Go to www.Cdincorp.com

Redundant Array of Independent Disks (RAID) for Small to Enterprise Storage Solutions

NEWS Announcement 


by C Jones

New phenom RAID configurations are evolving and coming of age as we write. One particularly adventurous and promising Solid State 0-1 RAID configuration demonstrated by Next Level Hardware (nextlevelhardware.com) includes:

9 X Mtron 16GB SSD's (Solid State Drive's) in Raid 0
Intel QX9650 Processor , Corsair Dominator DDR3-1800, 
Gigabyte X38T-DQ6, Sapphire HD 2900 XT, Silverstone OP1000
Raid Controllers Used: Areca ARC-1220, Areca ARC-1231ML
(Alternate  HDD's Used for comparison: Maxtor DiamondMax 300GB IDE, WD Raptor 150 SATA)

Results in a single drive configuration yielded 30-40% speed increase over platter drives for the Mtron, 800% using an Enterprise controller with a bay of 9 16GB SSDs; effectively a level "8" cache hard disk scenario. 

The Mtron Solid State Drive displayed an average read increase of almost 30 MB/s over the mechanical WD Raptor 150 (15000 RPM) in a single drive configuration. When scaled under RAID 0, is 120MB/s or higher. Scaling 5 SSD drives to 600 plus MB/s will cause a performance over Mtron's spec of 120 MB/s sustained read. A 9 SSD drive RAID 0 config will output an 830 MB/s sustained read, better or best world scenario.

So, in a carefully constructed enterprise situation utilizing a properly configured high end raid controller, SSDs will most like do almost 85,000 at a random 4k read, and 100,000 IOP's at a random 512 byte read. This phenomenal speed increase with the use of solid state drives and surpasses current implemented technology by over 10 fold. 

Power Consumption.
The Mtron drive uses a max 2.95 watts load as a single drive. The Raptor uses 10 watts. So, for 25 watts of load to power 9 Mtron drives at RAID 0, you can run only 3 Raptors. Considerable power savings.

Write Endurance
Some experts have stated that NAND based SSD's can deteriorate rapidly in a server environment that maxes out drive write space repeatedly.  However, OEM SSD manufacturers report common themes to address this by dynamically reallocating physical blocks to logical blocks and preading the load transparently across the entire disk. Unless you wrote to the entire disk for the entire number of life read cycles, which can reach 140 years at a daily write of 50GB, the drive will continue to perform.

Additionally, manufacturers use patented architectural algorithms that rewrite in a more complex wear leveling as well as having a higher percentage of spare flash blocks.  These are invisible to the host interface nor do they show up as spare storage, but, when block limits are reached, the data is invisibly switched to the spare blocks to relieve the usable blocks. These methods can offer a 4 times greater lifetime use rate than other enterprise flash devices and up to 100 times greater life rate usability than intrinsic flash memory. 

The Mtron Solid State Drive rules the roost with ultimate speed performance. Long range use in a server scenario might yield different stats than those on paper. Until testing is completed, real world usage will be tentative until a solid performance record is achieved.

There are a number of manufacturers that are jumping on the SSD bandwagon. Hopefully, the RAID controllers will follow suit and accommodate the lightning fast high performance ability of the SSDs.

For almost instantaneous data flow management and access, NAND SSD is being implemented by several Web services providers for database hosting.

For high end enterprise clients demanding superior performance and time sensitive data access, this new technology promises to provide a new generation of global millisecond data transfer and retrieval.
HDTach Scores for the Mtron
Make WD Scorpio Seagate Barracuda WD Raptor Mtron MSD-P25016
Avg. Read 45.5 MB/s 48.4 MB/s 75.4 MB/s 92.4 MB/s
Random Access 17.7 ms 14.2 ms 8.2 ms .1 ms
Burst 112.5 MB/s 116.8 MB/s 127.4 MB/s 93.8 MB/s
CPU Utilization 2% 1% 5% 2%
Mtron Stats
Make WD Scorpio Seagate Barracuda WD Raptor Mtron MSD-P25016
Model WD2500BEVS ST9160823ASG WD1500ADFD MSD-P25016
RPM 5,400 7,200 10,000 N/A
Platters /Heads 2/4 2/4 2/4 2/4
Interface SATA150 SATA300 w/ NCQ SATA150 SATA150
Buffersize 8MB 8MB 16MB 8MB

