Understanding RAID Technology

In recent years, RAID technology has transitioned from being a “luxury” component confined to large computer systems, workstations, and servers, to being integrated into desktop computers in a simplified manner. However, while buyers may know that their motherboards (BMC) support RAID technology, not everyone knows how to use it effectively. This article provides basic information about RAID as well as some practical tips for enhancing PC performance.

WHAT IS RAID?

RAID stands for Redundant Array of Independent Disks. Initially, RAID was used as a protective solution because it allows data to be written across multiple hard drives simultaneously. Over time, RAID has evolved into various configurations that not only ensure data safety but also significantly increase data retrieval speeds from hard drives. Below are five commonly used RAID types:

1. RAID 0

This form of RAID is popular among users due to its ability to enhance hard drive data transfer performance. Requiring a minimum of two hard drives, RAID 0 enables the computer to write data across them using a special method called Striping. For example, if you have 8 data segments numbered from 1 to 8, the odd-numbered segments (1, 3, 5, 7) will be written to the first hard drive, and the even-numbered segments (2, 4, 6, 8) will be written to the second. To simplify, imagine you have 100MB of data; instead of storing all 100MB on a single hard drive, RAID 0 allows you to store 50MB on each drive, theoretically reducing the processing time by half. Consequently, if you have 4, 8, or more hard drives, the speed will increase even further. However, despite its appeal, RAID 0 carries a risk of data loss. This is primarily due to the way information is fragmented; since data is not stored completely on any single hard drive, when the computer needs to access information (say a specific file), it must compile it from multiple drives. If one drive malfunctions, the associated file becomes irretrievable. Fortunately, modern technology has produced fairly durable hardware, so instances of data loss are not very common.

RAID 0 is particularly suitable for users who require rapid access to large amounts of data, such as gamers or graphic and video professionals.

2. RAID 1

This is the most basic RAID type that ensures data safety. Like RAID 0, RAID 1 requires at least two hard drives to operate. Data is replicated identically on both drives (Mirroring). Should one drive fail, the other continues to function normally. You can replace the faulty drive without worrying about data loss. With RAID 1, performance is not the primary concern, so it is not surprising that it is not the top choice for speed enthusiasts. However, for network administrators or those managing critical information, RAID 1 is indispensable. The final capacity of a RAID 1 system is equivalent to the size of a single drive (two 80GB drives in RAID 1 will be seen as a single 80GB RAID drive).

3. RAID 0+1

Have you ever wished for a storage system that is as fast as RAID 0 and as safe as RAID 1? You are not alone in this desire. That is why the RAID 0+1 configuration was created, combining the advantages of both “big brothers.” However, the cost for such a system is quite high, requiring a minimum of 4 hard drives to operate RAID 0+1. Data will be written simultaneously across 4 hard drives, with 2 drives using Striping for speed and 2 drives using Mirroring for backup. These 4 drives must be identical, and when integrated into the RAID 0+1 system, the final capacity will equal half the total capacity of the 4 drives; for instance, four 80GB drives will yield a usable capacity of (4*80)/2 = 160GB.

4. RAID 5

This is perhaps the most robust RAID type for home and office users, utilizing 3 or 5 separate hard drives. Data and backup copies are distributed across all drives. This principle can be quite complex. Returning to the example of 8 data segments (1-8) with 3 drives now, segments 1 and 2 are written to drives 1 and 2, respectively, while their backup is stored on drive 3. Segments 3 and 4 are written to drives 1 and 3 with their respective backups stored on drive 2. Segments 5 and 6 are stored on drives 2 and 3, with the backup on drive 1, and this pattern repeats, with segments 7 and 8 written to drives 1 and 2, and their backup stored on drive 3 as initially configured. Thus, RAID 5 ensures improved speed while maintaining a high level of safety. The final hard drive capacity equals the total capacity minus one drive. So, if you use 3 80GB drives, the final capacity will be 160GB.

