NET EXPRESS

sales[at]netexpresslabs[dot]com, Silicon Valley, California

Home Page

Contact and OrderInformation

Hard Drives

 

RAID Modules and Bay Coolers
Protect your hard disk investment

7200RPM and 10,000 SCSI hard disks are one the most expensive investments. They are also the most intolerant components to high temperatures. At temperatures above 130F disk failures will occur. During periods of peak disk activity disks heat up at alarming rates. Keeping the chassis temperature down can prevent disk failure and reduce the risk of data loss. RAID Hot Swap Modules and Bay Coolers can be used with any SCSI disk in any standard 5.25" chassis bay. They reduce the disk and chassis temperature by increasing ventilation.

RAID Hot Swap Modules may be used with a RAID controller or just a standard SCSI controller. High RPM Cheetah disks can initiate a high resonance frequency in some chassises. The RAID Hot Swap Modules dampen vibrations on high RPM Cheetah disks per Seagate's recommendation. Their solid aluminum construction assists in conductive heat dissipation while the two 40mm fans aid in heat convection.


JNL DataPort VI SCSI Hot Swap RAID Modules


California PC Products Bay Cooler

External SCSI Enclosures and RAID Enclosures

These are bare internal drives. For a variety of External or RAID solutions see our Tower Chassis, Rack Mount Chassis and External SCSI case pages. We provide a pre-configured RAID solutions for all operating systems. Our cable page lists cables and mounting hardware.

Western Digital and IBM Confirms Five Year Warranty on OEM SCSI Disks

Because the hard disk is the most likely component to fail in a computer system, you should consider the implications of warranty. This year we have been extremely pleased with the reliability and the performance of the Western Digital 7,200RPM SCSI disks and the IBM 7,200RPM and 10,000RPM SCSI disks. 

Western Digital has confirmed that they will honor a five year warranty on all their SCSI disks including OEM models. They will also allow RMA directly to the factory for warranty repair or replacement. Cross shipping is allowed with a credit card. Direct factory RMA's are much faster than RMA's through channel distributors.

IBM is honoring a five year warranty on all 7,200 RPM and 10,000 RPM disks that we sell. However, they do not honor a five year warranty on all OEM disks. For example IBM disks sold via Dell and Gateway have only a one year warranty according to their RMA department. They claim these vendors do not adequately care for their components. 

Quantum and Seagate will not allow direct factory RMA's on many of their products. Moreover they only honor a one year warranty on many OEM disks.

Western Digital has only been selling SCSI disk since January, 1997. However, we are pleased to report that their products have proved extremely reliable to date and their performance is excellent. IBM has been pushing the envelop on new disk drive technologies this year and has released some very exciting products.

Free Items

We will provide the following items for free but only if you ask for them on your order form:
o Screws
o 50-pin SCSI cable
o Internal Wide SCSI (68-pin) to Narrow SCSI (50-pin) Adapter (Adapter only sold with wide drives that need to be converted to 50-pin drives).

Contact Information

For your reference we have provided the telephone numbers for the RMA Department of each disk manufacturer:

Seagate

(800)468-3472

Fujitsu

(800)626-4686

Quantum

(800)345-3377

Western Digital IDE

(800)832-4778

IBM

(888)426-5214

Western Digital SCSI 

(888)932-7274

Samsung

(201)935-2520

 

Technical Considerations

 

Setting SCSI ID and Termination

Up to 7 or 14 SCSI devices can be connected each with a unique SCSI ID jumpers. All SCSI devices use the same binary logic to assign SCSI IDs. A three bit number maps to eight possible SCSI IDs (2^8). Three jumpers control this. On CD-ROM's they are labeled ID1, ID2 and ID4. On hard drives they are labeled ID0, ID1 and ID2. SCSI IDs order has nothing to do with termination. Many people think they must terminate the devices with the highest or lowest SCSI ID. This is NOT true. You terminate the devices on the opposite ends of the SCSI cable!

Jumpers 1 => 3
ID0	ID1	ID2	SCSI 	Hard Drive
ID1	ID2	ID4	ID	CD-ROM
0	0	0	0
0	0	1	1
0	1	0	2
0	1	1	3
1	0	0	4
1	0	1	5
1	1	0	6
1	1	1	7

Jumper on  = 1
Jumper off = 0

In SCSI chains, the voltage is maintained over up to 7-21 feet of internal and external cables with the use of terminators. Terminators must be placed on opposite ends of the SCSI cable set. If the SCSI cable is at one end of the SCSI cable set than it is terminated by placing a jumper labeled TE. Terminated devices maintain the voltage in the cable reducing the single to noise ratio. All SCSI devices between the two ends of the cables must be un-terminated by removing both the TE jumper and the two resistor packs.