3 years OEM/1 Year retail

5 years 5 years 5 years
MaximumPC dropped the Mtron repeatedly from a distance of 3 inches while running backup. They also rattled it on a hard surface from different angles. Ordinarily, a disk drive would stumble, lose data and/or die. Not so the Mtron. The drive was completely unfazed by the mistreatment.

RAID, in General
  • The Benefits of RAID provide real-time data recovery with uninterrupted access when a hard drive fails
  • Increases system uptime and network availability
  • Protects against data loss
  • Multiple drives working in parallel increase system performance

In the past computer systems were often restricted to writing information to a single disk.  This disk was usually expensive and prone to failure. Hard disks have always been the weakest link in computer systems because the devices are the only mechanical member of an otherwise all electronic system.  The disk drive contains a mass of moving mechanical parts operating at high speed.  The question is not whether a drive will fail,  but when a drive will fail.

RAID was designed to revolutionize the way computers managed and accessed mass storage of data by providing an inexpensive and redundant system of disks.  Instead of writing to one single large expensive disk,  RAID writes to multiple independent disks.   RAID accomplishes its goals of redundancy and fault tolerance by doing two things;  one is striping and the other is parity checking.  Striping means that files are written a block at a time over multiple disks.  The striping technique devides data across many drives and improves data transfer rate and total disk transaction times.  Such systems are good for transaction processing, but suffer from poor reliability because the system is only as reliable as the weakest individual drive.

Parity checking ensures that the data is valid by performing a reundancy check on all data following a transmission.  With parity, one of the disks on a RAID system can fail and the other disks have the ability to rebuild the failed disk.  In both cases, these functions are transparent to the operating system.  A Disk Array Controller handles both striping and parity control.

Adaptec has three powerful new microprocessor-based Ultra160 SCSI RAID controllers for entry- and mid-range servers. The Adaptec SCSI RAID 2100S, Adaptec SCSI RAID 3200S and Adaptec SCSI RAID 3400S controllers deliver reliable, high-performance data protection to support critical business applications.

These new controllers are part of Adaptec's "RAID Everywhere" initiative, a program aimed at making RAID ubiquitous on PC servers. "RAID Everywhere" is a revolutionary initiative that focuses on driving RAID costs down, making the technology easier to use, and more available.

An Intel solution is the IntelŪ RAID Controller SRCU42L, which offers the latest SCSI technology, Ultra320, in a low-profile form factor for high-density server applications. Featuring the IntelŪ 80303 I/O processor, the Intel RAID Controller SRCU42L includes two Ultra320 SCSI channels (one internal and one external)ūsupporting up to 30 SCSI devices (15 per channel). This feature-rich RAID controller enables you to deliver storage solutions with very high availability and enhanced data protection. The IntelŪ RAID Controller SRCU42L provides the latest in technological innovation. Featuring the IntelŪ 80303 I/O processor, the low-profile RAID Controller SRCU42L includes two Ultra320 SCSI channels, 64-bit/66MHz (3.3V and 5V) PCI support, and 64MB embedded ECC SDRAM, and it supports high-density solutions, and RAID levels 0, 1, 4, 5, and 10, and JBOD.

