The Art of Computing Series – Development of SATA Technology
In the past several years, Serial
Advanced Technology Attachment (SATA) has developed as a technology
in the low-end of enterprise class storage markets. SATA has shown
that there are alternatives technologies to expensive Fiber Channel
(FC) and Small Computer System Interface (SCSI). SATA has made
significant gains in not only in desktop applications, but also in
server applications.
First was SATA, then SATA II
followed addressing some on the shortcomings of the original SATA
implementation. SATA was successful at changing the approach to
mass storage, showing that there were other ways to achieve large
storage capacities while still preserving both performance and
management features. However, the SATA interface revealed the
shortcomings that made it virtually unusable for a large number of
applications, leaving a large portion of the market to FC and SCSI.
SATA II, the extension to the
original SATA standard, has served as a bridging technology from a
desktop-oriented storage interface to a server-level storage
interface, making it a viable, low-cost storage solution for server,
NAS, and SAN applications.
For the foreseeable future, there
will always be a need for high-end FC and SCSI drives, but SATA II
is making the server alternative viable. SATA II is bridging the
gap between SATA and FC/SCSI applications using Native Command
Queuing, Enclosure Management, and Port Multipliers.
Before SATA, the storage world was
clearly divided between ATA for personal computing and SCSI/FC for
professional and server-oriented computing.
ATA users typically worked with
personal computers, single-threaded applications, 8x5 usage models,
and believed that performance was nice, but cost was by far the most
important.
On the other hand, SCSI and Fiber
Channel users worked with servers and high-end workstations,
multi-threaded applications, 24x7 usage models, and while cost was
important, performance was first and foremost.
As the cost of servers and high-end
workstations decreased, servers and high-end workstations became
available to small offices and homes, making cost a more important
factor. Many manufacturers began introducing entry-level ATA storage
solutions in order to attract customers. However, SCSI and FC
manufacturers knew ATA was not a challenge to SCSI and FC in the
server and high-end workstation arena.
Then there came the SATA technology.
Fundamentally, SATA was the same as ATA except that instead of data
flowing in parallel, it would flow in a serial stream. The new
Serial ATA standard came out and was driven by very practical
reasons. ATA had passed the point where data could be transferred
at reliable speeds, cabling was cluttered and made the airflow
necessary to cool processors a challenge. Also, 5 volt power was a
dying technology because 0.13 micron and 90nm technology could not
withstand voltages that high.
SATA provided a new concept for how
storage can be used in professional environments. The serial
interface allows more ports to be integrated into the same piece of
silicon, so the traditional 2-channel ATA controller can be replaced
with an 8-port SATA controller. Also, the new cabling made it much
easier to route many more disks than ATA. Where ATA barely connected
up to four disks, SATA can easily add up to eight. This, coupled
with the low-cost of the Serial ATA controllers, has created a new
business model. Why use four SCSI disks to create a RAID 5 when you
can build it with four, five or six SATA disks for much less money?
Even with these new advancements in
Serial ATA, there are limitations that make it uncompetitive with
SCSI and FC in several areas. The greatest achievement of the
original SATA implementation was that it made people realize that
there are alternatives in storage and that SATA is blazing the new
trail.
The main limitation of SATA - just
as its name states - is that it is just a “Serial-ATA.” And with
that come many of the protocol, manageability, usability, and
reliability concerns that existed with Parallel ATA. These
limitations are divided into two separate issues
 |
Mechanical chassis,
and |
|
Protocol. |
SCSI chassis are designed for high performance, high reliability,
24x7, multi-threaded protocols. They are designed in tightly packed
configurations that are able to survive with high MTBF’s (mean time
between failures) and all the mechanical requests that a high-end
server environment demands. SATA chassis, on the other hand, are
designed to live alone in a PC, not in a server where they are
coupled with many disks whose vibrations stress each other’s
bearings. Unexpected heat generation also becomes a serious issue
due to multi-threaded activities.
Both SCSI and SATA have their own
value. SCSI, FC, and Serial Attach SCSI provide top-notch
performance while SATA is still much more economical and will serve
less-demanding markets or users.
As for the protocol issue, there are
at least three areas that taint SATA’s server potential:
|
Performance
|
|
Manageability, and
|
|
Connectivity |
In terms of performance, most of the
limitations are due to mechanical issues as discussed above. But
further, with SATA, there is no value in a server application where
access is largely random and command reordering gives top-notch
benefits. With SATA you only have one command to work with. So, if
you are using the disk for a video server, performance may be
adequate. However, if you are using is for any type of transaction
processing, you will most likely see substandard performance.
Regarding manageability, consider
when you purchase a RAID controller, it is because you want to
preserve data integrity and functionality in the event of a disk
failure. In the case of SATA, where the die-out rate of disks is
higher, you especially want to use RAID. This is where the
manageability limitation comes into the picture. Once you have a
failed disk drive, you need to replace it. With SCSI, FC, and SAS,
there is a mechanism that allows a dead disk to be reported through
light indicators. With SATA, there is no indicator to show you which
disk has failed. The user then runs the risk of losing invaluable
data by pulling a functional disk instead of the failed disk.
Regarding connectivity, although
SATA disks are inexpensive, the total SATA solution will not
necessarily be low-cost. The main reason for this is that while
using SATA 1.0 controllers, one can only connect one disk per port.
If you need eight disk drives, you have to use eight ports. Also,
using SATA 1.0 controllers is not an efficient use of the transfer
speed. Although SATA 1.0 touts a 150MB/sec peak speed, a SATA disk
drive may be around 60MB/sec. In essence, each port would only be
using a fraction of its potential.
The introduction of Serial ATA II
addresses the limitations of SATA 1.0 in order to make this
interface a fit for large-scale professional deployment. There are
five main features that epitomize Serial ATA II:
|
Higher per-port
transfer rate (3Gbs = 300MB/sec) |
|
Native Command Queuing
|
|
Enclosure management
|
|
Port multiplier |
|
Provides an upgrade
path to SAS |
It may sound strange to have
increased the per-port transfer rate given the issues discussed
above. With one disk connected to one port, 150MB/sec is a waste of
bandwidth because one disk cannot possibly utilize all that
bandwidth. However, because SATA II will allow users to connect
multiple disks to the same port using port multiplier, a higher
transfer rate is necessary to be able to connect 4-8 disks to a
single port.
Native Command Queuing enables the
hard drive to take multiple requests for data from the processor and
re-arrange the order to maximize throughput.
Remember the previous scenarios
where the SATA RAID user had a dead disk and did not know which disk
to replace? Enclosure management is the solution to that problem.
The same protocols used for SCSI and Fiber Channel monitoring have
been ported to SATA, completely bridging the gap on this front.
Port Multiplier allows up to fifteen
disks to be connected to the same port. Port Multipliers will likely
attach between four to eight disks to one port. From a cost
perspective, this is a very efficient solution because of savings on
controllers. Other advantages include fewer cables, more efficient
use of space, and scalability of up to 32 disk drives, which was not
possible with SATA 1.0.
Because the similar physical and electrical interfaces design of
SATA and SAS, Serial Attached SCSI system can use either Serial
Attached SCSI disks or Serial ATA disks based on the application
requirements to reduce overall solution cost. Businesses can
seamlessly upgrade to SAS by moving the SATA disks to SAS RAID
system. Server and storage suppliers provide SAS infrastructures
(boxes, backplanes, cables, etc) where a user can either plug in
high-cost, high-performance SAS disks or low-cost,
lower-performance, yet high-capacity SATA disks.
By
Derek Powell,
Network Engineer
Back to Top 
Information Request Form
|