TE = Termination Jumper on = Enabled Jumper off = Disabled

Note that on the NCR the TE jumper settings are reversed so that Jumper off = Enabled.

Also note that you must not only remove these jumpers, but you must actually remove the resistor packs to de-terminate a device.

These settings are labeled on top of the Toshiba CD-ROM. They are also labeled near the jumpers:

|      |    |     |    |        |     |
ID1	ID2  ID4  PRTY PRV/ALW  TEST   TE

On NEC CD-ROMs They are labeled:

|     |     |     |     |       |
1     2     3     4     5       6
ID1  ID2  ID4 Reserved TE  SCSI-2/SCSI-I

The jumper labeled PRV/ALW will Activate/De-activate removal of the CD-ROM Caddy:


PRV/ALW:
Jumper on  = Disables Eject Button; 
             Prevents Caddy from being Removed
Jumper off = Allows removal of Caddy

PRTY:
The jumper labeled PRTY
will Activate/De-activate Parity Checking
on the SCSI Bus. Even if the parity
Check is disabled, the drive will still
develop output data parity. 


PRTY:
Jumper on  = Drive executes Parity Check
Jumper off = Drive does not execute Parity Check
             Prevents Caddy from being Removed

TEST:
The jumper labeled TEST
Selects mode of operation. When jumper is present,
audio disc reproduction mode is possible (note: SCSI-ID
must be set to 0). When in Audio disc reproduction mode,
commands from the host computer are ignored.

TEST:
Jumper on  = Audio signal reproduction mode
	Pushing Eject for less than one second skips to next track.
Jumper off = Normal Operation

SCSI-2/SCSI-I:
Jumper on  = SCSI-I
Jumper off = SCSI-II

Reserved:
Leave this jumper off. Its for factory use only.

Hard Disk Performance Guide

Many people make the mistake of buying too much processor power at the expense of RAM, cache and hard disk I/O. In many cases the amount of RAM and the quality of controller and hard disk will have more impact on real world speed than processor speed.

When comparing different drives it is important to remember a few salient points. In general five factors are most important: average rotational latency (dependent on the RPM of a drive), cache buffer size, seek time, disk to buffer data transfer rate and buffer to host data transfer rate.

Hard disks are composed of up to 10 spinning aluminum-alloy platters with a magnetic coating. Each platter has two recording surfaces. Each recording surface has one head for up to a total of 20 heads. The platters are logically divided into concentric circles called tracks. The collection of all tracks at a given head position on all platter surfaces are collectively known as a cylinder. Thus the number of cylinders is the same as the number of tracks per surface. Each track is subdivided into 60 to 180 sectors. The number of sectors per track are greater on the outside rim of the platter and lesser near the center of the disk (in multiple zone recording). Within each sector 512 bytes of data are stored. Combining these ideas, we can see that the capacity of a drive is given by:

Drive Capacity (MBs) = (Bytes/Sector)x(Sector/Track)x (Cylinders) x (Heads) x(1MB/1048576Bytes)

Many advertisers list only a hard drives Buffer-to-Host Transfer Rates. These rates are actually the theoretical maximum through-put of a drives interface:

 
SCSI.............................5  MB/Sec
Fast SCSI.......................10  MB/Sec
FAST WIDE SCSI..................20  MB/Sec
Fast Ultra SCSI.................20  MB/Sec
Fast Ultra Wide SCSI............40  MB/Sec
ATA: IDE.........................4  MB/Sec
FAST ATA-I:  EIDE PIO MODE 3....11.1MB/Sec
FAST ATA-II: EIDE PIO MODE 4....16.6MB/Sec

There are two major types of drive interfaces. The first is the Intelligent Disk Electronics (IDE) interface (also known as ATA (Advanced Technology Attachment). The second is the Small Computer System Interface (SCSI pronounced scuzzy). SCSI is used in high performance servers.

However, most drives do not actually transfer data at these rates. Most Drives are generally slower. The net effective data transfer rate is effected by two factors: electromechanical motion (access times) and electronic functions (Disk and Host Transfer Rates).