5. JBOD

JBOD (Just a Bunch Of Disks) is not technically a standard RAID type, but it has some characteristics related to RAID and is supported by most RAID controller devices. JBOD allows you to attach any number of drives to your RAID controller (within the limit of available ports). They will then be “aggregated” into a larger virtual hard drive for system use. For example, if you connect 10GB, 20GB, and 30GB drives, the RAID controller supporting JBOD will present it as a single 60GB drive. However, note that JBOD does not provide any additional benefits: it does not enhance performance or provide data safety solutions; it simply connects and aggregates storage capacity.

6. Other RAID Types

In addition to the types mentioned above, you may encounter several other RAID configurations that are not widely used and are typically limited to specific computing systems. Examples include Level 2 (Error-Correcting Coding), Level 3 (Bit-Interleaved Parity), Level 4 (Dedicated Parity Drive), Level 6 (Independent Data Disks with Double Parity), Level 10 (Stripe of Mirrors, the opposite of RAID 0+1), Level 7 (a brand by Storage Computer that allows caching for RAID 3 and 4), and RAID S (an invention by EMC used in their Symmetrix storage systems). Additionally, there are some variants, such as Intel Matrix Storage that allows running RAID 0+1 with only 2 hard drives or DFI’s RAID 1.5 on the BMC 865 and 875 systems. While they have many differences, most of them are improvements of traditional RAID methods.

WHAT DO YOU NEED TO RUN RAID?

To run RAID, you need at least one controller card and two identical hard drives. The drives can be of any standard, including ATA, Serial ATA, or SCSI; ideally, they should be exactly the same due to a simple principle: when operating in synchronous mode like RAID, the overall performance of the system will be limited by the slowest drive present. For example, if you attempt to run a 160GB drive with a 40GB drive (regardless of RAID 0 or 1), you effectively waste 120GB because the controller sees them as a pair of two 40GB drives (except in the JBOD scenario mentioned). The number of drives will depend on the RAID type you plan to implement. The interface standard is not very critical, especially between SATA and ATA. Some newer BMCs allow mixed RAID configurations with both interfaces. Notable examples include the MSI K8N Neo2 Platinum or the DFI Lanparty NForce4 series.

The RAID controller is where the data cables connecting the drives in the RAID system converge, and it processes all data passing through. This controller comes in various forms, from separate cards to integrated chips on the BMC.

For PC systems, while not yet widespread, opting for a BMC with integrated RAID is advisable since this is generally one of the most effective and cost-efficient solutions to enhance system performance, not to mention the data safety it provides. If your BMC lacks RAID support, you can still purchase a PCI controller card on the market at a reasonable price.

Another component of the RAID system that is not mandatory but can be useful is hot-swappable drive bays. They allow you to replace malfunctioning drives while the system is running without having to power down (simply unlock, pull out the bad drive, and insert a new one). This device is often used with SCSI drives and is crucial for server systems that require continuous operation.

Regarding software, it is quite straightforward since most modern operating systems support RAID very well, especially Microsoft Windows. If you are using Windows XP, adding RAID is relatively easy. The most critical aspect is the drivers, which are typically bundled with the device. Setting up RAID may pose some challenges if you lack experience, but solutions are provided later in this article.

There are two scenarios when users upgrade RAID in their systems. If the added RAID system is solely for storage or high-speed data exchange, installation is quite simple. However, if you plan to use it for operating system installation or software, it becomes more complicated, requiring a complete reinstall from scratch.

CUSTOMIZING RAID FOR YOUR NEEDS

1. Choosing the RAID Type

Now that you’ve decided to elevate your system to new heights, selecting the right RAID type is not as straightforward as you might think. Given the conditions in Vietnam, you can choose from several RAID solutions, including 0, 1, 0+1, and 5. Among these, RAID 0 and 1 are the most economical and commonly found on most current BMCs. RAID 0+1 and 5 are typically only available on higher-end, more expensive models.