For entry-level and midrange Intel-based servers, the Intel Server RAID Product makes exceptional data protection and manageability affordable . Intel's Server RAID Controller U3-1  provides an outstanding combination of RAID price/performance with a 64-bit PCI connector for enhanced data throughput.  This controller is an Ultra160 LVD (Low Voltage Differential) SCSI, single-channel controller that supports a burst data transfer rate of up to 160 Mbytes/second. Designed and validated for Intel server boards and platforms, the SRCU31 controller enables ease of integration and flexible support for RAID levels 0, 1, 5, and 10 for the optimum balance of fault tolerance and data throughput. The high-integration Intel® i960® RN I/O processor supports best-of-class RAID performance.

The SRCU31 controller with Intel Server RAID Storage Console software enables end-users at all experience levels to choose from typical or more advanced RAID features. Examples include online capacity expansion, online drive roaming, and online RAID level migration without rebooting the system. With support for hot-plug drives, global hot spares, and automatic rebuild, data remains available, even in the event of a disk failure. Users can add a new hot-swap hard drive, do a bus scan and have the new disk available for RAID array expansion without interrupting workflow to reboot the system.

  • Grant selective network access to authorized remote and corporate users
  • Authenticate network users with strong authentication techniques before granting access to sensitive corporate data
  • Ensure the privacy and integrity of communications over untrusted, public networks like the Internet Provide content security at the gateway to screen malicious content, such as viruses and malevolent Java/ActiveX applets
  • Detect network attacks and misuse in real time and respond automatically to defeat an attack Protect internal network addressing schemes and conserve IP addresses
  • Ensure high availability to network resources and applications
  • Deliver detailed logging and accounting information on all communication attempts

With its Enterprise Security Management product family, Check Point Software Technologies offers a comprehensive set of solutions that meet these demanding requirements. Check Point’s FireWall-1/VPN-1 security suites enable all functionality to be deployed and managed with a single enterprise-wide security policy for straightforward management and administration.

Enterprise security solutions are unified by Check Point's OPSEC [Open Platform for Secure Enterprise Connectivity] policy management framework which provides central integration, configuration and management for Check Point products as well as other third-party security applications. Only Check Point provides organizations with the ability to define a single, integrated security policy that can be distributed across multiple gateways and managed remotely from anywhere on the enterprise network. There is never any need to individually reconfigure each security gateway.

All of Check Point’s security solutions are built on Stateful Inspection, the de facto standard for network security that was invented and patented by Check Point. Stateful Inspection provides full application-layer awareness without requiring a separate proxy for every Internet service and protocol. This provides unparalleled performance, scalability, and the ability to support new and custom applications and services quickly and easily. CDI can provide you with Intel RAID storage systems, service and support.
Additional CDI Favorites

As the process of managing IT struggles to keep pace with new technology, and storage and networking infrastructure demands increase, the industry has no choice but to unleash a plethora of new storage management tools and techniques.

Adaptec's long-anticipated Ultra320 bus, the most advanced interface available, takes a step toward meeting today's ever-changing speed and storage requirements. The Ultra320 SCSI is the next step in the SCSI evolution with a bus that has a staggering maximum data-transfer speed of 320 Mbps.

CRN Test Center engineers tested the integrity of the Ultra320 bus against Adaptec's Ultra160 bus. Both Adaptec SCSI cards (39320D and 29160N) were installed into a Hewlett-Packard NetServer LH 3000/3000r running Microsoft Windows Advanced Server 2000. The 64-bit slot allowed Test Center engineers to achieve the full bandwidth of the Ultra320 card. Once the cards were installed, two SCSI cables were attached, each with four unformatted 36.7-Gbyte Maxtor Atlas 10K III Ultra320 drives, for a total of eight drives.

Each card's performance was measured using Iometer from Intel, which can read and write data to a storage device according to script. An 11-stage script was used, and each stage had different settings for data-block size and data-position randomness. As Iometer moved from one stage to the next, the script decreased the block size while increasing the randomness of the data's position.

The first stage simulated a data-streaming application, which typically involves large data sets that are accessed sequentially. The last stage simulated database and multiuser access, which typically involves excessive seek time and small data sets. The settings of the other stages represent everything between those extremes. All stages were set to produce an equal number of reads and writes.

