10.4 Web Server Considerations


10.4 Web Server Considerations

During the past few years , the number of Web servers connected to the Internet has increased at an almost hyper-exponential growth rate. From a few thousand servers in 1995, the number of servers connected to the Internet is estimated to exceed several million as we enter the new millennium . Along with this increase in the employment of Web servers to provide corporations, academic, and government agencies with a presence on the Web is a pair of interrelated problems that result from the use of images. One problem is determining an appropriate WAN connection rate to the Internet from the LAN on which the server resides. A second problem involves examining and, when practical, altering Web page images to reduce the quantity of transmission to improve performance. As a side benefit, this action will decrease the cost of communications if your ISP bill is based on the level of utilization of the transmission facility connecting your organization's Web server to the Internet. In this section we examine both problems as well as methods to resolve those problems.

10.4.1 Determining a WAN Connection Rate

The selection of an appropriate WAN operating rate to connect your organization's Web server to the Internet can be considered to represent a double-edged sword, requiring a high degree of caution. To illustrate the problems associated with the selection of an appropriate WAN operating rate, consider Figure 10.13, which illustrates the manner by which most servers are connected to the Internet. In this example, a router attached to the organizational LAN on which the Web server resides is connected via leased line to a router at the ISP. If the WAN operating rate is too slow, your organization may lose visitors who, after attempting to access your site, simply give up and go elsewhere, or worse yet, forego a purchase if your site provides online sales. At the other extreme, let us assume you selected a WAN operating rate that greatly exceeds actual traffic requirements. In this situation, your organization will more than likely wind up paying for unused and perhaps unnecessary transmission capacity month after month. Thus, a methodology is required that can eliminate a degree of guessing that takes place when many organization attempt to select a Web server WAN connection rate.

click to expand
Figure 10.13: A Typical Web Server “Internet Connection

10.4.1.1 Variables to Consider

There are a number of variables you should consider when selecting a Web server WAN connection rate. Some variables can be easily quantified ; however, as we will shortly note, other variables may require a degree of estimation.

Table 10.6 lists the major variables you should consider as input to a decision process concerning determining an appropriate Web server connection rate to the Internet. By discussing each of the entries in Table 10.6, we can obtain an appreciation for the use of these variables in the WAN operating rate selection process. Thus, let us examine each of the variables listed in Table 10.6 and how we might apply information about those variables to the WAN operating rate selection process.

Table 10.6: Variables to Consider When Selecting a Web Server WAN Connection Rate
  • Web server page structure

    • Use of text and graphics

    • Storage requirements per page

  • Web page relationships

    • Estimated frequency of access per visit

  • Information content and audience

  • Expected hits during busy hour

  • Other LAN connectivity requirements

    • Workstations

    • Servers

10.4.1.1.1 Web Page Structure

The Web page structure variable is concerned with the relationship between the use of text and graphics on each page on your server. In examining the use of text and graphics, you should attempt to determine the storage requirements per Web page. In doing so, it is important to note that each object referenced on a stored page, such as a GIF or JPG image, is stored as a separate file. Thus, if a Web page contains nine icons, you must locate the size of nine images and add their cumulative storage to the storage of the Web page holding the text of the page.

10.4.1.1.2 Web Page Relationships

A second variable that warrants consideration is the relationship of Web pages to one another and an estimate of the frequency or probability of access per Web page per site visit. This means you should review how a Web surfer or customer might access your organization's home page and then jump to other pages on your server. In doing so, you would estimate how many times a user accessing the home page would access other pages. For example, if you believe a page would be accessed once by every three persons visiting your site, the page would have a probability of access of 0.33.

10.4.1.1.3 Information Content and Audience

You should examine the information content of your organization's Web server and its audience as this will probably provide you with an indication of its anticipated popularity. In doing so, you may wish to check with your marketing and public relations department to ascertain if there are any promotions or advertising planned that could affect Web server access and, if so, how access might be affected.

10.4.1.1.4 Determining Expected Hits

To correctly size the WAN transmission facility, you must consider the number of hits to your server during its peak or busy-hour period. If your organization does not already operate a server from which you can extrapolate such information, you may be able to use the information content, audience, and advertising and promotion information to query your ISP on page hits to other servers they support that appear to be similar in scope to your organization's planned server.

10.4.1.1.5 Other LAN Connectivity Requirements

If the Web server is not the only device on the LAN, you should consider the effect of other devices on the WAN operating rate. For example, if there are workstations or other servers on the LAN, their WAN transmission requirements must be considered. Now that we have an appreciation for the scope of the variables we should consider, let us turn our attention to the estimation process.

