10.1 Appliance or Software Solution

only for RuBoard - do not distribute or recompile

10.1 Appliance or Software Solution

For the most part, caching products can be classified into appliances and software solutions. Choosing an appliance means that the hardware and software come bundled together as a single product. With a software solution, on the other hand, you buy the hardware from a different company or perhaps use some equipment you already have. Appliances are usually designed to run just one application ”the web cache. Both types have their advantages and disadvantages.

10.1.1 Appliances

Many people are attracted to appliances because they tend to be easier to install and maintain. The software comes preloaded on the hardware. You don't have to think about disk space, file permissions, startup scripts, etc. Often, the product is up and running after you provide a few basic configuration details, such as the box's IP address, hostname, default route, and DNS server.

Appliances usually have a nice graphical interface for administration. That is, you simply use your web browser to configure and administer the cache. The interface probably also includes real-time graphs that show you current load, disk usage, and other useful metrics.

Another benefit of an appliance is that you don't have to worry about hardware compatibility or drivers. You can be sure that all components of the system work together. When buying your own system, you might end up with the wrong type of memory, for example, or the operating system may not support your particular network card.

Often, an appliance's operating system is optimized and customized for its particular hardware or task. For example, the product may use a specialized filesystem specifically designed for 9GB SCSI LVD disks. Another likely optimization is to have a tighter integration between the network and the application. Copying data buffers between layers is a potential source of delay. Eliminating those copies improves the product's performance.

Until recently, it was rare to find an appliance that used ssh for remote access. Telnet was the only option for a long time. Now it seems like most of the appliance products have adopted ssh as well. Even so, a general-purpose operating system may give you more security features than an appliance. With ssh on Unix, for example, you can require or deny certain types of encryption, key formats, and so on. Additionally, most Unix systems now have extensive packet filtering capabilities not likely to be found on appliances. On the plus side, appliance vendors should be very responsive to known security problems. You may not be able to receive software updates without a support contract, however.

In general, an appliance is more difficult to customize. You won't have access to the operating system or be able to run additional services on the cache machine.

You might have a difficult time upgrading an appliance. If you need more disk space, for instance, any number of problems might surface. Higher capacity disks may not be available for your particular model. Some appliances are designed to be very small and may not have open slots for additional network or disk controllers. If you need more processing power, you may not be able to upgrade just the CPU. Also note that your warranty or license may not allow you to upgrade the hardware yourself, even if you know how. These issues are perhaps less important for small, cheap appliances because you can simply add more units to expand your capacity.

The following companies offer appliances: Network Appliance, InfoLibria, Cisco, CacheFlow, Lucent, iMimic, IBM, Dell, Compaq, and Inktomi. A number of companies sell appliance products based on Squid. For a current listing, visit http://www.squid-cache.org/products.html.

10.1.2 Software

If you decide on a software solution, then you must buy the necessary hardware on your own. If you're lucky, you may already have some hardware you can use. When buying your own hardware, there is a risk that some of the components are not well matched. For example, you might get too much disk space and not enough memory, or you may spend too much money on an overly powerful CPU. There is also some risk of incompatibility . Your operating system might not support Gigabit Ethernet cards, or you might get the wrong type of memory.

The software solution allows you to reuse your hardware for other purposes. If you invest a lot of money in appliances and later decide you don't like them, you may be stuck with unusable equipment. Your general-purpose hardware retains some value even if you don't use it for caching. Also, it might make sense for you to run additional services, such as DNS and email, on your caching hardware. This reduces the number of boxes you have to administer.

The software approach allows you to choose hardware and operating systems that you are comfortable with. Your company may have invested a lot of time and expertise in a specific platform, such as Digital Unix or Solaris. Furthermore, you may have a good business relationship with certain vendors. If your company is large enough, then you probably get good discounts on hardware and support contracts.

A software solution allows you to customize many aspects of the system. For example, you can write your own scripts to analyze and archive log files. For additional security, you can utilize packet filtering features of the operating system or use third-party applications.

Open source products such as Squid offer another level of customization. If you are so inclined, you can add your own features, change some hardcoded values, and even fix some bugs .

On the other hand, a general-purpose operating system is probably not tuned and optimized for proxy caching. Historically, proxy caches that run on Windows or Unix have used the native filesystems (i.e., NTFS, UFS). Soon we will see more software solutions that have their own filesystems. Even so, general-purpose operating systems have a lot of code, including drivers, that could be optimized or simply removed for a system dedicated to proxy caching.

A software solution may be easier to debug because you have full access to the operating system. You can use native tools to monitor disk performance, CPU utilization, memory usage, etc. For example, you might find it useful to run tcpdump when debugging network problems.

You are likely to spend more time administering a software solution than an appliance. With an appliance, the operating system basically takes care of itself. A software product, on the other hand, requires you to create users, monitor disk usage, write cron jobs, etc. You also need to stay up-to-date with bug fixes and security patches from the operating system provider.

Finally, software products may be easier to expand and upgrade, especially if you use inexpensive PC hardware. You should have few problems upgrading to a faster CPU, getting more memory, and adding additional disk drives .

Squid is available as software (source code , although a number of companies sell Squid-based appliances. The following companies sell software products: Netscape, Microsoft, Novell (Border Manager), and Inktomi.

only for RuBoard - do not distribute or recompile