Electromechanical motion is critical in determining transfer rate. The actual net transfer rate is given by:

Transfer Rate in MB/Second = (Bytes/Sector)x(Sector/Track)x(RPM)x(1Min/60Sec)x(1MB/1048576Bytes)

If you apply this formula you will find that many drives can not sustain data transfer anywhere near to the Advertised SCSI or EIDE Transfer rates! In order to fairly compare two drives you must compare the respective RPM's and the sector and byte densities.

Many advertisers also over emphasize the importance of random average seek times. 65% of all data is read sequentially in real world applications whereas 35% is actually random. This indicates that in faster drives average rotational latency accounts for a substantial percentage of the access time. The rotational latency is the time it takes, on average, for target data to spin towards the head once it has reached the appropriate track. Since on average the head lands 1/2 a turn away from the target data we find that:

Rotational latency = 1/2 x (60Min/Sec) / (Revolutions/Min).

Indeed, track to track (sequential) seek read/write times are more important than random average seek times. Note that this indicates that 7200RPM drives have a substantially greater speed than there 5400RPM counterparts.

Random average seek times come in two flavors: read and write. The former accounts for 60% of all disk operations. A reasonable formula to compare drives is:

(0.21 x Random Average Read Seek Time) +
(0.14 x Random Average write Seek Time) +
(0.39 x Track to Track Seek Read Time) +
(0.26 x Track to Track Seek Write Time) +
(Average Rotational Latency)

The primary rate determining electronic function is the buffer or cache size. On ISA, EISA, MCA and NuBus bus architectures, disk to buffer data transfer rates are rate determining for most SCSI systems while buffer to host data transfer rates are more critical for IDE drives. However, the PCI bus allows burst rates of up to 135MB/sec. This means that faster buffer to host data transfer rates are now the rate limiting steps for both SCSI and IDE drives on a PCI bus.

Moreover IDE and SCSI buffer to host data transfer rates are not directly comparable. The IDE and the EIDE buffer to host data transfer rates are overestimated because they include large command overheads. In contrast, SCSI drives will deliver much more data than an IDE or EIDE at the same buffer to host data transfer rate.

There are currently three independent methods used to double the SCSI host data transfer rate:

  1. Fast SCSI (doubles the clock speed from 5MB/s to 10MB/s)
  2. Wide SCSI (doubles the data path from 8-bit to 16-bit)
  3. Ultra SCSI (double the clock speed again)

Because these methods are orthogonal, these two techniques can be implemented together to increase the host data transfer by up to eight times (40MB/Sec). It is important to note that 8-bit SCSI drives use a 50 pin connector and Wide 16-bit SCSI drives utilize 18 extra pins (8 pairs for the data path and one pair of grounds) for a total of 68 pins. Therefore a 68 pin controller such as the Adaptec 2940W or the BusLogic 956C are needed to run both Wide 16-bit SCSI and 8-bit SCSI. Each controller sports one 50-pin and two 68-pin connectors. SCSI-III is NOT the same as WIDE SCSI. There are SCSI-II and SCSI-III versions of both WIDE SCSI and 8-bit SCSI. SCSI-III has added on some new commands and standards to the SCSI-II set. 8-bit SCSI-III and WIDE SCSI-III are both backward compatible with SCSI-II. (SCSI-II is a subset of SCSI-III).

Enhanced IDE (EIDE) uses similar clock doubling methods to achieve its PIO mode 3 and 4 rates. However, EIDE is not well designed for preemptive multitasking. EIDE drives can not do more than one task at a time. In contrast, SCSI devices are handle more jobs and SCSI controllers can tag queue several commands. Furthermore, since there are no terminators used in EIDE, the cable is limited to only 18 inches and can only connect to two devices per controller. The two devices are differentiated by two jumpers marking one drive as master and the other as slave. In many respects, EIDE can not compare to SCSI. Indeed, SCSI is the de facto standard of all high end servers. It is interesting that the cost to manufacture SCSI drives is not more than EIDE.

Keep in mind that each SCSI channel will start to slow down as more devices are added. For this reason, high-end servers are designed with two or more SCSI Buslogic or Adaptec controllers. This will substantially increase speeds under all operating systems. The boot drive must be on the bottom controller. Similarly, one should put EIDE drives and CD-ROMs on separate EIDE controllers if available.

