MIIS 2003 Capacity Planning Test Summary - Network
Applies To: Windows Server 2003 with SP1
Previous Sections in This Guide
The following section discusses the network requirements for MIIS 2003 based on data collected during various tests presented throughout this document.
Test Description
Network traffic during staging and synchronization operations was monitored between an MIIS 2003 server and a remote SQL Server hosting the MIIS 2003 database. Network traffic during staging operations also monitored between an MIIS 2003 server and a connected data source.
Expected Results
Many believe that increasing available bandwidth increases performance. In an MIIS 2003 environment, a significant number of factors affect performance. Testing results indicated that other components of the MIIS environment are strained much more than the network bandwidth. As such, the requirements on the network should not be too intensive. One possible exception is the situation where the SQL server is remote to the MIIS server, which might be a greater load on the network.
Test Results Summary
Test results show that MIIS 2003 staging operations have a low impact on available bandwidth. In the tests presented here, the impact was consistently less than 1.5% of the available bandwidth. Even the use of a remote SQL server database had low load on the network.
The tests presented here are based on sequential processing of each management agent. If run profiles are executed concurrently in your environment, it is recommended that you test the impact of multiple sources passing information across the network.
The results confirm that the network requirements for MIIS 2003 are not severe. During the testing, however, it did become apparent that reliability of the network connection played a much more important role than speed.
Result highlights:
A 100Mpbs network connection should be more than sufficient
The reliability of the connection is a very important factor for a stable and efficient MIIS 2003 environment.
During sequential run profile execution, the network requirements never exceeded 1.5 Mbps.
Recommended Best Practices
Use at least a 100 Mbps local network connection between the MIIS 2003 server and its connected data sources. If you use a remote SQL Server to host the MIIS 2003 database, use at least a 100 Mbps local network connection between the MIIS 2003 server and the SQL Server. The network layer will not be a bottleneck if this requirement is met.
If the data source is remote to the physical location and 100 Mbps service is not available, use a reliable link that is no slower than 5 Mbps. Operating MIIS 2003 across a slow link (less than 5 Mbps) is not recommended. However, if your environment makes it necessary to operate MIIS 2003 across a slow link, reliability is the primary concern.
Note
If the link is unreliable, resolve the unreliability issues before you deploy MIIS 2003. Although the time-outs on the various management agents can be increased, this should not be a workaround for unreliable links. An unreliable link can significantly impact the performance and operational efficiency of MIIS 2003.
Test Scenario
The data used for this section was gathered during the SQL Server and processor tests presented in the "MIIS 2003 Capacity Planning Test Summary - SQL Server" and "MIIS 2003 Capacity Planning Test Summary - Processor" sections of this document. Those specific tests were chosen because of the increased network load created by remotely locating the SQL Server.
Two sets of the SQL Server tests involved staging and synchronizing a large number of objects where the MIIS 2003 server was located on a different server than the SQL database used by MIIS 2003. The first set of network test results summarize the traffic generated when staging and synchronization data was sent between the MIIS 2003 server and the SQL Server hosting the MIIS database during those tests. The results of these network tests demonstrate the impact on network traffic if you decide to remotely locate the SQL Server that hosts the MIIS 2003 database.
The second set of network test results summarize the traffic generated between an MIIS 2003 server and a connected data source during staging operations. This data was gathered during the processor tests presented earlier in this document.
Server Hardware Configuration
The server configuration used to gather this data is the same configuration used for the SQL Server tests presented in the "MIIS 2003 Capacity Planning Test Summary - SQL Server" and "MIIS 2003 Capacity Planning Test Summary - Processor" sections of this document. For the first test, this data was collected from the two split server scenarios using the two test platforms designated as DMQS and QMDS. For a detailed description of the hardware configuration of those two platforms and a complete description of how MIIS 2003 was configured on those platforms, see the "Server Hardware Configuration" information in the section "MIIS 2003 Capacity Planning Test Summary - SQL Server" in this document.
Data for the second test was collected during the processor tests by monitoring traffic between the MIIS 2003 server and the SQL server that acted as a connected data source. For a detailed description of the hardware configuration of those two platforms and a complete description of how MIIS 2003 was configured on those platforms, see the "Server Hardware Configuration" information in the section "MIIS 2003 Capacity Planning Test Summary - Processor" in this document.
Network Configuration
During this test, the network between the two servers was a 100-Mbps VLAN Switched connection. Monitoring of performance was conducted on a separate monitoring network interface.
Test Results
The first test presents results of monitoring staging and synchronization traffic between the MIIS 2003 server and a remote SQL Server that hosts the MIIS 2003 database. The second test monitors staging traffic between an MIIS 2003 server and a remote SQL Server that acts as a connected data source.
MIIS Server and Remote MIIS Database
.gif "Chart: % Utilization of 100Mbps net connection")
Table 30: Source data for Figure 18
| Operation | % of 100Mbps |
|---|---|
TXT01 Import |
1.23 |
TXT02 Import |
1.23 |
TXT01 Sync |
1.30 |
TXT02 Sync |
0.87 |
Observations
During this test, 100,000 objects were staged and synchronized to a remote SQL database server.
Even with a remote SQL server, the network load does not exceed 1.3Mbps.
Note the decline in network traffic during the second synchronization run profile (TXT02 Sync). This is because more time is now spent doing index queries to attempt to join objects together as opposed to the large amount of T-SQL INSERT’s that happened in the preceding operations.
MIIS Server and a Remote Connected Data Source
.gif "Chart: Traffic to and from remote SQL Server")
Table 31: Breakdown of inbound and outbound traffic from the remote SQL Server. (Measured in Bytes/ Sec)
| 100,000 Objects | Total SQL Traffic | Sent from SQL Server | Received by SQL Server |
|---|---|---|---|
TXT01 Import |
1,231,861 |
463,843 |
768,019 |
TXT02 Import |
1,229,771 |
462,704 |
767,068 |
TXT01 Sync |
1,302,425 |
443,643 |
858,937 |
TXT02 Sync |
868,961 |
588,059 |
280,902 |
Observations
Note that the network load during staging is quite similar, but more traffic is received by MIIS 2003 from the SQL Server when synchronization operations begin. This is due to the fact that following the INSERT statements executed during the staging, the SQL server has to return information about objects in the metaverse that can possibly be joined to objects being synchronized. The rules for join and projection are evaluated in the MIIS 2003 server and thus these records concerning these operations have to be passed from the SQL server to the MIIS 2003 server.
During the second synchronization, this becomes very apparent as the amount of data sent from the SQL server to MIIS 2003 increases by almost 100Kb/sec.