|
|
There are basically three different storage technologies you can use: Storage Area Networks (SANs), network-attached storage (NAS), and direct-attached storage (DAS). They each have their advantages and disadvantages, naturally; but by combining them, you can perhaps overcome their individual shortcomings. We will look at these in more detail in the following sections.
A SAN is a specialized network, the sole purpose of which is to provide access to high performance and highly available storage subsystems. This solution is quite interesting. The SAN is constructed of several devices, all interconnected by fiber or copper wiring. The subsystem is available to multiple applications at the same time, just like a network-attached storage appliance. The major difference is that it provides higher scalability and performance. Even though a DAS is still faster than a SAN, this performance gap is diminishing constantly with the evolution of SAN technology. (For more on DAS, see the section "Direct-Attached Storage.")
The SAN concept is quite simple. An external RAID (short for redundant array of independent disks, sometimes called redundant array of inexpensive disks) cabinet has a connection directly from the Host Bus Adapter (HBA) to the external RAID subsystem. The SAN, on the other hand, connects the HBA to a switch instead, providing other servers access to the data storage that way (see Figure 8-1).
Next we will take a closer look at NAS and DAS, after which we will introduce you to the logical designs of storage solutions.
A network-attached storage system is a file server, to put it simply. This solution is built on an operating system, often designed for the sole purpose of providing file services. It thereby provides a flexible and scalable solution for most enterprises. Many applications can access a NAS appliance over the LAN by using protocols like TCP/IP. Since this solution is accessed across a LAN, performance might be slow. When compared to a DAS system, performance is lower, but with the help of Gigabit Ethernet you can decrease this performance hit.
 |
Microsoft has developed a dedicated file server based on Windows Server 2003. This edition of the Windows Server family is called Windows Storage Server 2003 and is only available through OEMs. The Storage Server is enhanced to offer a scalable, available, quickly recoverable solution for enterprise file servers.
Besides being a file server, the Windows Storage Server 2003 also acts as a gateway to SANs, so you are not limited to the features offered in the server itself. To learn more about this server, direct your Web browser to http://www.microsoft.com/storage.
|
As the name implies, direct-attached storage refers to a storage device directly attached to a server, specifically local hard drives and RAID systems attached with an IDE or SCSI interface to a computer.
This solution is fast and allows great performance on the server it is attached to, and often only the attached server can access it. This could be a great solution for a small SQL Server perhaps, but if you are planning a larger enterprise application with clustering, this is not a good choice, because many servers might need to access it to provide redundancy.
We feel that it is better for an enterprise to develop a storage architecture that all projects comply with, rather than letting every project develop its own strategy. With a well-thought-out plan for storage, teams working on new projects can spend time on designing and building applications without bothering with planning how to store data. Furthermore, the more you centralize your storage, the more you can simplify management and decrease complexity.
One of the buzzwords these days is consolidation. Everything should be consolidated: databases, servers, applications, and so on. This is not a bad idea if you look behind the hype. Take storage, for example: Many companies have little control over where all data is located, since a lot of smaller databases are distributed around their networks. If you would have a strategy or a policy that dictates all data be located at the same place, you could utilize your storage space better, as you will see later in this chapter in the section "Centralized Model." SANs and NASs are great for accomplishing this.
To illustrate this, we will now show you some examples of how you can develop your logical design for storage purposes.
In a distributed model solution, such as the one shown in Figure 8-2, all servers have their own data stored on a DAS. An example that fits this model is a SQL Server that has its data files and log files on local RAID systems.
Although there are several benefits, such as flexibility, to this design, the disadvantages can outnumber these benefits. If you deploy this kind of solution, all projects need to have expertise in data storage applied during the design phase. The same storage decisions have to be made again and again, and this process is costly.
Because data will be scattered in data islands all over the network, manageability will suffer. As you saw earlier, this situation can also lead to underutilization of the total storage capacity in the enterprise. This will inevitably lead to greater TCO than is necessary, making it an expensive solution in the end. Although this might seem like a cost-effective path to follow for a single project, the total cost for your company will be greater as time goes by.
The opposite of a distributed solution is a completely centralized one, where all data is located on the SAN (see Figure 8-3). In this model, the operating system and application files are the only things placed on local hard disks. By using a powerful SAN, you can supply a solution that can fulfill the business requirements of an enterprise.
The advantages of this solution are numerous. Since you keep all data in a central place you can also manage it easier. This lowers the TCO and makes it less expensive in the long run. You must be aware that the startup costs are high when implementing this design, however, but over time it is worth it. This is why it is important for an enterprise to have a clear view of how data should be stored from the start. If no guidelines exist, many projects will continue to invent their own solution every time. In the end, this will be more expensive than if a clear policy exists, even if the chosen policy might have a high initial implementation cost.
Another benefit is that you can provide high security for the enterprise data, since you have it all in one place. This security does not come without manual labor, however, but is easier to set up with a centralized solution.
A disadvantage with this solution, besides initial costs, is that you cannot distribute your data store to separate geographic locations. So if you are unlucky, you can lose a lot of data if the geographic placement of the SAN is destroyed. Without a doubt, this is sure to change in the near future; in the meantime you do have opportunities available to back up data to a separate geographic location, so you can overcome this problem.
The last example is a middle-way data storage design. Here you use various technologies to store your data. Some of it is stored on DAS, some on SAN, and the rest on NAS appliances. This solution is very flexible, and offers several possibilities for your projects. You can let team leads on each project decide what best serves their purposes.
This could be a more complex design to implement, however. Many design considerations have to be made, just as with the distributed model, making the process long and complicated before it can be implemented. Without proper guidelines, it will also be costly. Since this solution involves a lot of storage technologies from different vendors, TCO risks being higher than a centralized solution.
|
|