On high end servers, disks can be striped together to form on logical disk with double the transfer rate (RAID ONE). Disks can also be mirrored for redundancy or combined into a single volume (RAID ZERO). Under parity striping, three or more disks can be striped and the last disk can be used to store an XOR of the all the disk so that if a drive should fail, the data on any drive can automatically be reconstructed (RAID FIVE). You can do this with free software under NT, Linux and BSD. You can find drive specs from our WWW links.

Comparing 9GB SCSI Disks

 

Supplier
Product
Western Digital
WD Enterprise
WDE9100
Seagate
Barracuda 9
ST19171
Quantum
Atlas II
XP39100
IBM
Ultrastar 2XP
DCHS-39100
Micropolis
Tomahawk 9
3391
Fujitsu
M29X Series
M2949
Hitachi
DK308H Series
DK318H-91
               
Capacity (MB) 9105 9100 9100 9100 9100 9100 9100
Form Factor 3.5",
low-profile
3.5",
half-height
3.5",
half-height
3.5",
half-height
3.5",
half-height
3.5",
half-height
3.5",
half-height
               
Configuration              
Disks 6 10 10 9 11 10 10
Heads/Channel 12,
MR/PRML
20,
MR/PRML
20,
MR/PRML
18,
MR/PRML
22,
MR/PRML
18+1, MR/PRML 20,
MR/PRML
Areal (Gb/in2) 1.10 N/A 660 829 N/A 766 845
               
Performance              
RPM 7200 7200 7200 7200 7200 7200 7200
Track-to-Track Seek (ms, read) <0.8 0.6 1.0 0.7 1.0 1.0 1.3
Average Seek
(ms, read)
<7.9 8 8 8.5 7.9 9 9
Full Stroke Seek (ms) <17 19 <18 18 19 N/A 19
Internal Data Rate (Mb/s) 99 to 160 80 to 124 up to 121 82 to 123 76 to 125 79 to 119 75 to 125
               
Interface              
SCSI Support SCSI-3 SPI Compliant
Ultra SCSI
SCSI-3 SPI Compliant
Ultra SCSI, FCAL
SCSI-3 SPI Compliant
Ultra SCSI
SCSI-3 SPI Compliant
Ultra SCSI
SCSI-3 SPI Compliant
Ultra SCSI
SCSI-2
Ultra SCSI
SCSI-3 SPI Compliant
Ultra SCSI
Buffer Size 512 KB, 1 MB opt 512 KB, 2 MB opt 1 MB 512 KB, 1 MB opt 512 KB, 2 MB opt 506 KB 512 KB
               
Reliability              
MTBF (hours) 1,000,000 1,000,000 1,000,000 1,000,000 1,000,000 1,000,000 1,000,000
               
Acoustics              
Idle (A-bels) 4.0 4.6 4.2 5.0 4.8 N/A 4.5
               
Power              
Typical Power (w), Idle 9.4 13 12 15.6 14 15 13.6
               
Production Date October 1997 July 1996 October 1996 January 1996 October 1996 1997 January 1997

 

Comparing 4GB SCSI Disks

Supplier
Product
Western Digital
WD Enterprise
WDE4360
Seagate
Barracuda 4LP
ST34371
IBM
Ultrastar 2
DCHS34550
Quantum
Atlas 2
XP34361
Fujitsu
M2954
Capacity(MB) 4360 4350, 4500 (opt) 4550 4360 4400
Form Factor 3.5", low profile 3.5", low profile 3.5", low profile 3.5", low profile 3.5", low profile
Configuration
Disks 4 5 5 5 5
Heads 8, TFI/PRML 10, MR/PRML 9, MR/PRML 10, MR/PRML 9+1, MR/PRML
Areal (Mb/in2) 769 660 829 636 767
Performance
RPM 7200 7200 7200 7200 7200
Track to Track Seek
(ms)
1 0.8 0.5 <2 N/A
Average Seek (ms,
read)
8 8 7.5 <7.9 8
Full Stroke Seek (ms) <18 17 17 <18 14
Internal Data Rate
(Mb/s)
83 to 140 75 to 120 81 to 123 70 to 106 76 to 112
Interface
SCSI Support SCSI-3 SPI
Compliant
Ultra Supported
SCSI-3 SPI
Compliant
Ultra Supported
SCSI-3 SPI
Compliant
Ultra Supported
SCSI-3 SPI
Compliant
Ultra Supported
SCSI-3 SPI
Compliant
Buffer Size 512 KB, 1 MB opt 512 KB 512 KB 512 KB 512 KB
Reliability
MTBF 1,000,000 1,000,000 1,000,000 1,000,000 1,000,000
Power
Nominal Voltage (V) 5, 12 5, 12 5, 12 5, 12 5, 12
Typical Power (w)
Idle
7 10 11.2 9 10

 

Ultra SCSI and Fast SCSI are backwards compatible

Ultra SCSI and SCSI-III drives will work with older non-Ultra and SCSI-II controllers in addition to the newer Ultra SCSI-III controllers.

