Advanced Network Monitoring Using Oracle Solaris 11 Tools

How to use Oracle Solaris 11 networking tools to generate, observe, and analyze network statistics.

Published September 2013

In this article, we will examine three use cases in order to demonstrate the capability of Oracle Solaris 11 tools. The first use case will show how to measure the network bandwidth between two systems. The second use case will demonstrate how to observe network statistics on a specific TCP port, and the last use case will demonstrate, by example, how to collect and display historical statistics of network resource usage for a given date and time range.

Want to comment on this article? Post the link on Facebook's OTN Garage page. Have a similar article to share? Bring it up on Facebook orTwitter and let's discuss.

The following are benefits of using these tools:

The tools are built into the Oracle Solaris 11 operating system, so there is no need to install or buy third-party applications or use special hardware equipment in order to run the network measurements.
The tools have low performance overhead (about 5 percent) during the network statistics collection, so it is safe to use them on a production system, and they have a minimum performance impact on the network I/O workload. In addition, we can enable or disable the tools without the need to reboot Oracle Solaris 11.
It is possible to monitor physical or virtual network elements—such as virtual network interface cards (VNICs) or virtual operating systems (for example, Oracle Solaris Zones) in a multitenant cloud environment—and very accurate observations are obtainable based on the specific virtual element being monitored.
It is feasible to collect historical network statistics in order to generate a baseline for the network activity. This baseline can be used to compare current network activity to previous activity and to help with capacity planning for future growth. In addition, you can generate statistical reports based on a given date and time range and visualize your network activity.

In the example presented in this article, all the building blocks will be installed using Oracle Solaris Zones, ZFS, and network virtualization technologies. Figure 1 shows the architecture:

Figure 1

Measuring the Network Bandwidth Between two Environments

In the first use case, we will measure the network bandwidth between two environments.

Note: You can use these tools and procedures between physical systems or virtual systems (for example, Oracle Solaris Zones).

When doing performance benchmarks, it is desirable to measure network bandwidth, which in turn helps to measure network scalability. In addition, testing the network capability helps eliminate the network as a limiting factor in a given environment.

A useful tool for this kind of task is the iperf(1) tool, which provides the ability to the measure the performance of a network, and it can observe TCP or UDP throughput and provide real-time statistics. We will use this tool in order to measure the maximum network bandwidth between two Oracle Solaris Zones.

Oracle Solaris 11 network virtualization allows us to build any physical network topology inside the Oracle Solaris operating system including VNICs, virtual switches (vSwitches), and more-sophisticated network components (for example, load balancers, routers, andfirewalls), as shown in Figure 2.

The benefits for using this technology include reducing infrastructure cost, since there is no need to invest in superfluous network equipment. In addition, the infrastructure deployment is much faster, since all the network building blocks are based on software, not on hardware.

Figure 2

In this use case, we will leverage the ability to create VNICs on top of the physical network card.

Creating the VNICs

First, let's create the VNICs (in a later step, we will associate the VNICs with Oracle Solaris Zones):

root@global_zone:~# dladm create-vnic vnic0 -l net0
root@global_zone:~# dladm create-vnic vnic1 -l net0

Verify the VNIC creation using the dladm show-vnic command:

root@global_zone:~# dladm show-vnic
LINK                OVER         SPEED  MACADDRESS        MACADDRTYPE       VID
vnic0               net0         1000   2:8:20:c0:c:d0    random            0
vnic1               net0         1000   2:8:20:7c:9:c6    random            0

We can see the two VNICs, vnic0 and vnic1, have been created, and we can see their MAC address and the physical link that they are associated with (net0).

Creating the First Zone (`source_zone`)

Oracle Solaris Zones are built-in virtualization technology available in Oracle Solaris. In this first use case, we will use them to contain our testing environments.

