| ||||||
Search Microsoft.com for: |
MS Site Server 3.0 Direct MailCapacity and Performance Analysis April 1999 Microsoft Corporation On This Page
Definition of Terms
Chapter 1OverviewThis document evaluates the performance and scalability characteristics of Microsoft® Site Server version 3.0 Direct Mail. Also demonstrated are procedures for identifying these characteristics. This document can be used to assess the value of adding resources, and to identify which resources would satisfy greater capacity needs. Analyzing Direct Mail System Configuration In the test scenarios, five computers were used in the following configurations: Direct Mail Processor Scaling
Test Profiles
Direct Mail DescriptionIntroduction to Direct Mailer Direct Mail sends electronic mail to groups of users. In addition to sending bulk mail to existing customers, you can use Direct Mail to send bulk e-mail to Web site users. You can also personalize messages according to these factors:
The mail process can be automated. For example, every Monday morning, Direct Mail can check which users have browsed the site for your new product in the past week. E-mail can then be sent to those users. Note that users who have never seen the site before will receive an introductory message; users who have been to the site more than once will receive more detailed information about the product. This task can be scheduled for low activity periods, such as before 4:00 A.M., so that your computers are free for your normal business activities. Direct Mail utilizes several components:
Interaction and process of Direct Mail The following diagram shows the Direct Mail service broken down by the physical servers that house the interacting components. The process of sending mail operates in three distinct phases: Distribution List Resolution. The Message Builder Service steps through the list and performs an LDAP search query to retrieve the e-mail name from the Membership Directory. Mail Address Sort. The Message Builder Service sorts the list to remove duplicates. Send and Receive: Non-personalized static HTML. The HTML document is retrieved from the Web server and sent to the SMTP server for each e-mail recipient. Non-personalized server side scripting (ASP). The ASP script is executed from the Web server, personalized for the first user in the distribution list, and then sent to the SMTP server for each e-mail recipient. Personalized static HTML. For each recipient, the HTML document is retrieved from the Web server and then sent to the SMTP server. Personalized server side scripting (ASP). The ASP script is executed and personalized for each e-mail recipient and then sent to the SMTP server. Summary of Scalability and PerformanceBased on the data collected in this document, the following assertions can be made about scaling and performance for Direct Mail:
Chapter 2Detail Discussion of Scalability and PerformanceTime to Complete a Mailing This chart shows that the time resource cost for adding personalization to a simple HTML is approximately 40 percent greater. It also shows that a simple dynamic script can increase the time to complete a mailing by 20 percent over a static script. The script in this example is in Appendix B. It is an uncomplicated script. More complex scripts will increase maximum throughput and time to completion. Time per mailing by mail type
Maximum transaction rates These tables display the maximum mail send rate when it is in the send and receive mail phase. The rates show the maximum sustainable speed at different list sizes.
Maximum send rate by mail type
Memory Utilization on the Message Builder/List Builder Computer This chart shows the maximum private bytes utilized by the Message Builder and List Builder Services while sending the different transactions. These memory consumption values are only for the mail services and do not include the memory used by Windows NT or any other process running on the server. Memory utilization on Message Builder/List Builder by mail type
SQL Resource Utilization The SQL server maintains the database for the Membership Directory. During the first phase of the mailing operation it handles the requests from the LDAP server for the user objects. This table details the CPU utilization for this system. The disk subsystem on the SQL server is also utilized. The specifications and recommendations for this specific subsystem are documented in a separate paper. The SQL server should be optimized as described in the Site Server 3.0 Membership Directory Configuration and Tuning document, which is also available from Microsoft.
The phase length in this table refers to the active period for the SQL server specific to the mailer operation. According to the description above this would be phase one. However, if the personalized script makes use of the AUO object then another request could be made through the LDAP server to the SQL server. LDAP Server Resource Utilization The LDAP Service handles the requests for name resolution during the first phase of the mailing operation. The Message Builder Service computer opens a single connection to the LDAP server and makes a continuous stream of requests. The table below details the resource utilization for the LDAP server in this phase.
The phase length in this table refers to the active period for the LDAP server specific to the mailer operation. According to the description above this would be phase one. However, if the personalized script makes use of the AUO object, then another request could be made through the LDAP server to the SQL server. WWW Resource Utilization In this table, the second phase of the mail operation is detailed. The primary resource utilization for this server is the CPU. It should be noted that the script that is run will greatly affect the overall outcome of these results. The script that was used in this test is contained in Appendix B.
SMTP Resource Utilization The SMTP server was running an SMTP server emulator1 that did not actually send the message. Refer to the SMTP scaling documents available from Microsoft that address specifications for this component. The transaction rate generated by the mailer directly corresponds to the amount of mail traffic the server should handle. Appendix ACritical Monitoring CountersAll counters noted can be found in the Microsoft® Windows NT® Performance Monitor. These counters will be distributed among the machines in the Personalization and Membership (P&M) service group. The counters in the system and memory objects can be used to monitor capacity. Physical Disk Disk Writes/sec Disk activity should not sustain maximum transaction rate Disk Reads/sec System Object Context switches/sec should be less than 15,000 %Total Processor should be less than 80% Processor Object %Processor Utilization (average) should be less than 80% (for each processor) Memory Object Available Bytes should be greater than 4 MB of RAM Pages/sec should be less than 1 per second Site Server Message Builder Service Messages Sent/sec SMTP Server Messages Delivered/sec Process Object Private Bytes for the inetinfo, TMMLBSvc, TMsgBld processes Appendix BASP Script for AUO<%@LANGUAGE=VBScript%> <% m_BindAsUserName = "cn=Administrator" m_BindAsPassword = "password" %> <!-- #INCLUDE FILE="dmheader.asp" --> <!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML//EN"> <html> <head> <meta name="GENERATOR" content="Microsoft FrontPage 3.0"> <!-- meta http-equiv="Content-Type" content="text/html"; charset="iso-8859-1" --> <meta http-equiv="Content-Type" content="text/html"> <title>Direct Mailer Test - Simple AUO page</title> </head> <body> <% response.write "Hello " & m_sFirstName & ", where do you want to go today?" & chr(13) %> </body> </html> HTML Page for Static Tests <html> <head> <meta name="GENERATOR" content="Microsoft FrontPage 3.0"> <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1"> <title>Direct Mailer Test - Simple HTML Page</title> </head> <body> Hello, World! </body> </html> 1 The program is a tool called SuperSink.exe. |