RAID 0 is undoubtedly the option that delivers the highest speed but is also the most fragile. For example, using 4 hard drives in RAID 0 can achieve data transfer speeds exceeding 100MB/s. This figure is extremely enticing for any PC user. However, the risk of data loss also increases fourfold. Hard drives are mechanical products with moving parts and will gradually “age” over extended use (fortunately, the lifespan of hard drives is quite long). Additionally, power surges or controller errors can lead to disaster. Therefore, RAID 0 should not be used for long-term data storage, but it is the top choice for temporary drives requiring high speed, such as web database storage. If you plan to use it long-term, consider adding a few more drives and switching to a RAID 0+1 setup. This is truly ideal if you have ample financial resources.

RAID 1, when running alone, serves no purpose other than creating an identical backup drive. Regular users may not find RAID 1 appealing, except for those who must store and manage genuinely important documents, such as customer information servers or account data. If using RAID 1, consider adding hot-swappable bays to facilitate quicker data recovery (you can remove the drive and create a backup on the new drive while the system is functioning normally).

RAID 5 is currently the top choice for all types of computers due to its ability to both correct errors and increase speeds. If you plan to build a RAID system with 4 or more hard drives, RAID 5 is certainly the optimal solution.

Combined RAID types, such as RAID 0+1 or RAID 50 (5+0), typically inherit characteristics from their component RAID types; however, be cautious and only use them if necessary, as the cost of components can be quite high. We can summarize as shown in the table.

2. Choosing Hardware

The first step is to select compatible components. For RAID controller chipsets, your options are limited since they are primarily integrated into the BMC. However, you should pay attention to the following points.

Currently, integrated RAID controllers typically come in two main types: controllers mounted on the BMC or those supported directly from the chipset. Common options include:

Integrated Chipsets:
+ Intel ICH5R, ICH6, ICH7. These southbridge chipsets are paired with the i865/875/915/925/945/955 series.
+ nVIDIA nForce2-RAID (AMD), nForce 3 Series (AMD A64), nForce 4 Series (AMD A64/ Intel 775).

External Controllers: Numerous brands exist, such as Promise Technology, Silicon Image, and Adaptec, but the most commonly seen are the Silicon Image Sil3112 and 3114 series.

Integrated chipsets generally have lower latency and are easy to use. However, they often have fewer features, and the software may be limited, sometimes “overutilizing” system resources for read/write tasks. Third-party chip solutions usually have slightly higher latency (which is negligible), more varied software and features, and minimal resource usage; however, dedicated cards are easier to replace or upgrade when necessary. It’s crucial to note that the Silicon Image Sil3112 has relatively poor compatibility, which may lead to data loss when transitioning to other RAID systems. The Sil3114 and higher have resolved this issue. nForce and ICH5, 6, and 7 systems can easily swap hard drives, and their RAID BIOS is smarter, often capable of recognizing pre-formatted RAID drive groups.

Some of the latest BMCs, such as the DFI Lanparty NF4 SLI-DR, also support RAID 5. For desktop systems, you typically only find PATA or SATA interfaces, and because multiple cables are required, SATA is a wiser choice even without considering faster speeds and numerous technological improvements. If finances allow, you might consider some relatively professional products that enable additional RAM to be used as a substantial cache to significantly boost speed.

In terms of hard drives, you should choose ones with high data transfer rates and quick access times. Access Time indicates how long it takes for the hard drive to find the needed data, and this figure should be as low as possible. Additionally, the hard drive should have a large cache (8MB or more); some newer models offer 16MB caches and technologies that can significantly enhance performance (such as Seagate NCQ). Ideally, the hard drives should be identical, as RAID performance can be affected by slow or smaller capacity drives, adhering to the principle of “a single weak link can jeopardize the whole chain.”

3. Setting Up RAID

Setting up RAID is generally straightforward and primarily relies on the BIOS of the motherboard, the RAID controller, and is not particularly difficult.

After connecting the hard drives to the appropriate RAID slots on the motherboard (refer to the product documentation), enter the BMC BIOS to activate the RAID controller and assign the relevant ports (usually found in the Integrated Peripherals section).

After this step, save the settings and reboot the computer. Pay careful attention to the notification screen and press the correct key combination when prompted (it may be Ctrl+F or F4 depending on the RAID controller) to access the RAID BIOS.

