specific performance
attributes, to help you more effectively decide what drive
technology is best for your dedicated hosting needs.
SATA vs. SCSI
One of the most prevailing misconceptions regarding the
SATA and SCSI drives is that the primary difference is
in the way the drives interface to a computer. Although
they are of course different in that respect, the more
fundamental differences have to do with specific aspects
of construction of the drive -- specifically, the mechanics,
materials, electronics, and firmware. Thus, the true
distinction is between Personal Storage (or "PS") drives
and Enterprise Storage (or "ES") drives. In almost all
circumstances, SCSI drives are ES devices, while SATA
drives are PS devices. Following is an overview of some
of the fundamental differences:
PS vs. ES (SATA vs. SCSI): Design Considerations
It can be quite convincingly argued that the single most
significant aspect to a PS drive is that its cost is
roughly proportional to the cost of the overall computer
system in which it is utilized. This can be easily seen
by considering that an entire computer system, with 80GB
of hard drive capacity, can be procured for approximately
the same price as a 73GB 15,000 RPM Enterprise Storage
drive. Simply put, cost pressures assure that PS drive design
is dominated not by high reliability, but by high affordability.
That's not to say that PS drives aren't suitable for their
purposes, but it does indicate that they will not have the same
quality characteristics as an ES drive.
ES (SCSI) drives on the other hand, have consistently been
used on large computer systems. They thus reflect qualities
that make them suitable for aggregation (i.e. group configurations),
that make them optimised for randomly accessing small portions
of data (typical "server" behaviour), and that make them intrinsicly reliable. Indeed,
as one group of researchers succinctly put it, in a commercial
environment, "[drive] failure could idle a considerable number
of employees and directly impact business operations."
Following is a brief table that gives a summary of some of the
more specific differences between drive types:
||Lower media spin speeds result in
comparatively higher latency vs. ES drives.
Innovation tends to focus on cost savings
(e.g. lowering the net cost to make a 7,200 RPM motor).
||Higher spin speeds result in lower latency
vs. PS drives. Innovation tends to focus on
new technologies (e.g. building a 15,000 RPM
motor for the first time).|
|Aggregation (i.e. > 1 drive)
||Designed for single-drive systems. Not
as effective handling of rotational vibration,
causing performance loss in multi-drive configurations.
||Designed for multiple-drive systems. Internal
drive mechanisms better handle rotational
vibration, thus avoiding related performance loss.|
||Lower-cost components, lower-power requirements
lead to simpler but less failure-proof designs.
||Goal is to achieve 1,000,000 hours between failures,
and to compensate for higher spindle speeds and
other effects such as rotational vibration. More
effective cooling, environmental protection
built into the drive, O-ring sealed spindle motors,
dessicant to control humidity, etc., are present.|
||Simpler IDE/SATA interface allows for less
complicated electronics, but this also
translates to less performance gain at the
||SCSI/FC drives typically have two processors; one for the servo, another for interface and read/write handling. ES drives also tend to have larger SRAM space, flash memory, data SRAM and cache SRAM.|
||Much more basic manufacturing and post-manufacture
testing than Enterprise Storage drives.
||Significantly longer build and test times
compared to PS; not only ensures reliability
but improves performance as drive "learns" of
track irregularities and media flaws.
||Tends to use a larger number of drive platters / sacrificing performance to gain capacity. Lower RPM and larger platter sizes than ES degrade performance.
||Small drive platters, higher RPM, lead to better intrinsic performance. Aggressive seek scheduling improves performance and lowers mechanical duty cycle.|
||Since not normally designed for multi-drive use,
rotational vibration (from the spinning of
other drives within the chassis) throws drive actuator "off-track" -- causing aborted writes + failed seeks (and in extreme cases, inoperability).
||Explicitly designed to operate in close proximity to other spinning drives. External rotation detection can compensate in servo processing; combined with better build quality, this ensures lower performance loss from rotational vibration.|
||Reliability specifications designed for expected
power-on-hours of 8 hours/day, 300 days/year. Using in a 24/7/365 server environment forces drive to work beyond specifications.
||Reliability specifications based on 24/7/365 drive use. Better scheduling algorithms also contribute to lower duty cycles, by making more efficient use of drive mechanics. Designed for enterprise-grade storage.|
Choosing a Drive for your Dedicated Server
Of course, given the above, we normally recommend an
enterprise-grade drive technology for dedicated server use.
However, when reliability and efficiency are not major concerns,
PS drives (SATA-based) can be effectively utilised, and do
in fact enable higher-capacity-per-dollar-spent.
Seagate Cheetah - 15,000 RPM Drives
These are the cornerstone of our VPS hosting platforms,
and of course are available for dedicated hosting clients.
These drives hold a variety of distinctions, including:
• Being the first drive to break the 100-Mbyte/s sustained
transfer rate barrier (peaking at 125 Mbytes/s)
• Perpendicular recording enables capacity up to 300GB
while offering 30% higher performance.
• 30% higher IOPS and more than 20% faster response times
when compared to 3.5" 10,000 RPM drives.
• Highest reliability of any 3.5" drive in the industry, with
a 0.62% Annualized failure rate.
• A focus on reliability and data protection has led to
innovations that have improved the nonrecoverable error rate
by a factor of 10 over earlier models.
• Exclusive IRAW (Idle Read After Write), Error Correction Code
and second generation Background Media Scan significantly
improves performance and reliability.
• Average Read/Write (msec) - 3.5/4.0
• Nonrecoverable Read Errors per Bits Read - 1 sector per 1016
10,000 RPM vs. 15,000 RPM Seagate Cheetahs
Following are specification comparisions for both
the Seagate Cheetah 10K and 15K. This data is based on
information provided by Seagate:
|Spindle Speed||10,000 RPM||15,000 RPM
|Average Read/Seek Time (msec)
|Transfer Rate (sustained Mbytes/s)
||39 to 80||73 to 125
|Nonrecoveral Read Errors per Bits Read
||1 sector per 1015||1 sector per 1016
|Annualized Failure Rate
Based on a 14 drive clustered configuration
|725||955 (32% faster)
Based on a 14 drive clustered configuration