The Utilization Graph provides a graphical representation of the data in MobiLink Monitor.
This tab has the following components:
Utilization statistics list Select and clear the properties listed to specify which properties appear in the utilization graph.
Show Select this option to show the property in the utilization graph. Properties without a checkmark do not appear.
Property Shows the available properties that you can choose to appear in the utilization graph. Only properties with a checkmark appear in the graph.
The list of properties comes from the MobiLink server or the .mlm file, and is saved in .mlmonitorsettings. If you delete .mlmonitorsettings, the properties have placeholder values (19 through 24).
The possible properties are:
TCP/IP work queue This queue represents work done by the low-level network layer in the MobiLink server. This layer is responsible for both reading and writing packets from and to the network. The queue is full of read and write requests. It grows when it gets notified of incoming data to read off the network and/or when it is told by the Stream worker to write to the network.
If this queue gets backed up, it is usually due to a backlog of either reads or writes—sometimes both, but usually one or the other. Reads can get backed up if the MobiLink server is using a lot of RAM and memory pages are being swapped in and out a lot. Consider getting more RAM in this case. Writes can get backed up if the network connection between the clients and server is slow. If this queue is the only queue that is backed up, look at the CPU usage. If CPU usage is high, there may be read/memory problems. If CPU usage is low, you may have slow writes. Consider using a faster network.
Stream work queue This queue represents work done by the high-level network layer in the MobiLink server. This layer is responsible for higher-level network protocol work such as HTTP, encryption, and compression. This queue grows when a lot of reads come in from the TCP/IP layer and/or when a lot of write requests come in from the Command processor layer. If this queue is the only queue that is backed up, consider removing some network protocols, such as HTTP or compression. If this isn't possible, consider reducing the number of concurrent synchronizations allowed using the -sm option. See -sm mlsrv12 option.
Heartbeat work queue This queue represents the layer in the MobiLink server that is responsible for sending pulsed events within the server. This layer is responsible, for example, for triggering the one-per-second pulses of samples to the connected MobiLink Monitors.
It is highly unlikely that this queue gets backed up so it is not visible by default.
Command processor work queue This represents work done by the MobiLink server both to interpret internal MobiLink protocol commands and to apply these commands to the consolidated database. This queue grows when a lot of requests come in. Request types include synchronization requests, MobiLink Listener requests, MobiLink File Transfer utility (mlfiletransfer) requests, and so on. The queue also grows when the consolidated database is busy working on synchronizations, but more synchronization requests keep coming in.
If this queue is the only queue that is backed up, look at the CPU usage. If CPU usage is high, the volume of requests may be too high. Consider reducing the number of concurrent synchronizations allowed, using the -sm option. If CPU usage is low, consider improving the performance of the consolidated database. See -sm mlsrv12 option.
Busy DB workers This value indicates how hard the MobiLink server is pushing the consolidated database. Each unit in the value represents a database worker thread that is doing something in the database. There is no distinction between inserts, updates, deletes, or selects. When this value is zero, the server is not operating on the consolidated database.
When this count is high (when it is close to the maximum set with the mlsrv12 -w option), the MobiLink server is pushing the consolidated database as hard as it can. In this case, if your throughput is satisfactory, there is nothing to do. If your throughput is not satisfactory, consider increasing the number of database worker threads via the -w option. Note that a higher -w value leads to greater contention between connections. This is particularly bad when all connections are performing uploads, so you may need to use the mlsrv12 -wu option to set a lower limit for database workers doing uploads. If you cannot seem to find settings for -w and -wu that provide adequate throughput, examine your synchronization scripts for possible contention issues. Finally, you can consult your RDBMS documentation for ways to improve the overall performance of your consolidated database. See -w mlsrv12 option and -wu mlsrv12 option.
OE Work Queue This represents the size of the Outbound Enabler (OE) queue. The queue size is proportional to the network I/O activity between the MobiLink server being monitored and the Relay Server(s).
Notifier Work Queue This represents the size of the Notifier work queue. The queue size increases if the consolidated database is slow to respond with the results of the Notifier's queries.
The Notifier's queries for Remote Tasks are very simple, but may return many rows in large deployments. The queue size is proportional to the length of time it takes the Notifier to retrieve the Remote Task notifications.
Dynamic Cache Work Queue This queue represents the length of time it takes to perform dynamic cache sizing. This line is hidden by default.
Color For each property, click Color and select a color from the dropdown list that is used to represent the selected property in the utilization graph.
Scale Double-click the Scale field and type a value to change the scale used for the graph.
The vertical scale on the Utilization Graph always goes from 0 to 100. This represents zero to one hundred percent of the scale. Each value has its own scale. By default, all scales are 5, meaning that values are expected to be in a range of 0 to 20, scaled (by 5) to the range 0 to 100. If a value becomes greater than 20, the scale is automatically adjusted such that the largest value is at 100.
To determine the maximum value in the display, divide the scale into 100. For example, if the TCP/IP work queue scale is 2.381, then the maximum value is ( 100 / 2.381 ) = 42. The actual maximum isn't usually important. What is important is that values towards the top of the graph are approaching the largest currently-known value for the given property—in other words, the peak load for that property as observed in the current monitoring session.
When the graphs are consistently towards the top of the display and you notice that synchronization throughput is down, you may have a performance problem that needs investigation. Similarly, if one or more values creeps upward over time without diminishing, then there is likely a performance problem. Note that the graphs may often be towards the top of the display with MobiLink server performing well. This indicates that MobiLink server is busy and successfully processing the job.
Selection border Select a color from the dropdown list for the border of the utilization graph. By default, the border is Black.
Grid lines Select a color from the dropdown list for the grid lines. By default, the grid lines are Light Gray.
Graph background Select a color from the dropdown list for the graph background. By default, the graph background is White.
Use antialiasing to smooth graph Antialiasing is a technique used to make graphics look smoother. Selecting this option makes the graph look better, but it may be slower to draw.
See also |
Discuss this page in DocCommentXchange.
|
Copyright © 2010, iAnywhere Solutions, Inc. - SQL Anywhere 12.0.0 |