In the RAID BIOS, while each type has a different interface (refer to the product documentation), you generally need to perform the following steps:

+ Designate the hard drives to participate in the RAID.
+ Select the RAID type (0/1/0+1/5).
+ Specify Block Size: This parameter significantly impacts the performance of the RAID hard drive array. For RAID configurations utilizing Striping, Block Size also refers to Stripe Size. If this setting is not aligned with your usage needs, it can lead to wasted memory and reduced performance. For instance, if the Block Size is set to 64KB, a minimum of 64KB will be written to the drive in all cases, even if it’s a 2KB text file. Therefore, this value should closely match the average size of the files you use. If your drives contain many small files, such as Word documents, you should set a smaller block size; if they contain many videos or music files, a larger block size will yield better performance (especially with RAID 0 systems).

Additionally, Block Size has another function that determines where files will be written. Returning to the example of a Block Size of 64KB, if a file is smaller than 64KB, it will only be written to one hard drive within the RAID system, resulting in no performance improvement. In another case, a 150KB file will be written across 3 drives in segments of 64KB + 64KB + 22KB, enabling the controller to read information from all three drives simultaneously, allowing for considerable speed increases. If you select a Block Size of 128KB, that file will only be written to 2 drives as 128KB + 22KB. In practice, you should choose a Block Size of 128KB for desktop computers unless you have specific needs.

Once the controller has properly recognized the new hard drives, proceed to install the operating system and format the RAID drive. Windows XP is a smart choice.

Installing Windows follows a standard procedure, but you need to prepare a floppy disk and a floppy disk containing the drivers for the RAID controller. As soon as you press the keyboard to begin the installation, pay attention to the text at the bottom of the Windows installation screen to press F6 in time. Then wait a moment, and when prompted, press S to input the RAID driver into the installation.

The remaining steps proceed just like installing on a standard hard drive.

After stabilizing the system, be sure to install additional RAID management utilities to take advantage of extended features and sometimes even performance improvements. Some programs to consider include Intel Application Acceleration RAID Edition or nVIDIA RAID Manager…

General Notes:

If you currently have a drive full of data and wish to set up RAID 0, you must format the drive and redo everything. Therefore, be sure to find a suitable backup solution. If using RAID 1, this is unnecessary.

Typically, with a RAID 0 system, you should have an additional smaller hard drive to store vital information in case of malfunctions, even though such occurrences are rare.

When the computer restarts (especially after an abnormal shutdown), the system may take a considerable amount of time to identify the drives through the RAID controller, and you might hear unusual sounds coming from the hard drive. There is no need to worry, as this is entirely normal because the controller must synchronize the operation of all the drives in the RAID group it manages.

RAID disk groups often consist of several hard drives operating in proximity to each other, generating significant heat, which is not beneficial in the long run. If conditions allow, seek cooling solutions to prevent unexpected issues.

SUMMARY

The value that RAID brings to a system is undeniable – safety and higher performance depending on the configuration. In fact, RAID 0 and RAID 0+1 are the most favored in home environments. RAID 0 is the fastest but also the most dangerous; a single failure can lead to complete data loss. Meanwhile, RAID 1, while offering the highest data security, often gives users a sense of waste (spending money on two hard drives while only getting the performance and capacity of one). RAID 5 provides both high performance and safety, but the controller equipment is often quite expensive, not to mention the higher cost of hard drives, which makes it less appealing unless specifically needed for work. For this reason, some users turn to SCSI drives to address performance and data security issues; however, the cost for a good SCSI system can be even higher.

On closer inspection, we can see that IDE standards have many issues, such as drives not being designed for continuous operation (which is crucial for server systems), and the current ATA cables are too bulky, which can lead to a cramped case when using multiple drives. In the worst-case scenario, the heat generated can lead to system malfunctions. However, with continuously advancing technology and newer standards like SATA being introduced, RAID is sure to have a bright future and become an ideal companion for high-end personal computer systems.