Acoustical Considerations

The penultimate generation of SCSI disks including most Fast SCSI-II and Fast Wide SCSI-II 7200RPM drives produced very loud, high pitches and they also produced a lot of heat. This older generation of drives were designed for the high performance server market where acoustics were not as much of a consideration as data through-put. Spinning disks release energy in the form of sound and heat. Heat can be destructive to the drive and sound can be annoying. Generally sound volume and thermal dissipation increase as the RPM and number of heads increase. Newer Ultra and Ultra Wide SCSI disks use less heads, have a lower frictional coefficient and have a higher surface to volume ratio. Therefore, they are much quieter and produce much less heat. This is an important accomplishment in the hard disk industry and may be even more important than the mere improvement in speed brought about by the Ultra SCSI interface. These drives should fail less and be more enjoyable. Indeed, in recent tests the new Atlas-II Ultra Wide SCSI disks and the Western Digital Ultra Wide SCSI disks proved quieter than older Fast and Fast Wide SCSI-II 4500RPM, 5400RPM and 7200RPM drives. Newer disk idle at 32-40dba at one meter depending on The model. Mounting the disk inside a computer chassis reduces noise even more so these new disks are quite acceptable.

If you are concerned about background noise you should buy an Ultra or Ultra Wide disk. Ultra and Ultra Wide SCSI disks demand a premium price right now, but they are well worth it if you can afford one.

Evaluating the Silent System, Inc.'s "Silent Drive" kit

Over several months our Berkeley staff and the Stanford Computer science staff corresponded with Silent Systems about their Silent Drive product. We were concerned that this product would not provide adequate ventilation for a hot SCSI disk and could lead to disk failure. Silent Systems insisted that their product was safe and would work with even the hottest 7200RPM SCSI disks. We asked for this in writing as we intended to buy and mount thousands of dollars in disk with these kits. They signed a letter stating their product would be safe for even the hottest 7200RPM SCSI disks.

We decided to evaluate Silent System, Inc.'s claims about the "Silent Drive" kits. These units are designed to reduce acoustical output from computer hard disk drives while maintaining adequate ventilation.

We evaluated the Silent Drive kit with an Atlas-II XP34550W 4.5GB Ultra Wide SCSI 7200RPM Quantum disk (rated 34 dba at 1M, idle) mounted in a standard Enlight Medium tower ATX case (Model 72300M). Three fans where installed: two 80mm muffin fans and one PC Power and Cooling 3.2 CPU fan. We tested the unit under Linux running on a Pentium Pro 200 exercising disk access by transferring 1GB of data with the UNIX dd command. Room temperature was 68F. Quantum and Western Digital both specify maximum operating temperature of 131F for their 7200RPM models. Higher temperatures can cause disk failure.

After only a short period of testing we measured disk temperatures of 139F and climbing. The rate of temperature increase showed no sign of reaching a plateau. Therefore we quickly stopped our tests to prevent disk failure. Moreover the unit mounted to the chassis with rubber screw mounts that easily broke off and in our opinion were inadequate to protect an expensive disk from dislodging. We stopped short of collecting acoustical data as the unit had in our opinion already proved a failure. However, we noticed little difference in the acoustic output when comparing this test system to an identical system without the Silent Drive kit.

We asked Silent Systems to retract their claim that their Silent Drive product can adequately ventilate 7200RPM SCSI disks. We also asked them to post information about this on their WWW site. After several weeks they have finally modified their WWW site and retracted their claims about supporting 7200RPM disks. However, they do continue to claim support for 5400RPM SCSI disks. In my opinion this is not the case. We don't feel that their product can adequately ventilate a 5400RPM SCSI disk under a full load in real world conditions without causing disk damage.


Copyright (c) 1989 - 2008 Net Express All Rights Reserved.