10.4.1.2 Estimating the WAN Operating Rate

There are six specific steps you should consider when attempting to estimate an appropriate operating rate to connect your organization's server to the Internet. Those steps are summarized in Table 10.7, where the sixth step, while not necessary for determining the WAN operating rate, is important to perform as it will indicate if the operating rate can be used to its full extent. To illustrate the use of the steps presented in Table 10.7, let us assume your organizational server will have a home page and three "order" pages, with each order page under the home page. Let us further assume that the home page contains 100 Kbytes of images and text, while each order page contains 25 K-bytes of text and images. Thus, these assumptions would result in completing the first step listed in Table 10.7.

Table 10.7: Steps in the WAN Operating Rate Estimation Process
  1. Determine Web page storage, to include graphics and text.

  2. Estimate probability of access per Web page.

  3. Estimate average Web page storage based on items 1 and 2.

  4. Estimate number of hits during the busy hour.

  5. Multiply expected hits during busy hour by average storage for typical page. Add 15 percent for communications overhead and divide by 3600 seconds per hour to compute operating rate in bps.

  6. Examine the LAN to determine if the server can use the LAN to keep up with the rate determined in step 5.

The second step in the WAN operating rate estimation process involves assigning a probability of access per Web page. Let us assume that for every hit on the home page, an order page will receive 0.3 hits. While 0.3 times the 3 order pages only equals 0.9, in this situation we are assuming that 10 percent of traffic to the home page will result in a browser operating decision that they either arrived by mistake, decided the product is not what they want after reading the home page, or decide upon another reason not to proceed further.

Under step 3 we want to estimate the average Web page storage based upon their retrieval. Thus, using the preceding assumptions, we obtain:

Next, we need to estimate the number of hits expected during the busy hour. As previously mentioned, we can contact our ISP to obtain an estimate based upon the type of site content we plan to place on the server, contact our organization's marketing and PR departments, or extrapolate figures from usage on a different server. Let us assume that we determined that during the busy-hour we can expect 3000 hits on the server.

Moving on to step 5, we would first multiply 64.47 KB by 3000 hits per busy-hour and then add 15 percent for protocol overhead. Next, we would multiply the result by 8 bits/byte and divide the number by 3600 seconds in an hour to determine the data rate in bits per second (bps). Thus, we would compute the following:

Based on the preceding computation, we would be inclined to order a 512-Kbps fractional T1 line to connect our server to the Internet. However, prior to doing so we should perform step 6, which involves examining the LAN to determine if it can support a data transfer of 506,133 bps from the server to the router for transmission to the Internet. To accomplish step 6, you would place a protocol analyzer on the existing LAN to determine the average frame length and frame rate on the network prior to providing Internet access. Suppose the existing frame rate is 620 frames per second and the average frame length is 120 bytes. Then, 620 frames /second * 120 bytes/frame * 8 bits/byte results in a data rate of 595,200 bps on the LAN. Obviously, an additional 506,133-bps transmission flowing from the server to the router via the LAN will not bring the LAN to saturation and the network can support the added transmission to the router.

10.4.2 Images and Network Charges

In concluding this chapter we turn our attention to the effect of images on the cost of an Internet connection. Let us assume that your Internet service provider bills for a corporate Internet connection based upon the operating rate of the leased line and its level of utilization, a billing technique many ISPs are beginning to employ . Let us further assume that the monthly charge of your ISP for a T1 line is $750 per month plus $0.01 per Mbyte of data transmitted onto the Internet.

To illustrate the potential effect of Web page images on network charges, let us assume that your organization's home page contains a picture in JPEG format of your firm's building, your CEO, or another object that results in 100,000 bytes being downloaded each time a person accesses your home page. If you import the picture into an image manipulation program and change its quality value slightly lower, you more than likely can reduce the size of the file by at least 20 percent, or 20,000 bytes, without adversely affecting its visual appeal . If your organization's Web site home page is accessed 15,000 times per week, this one simple action would reduce the quantity of data transmitted to the Internet by 20,000 bytes * 15,000, or 300 Mbytes per week. Over a four-week period, this results in the elimination of 1.2 Gbytes of traffic. At a cost of $0.10 per Mbyte, you just saved your organization $12 per month, or $144 on an annual basis for a few minutes of effort. If you continue your examination of other pages on your Web server and perform similar operations, it becomes possible to save several thousand dollars or more for a lightly utilized server, and possibly tens of thousands of dollars per year for a heavily utilized server by adjusting the quality scale of JPEG images, converting images into better formats with respect to their data storage requirements, or simply cropping images. Thus, images matter, especially when you have to pay for their transmission.




Enhancing LAN Performance
Enhancing LAN Performance
ISBN: 0849319420
EAN: 2147483647
Year: 2003
Pages: 111
Authors: Gilbert Held

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net