The block size of the first stage was set to 1 Mbyte and decreased by half, using geometric regression, each time Iometer moved to another stage. Simultaneously, the script increased the data-position randomness by 10 percent as it moved from one stage to the next. The first stage had zero percent randomness; the second had 10 percent, and so on, until the 11th stage, which invoked 100 percent data-location randomness.

CRN Test Center engineers found there were distinct advantages when it came to using the Ultra320 bus in a multiple-drive array. Overall, there was a 9 percent performance increase in operations per second and a 16 percent performance increase in throughput. The Ultra320 bus shined brightest in instances when the data was more random and the blocks of data were smaller. Once randomness hit the 70 percent mark, the Ultra160 bus started to drop off at an alarming rate in both operations per second and throughput.

For example, under operations per second, when block size was 2 Kbytes and randomness was at a 90 percent clip, the Ultra320 card outperformed the Ultra160 card by an amazing 47.65 percent. At 4-Kbyte block size and 80 percent randomness, the Ultra320 card surpassed the Ultra160 at a rate of 33.92 percent.

The added bandwidth really showed in the area of throughput. When the block size was 1 Kbyte and randomness was at 100 percent, the Ultra320 operated 82.12 percent better than the Ultra160. Also, the maximum throughput under the Ultra320 card was 156.21 MBps, while under the Ultra160 the maximum throughput was 85.77 MBps,a difference of 55 percent.

The Ultra320 bus can be advantageous in database use. Companies that have many employees accessing a centralized server would be able to cut down on the delay time in retrieving and transferring files. The card also would facilitate the transferring of large amounts of data onto SANs or external devices much more quickly and efficiently.

CRN engineers found that testing the bus speed with a single drive attached showed no distinct advantage between drives of low rotational velocity and density. For example, attaching a single 36.7-Gbyte Maxtor Atlas 10K III Ultra320 drive to either the Ultra320 or Ultra160 bus showed no significant difference. However, testing the buses on drives with higher density and a faster rotational velocity was an entirely different story.

Test Center engineers tested Maxtor's 73.4-Gbyte Atlas 10K III Ultra320 drive in a head-to-head comparison with Seagate's 15K ST373453LC, 73.4 Gbytes in density. Results from top to bottom were overwhelming, regardless of block size or randomness. The percent difference in operations per second in comparing the two drives ranged between 31.64 percent and 78.72 percent. The pattern was the same as in the earlier multiple-drive array tests,the percent difference increased when data was more random and the blocks of data decreased. The maximum throughput under the Ultra320 card was 57.30 MBps, while under the Ultra160 the maximum throughput was 32.06 MBps.

Some might argue that this is an unfair comparison, since one drive is spinning faster. But because the drives' densities were the same, Test Center engineers felt the margin of difference should not have been so great, even with a difference of 5,000 rpm. Whether using it as a single drive or in an array, the Seagate 15K ST373453LC drive will offer the optimum performance.

If speed is a necessity and price is not an issue, the Seagate drive is the better choice. The price of the drives increases as drive density and rotational velocity increase. The current suggested retail price of Seagate's 15K ST373453LC is $939, while Maxtor's 73.4-Gbyte Atlas 10K III Ultra320 drive is priced at $799. Depending on a customer's deployment, solution providers must decide whether the added speed is worth the price difference.

If a customer needs a multiple-drive array and price is a concern, opting for the Ultra320 bus with the Maxtor drives would offer a modest array. If price is not a concern, Test Center engineers would recommend the Seagate 15K ST373453LC drives in a multiple-drive array, getting the maximum performance out of their platform.

As data continues to grow exponentially, a faster and more efficient means of accessing this data and transferring it to storage will become an issue for more companies. Adaptec's Ultra320 bus is a good choice for high capacity and high reliability, as well as an obvious speed upgrade for those who need it most.



To order, Call or e-mail us at Order@CDInternational.com today.