If you don't already have a file system for the source_zone and target_zone zones, run the following command:

root@global_zone:~# zfs create -o mountpoint=/zones rpool/zones

We can use the zonecfg command in order to create our first zone, source_zone. The minimum information required to create a zone is its name and its zonepath. In addition, we will add the VNIC named vnic0, which we created in the previous section.

root@global_zone:~# zonecfg -z source_zone
Use 'create' to begin configuring a new zone.
zonecfg:source_zone> create
create: Using system default template 'SYSdefault'
zonecfg:source_zone> set zonepath=/zones/source_zone
zonecfg:source_zone> add net
zonecfg:source_zone:net> set physical=vnic0
zonecfg:source_zone:net> end
zonecfg:source_zone> verify
zonecfg:source_zone> commit
zonecfg:source_zone> exit

The next step is to install the zone:

root@global_zone:~# zoneadm -z source_zone install

Then, we need to boot the zone:

root@global_zone:~# zoneadm -z source_zone boot

root@global_zone:~# zlogin -C source_zone

Specify the following information during the interactive screens in the System Configuration Tool:

Computer Name: source_zone
Network connection configured: Manually
Wired network connection to be configured during installation: vnic0
IP address: 10.0.0.1
Netmask: 255.255.255.0
DNS Name Service: Do not configure DNS
Alternative Name Service: None
Time Zone Region: Europe
Time Zone Location: Britain (UK)
Time Zone: GB

In addition specify a root password.

When you are finished, you should see the see the zone boot messages. Log in to the zone at the zone console login prompt as root.

source_zone console login: root
Password:

After logging in to the zone, verify the networking configuration using the ipadm command:

root@source_zone:~# ipadm show-addr
ADDROBJ           TYPE     STATE        ADDR
lo0/v4            static   ok           127.0.0.1/8
vnic0/v4          static   ok           10.0.0.1/24
lo0/v6            static   ok           ::1/128
vnic0/v6          addrconf ok           fe80::8:20ff:fec0:cd0/10

As you can see, vnic0 has IP address 10.0.0.1 .

The next step is to install the iperf tool inside source_zone using the command shown in Listing 1:

root@source_zone:~# pkg install iperf

Packages to install:  1
       Create boot environment: No
Create backup boot environment: No

DOWNLOAD                                PKGS         FILES    XFER (MB)   SPEED
Completed                                1/1           6/6      0.0/0.0  942k/s

PHASE                                          ITEMS
Installing new actions                         20/20
Updating package state database                 Done
Updating image state                            Done
Creating fast lookup database                   Done

Listing 1

Exit from the zone console using the "~." escape sequence.

source_zone:~# ~.
[Connection to zone 'source_zone' console closed]

Creating the Second Zone (`target_zone`)

The next step is to set up the target zone from the global zone using the zonecfg command.

During the zone configuration, we will provide the zone name, the zonepath, and the VNIC name (vnic1).

root@global_zone:~# zonecfg -z target_zone
Use 'create' to begin configuring a new zone.
zonecfg:target_zone> create
create: Using system default template 'SYSdefault'
zonecfg:target_zone> set zonepath=/zones/target_zone
zonecfg:target_zone> add net
zonecfg:target_zone:net> set physical=vnic1
zonecfg:target_zone:net> end
zonecfg:target_zone> verify
zonecfg:target_zone> commit
zonecfg:target_zone> exit

The next step is to install the zone:

root@global_zone:~# zoneadm -z target_zone install

Then, boot the zone:

root@global_zone:~# zoneadm -z target_zone boot

Check the status of the zones that we've created:

root@global_zone:~# zoneadm list -cv
  ID NAME             STATUS     PATH                           BRAND    IP
   0 global           running    /                              solaris  shared
   1 source_zone      running    /zones/source_zone             solaris  excl
   2 target_zone      running    /zones/target_zone             solaris  excl

We can see that the status of target_zone and source_zone is running.

root@global_zone:~# zlogin -C target_zone

Specify the following information during the interactive screens in the System Configuration Tool:

Computer Name: target_zone
Network connection configured: Manually
Wired network connection to be configured during installation: vnic1
IP address: 10.0.0.2
Netmask: 255.255.255.0
DNS Name Service: Do not configure DNS
Alternative Name Service: None
Time Zone Region: Europe
Time Zone Location: Britain (UK)
Time Zone: GB

In addition specify a root password.

After finishing the zone configuration, as user root, log in to the zone at the zone console login prompt:

target_zone console login: root
Password:

Verify the networking configuration using the ipadm command:

root@target_zone# ipadm show-addr
ADDROBJ           TYPE     STATE        ADDR
lo0/v4            static   ok           127.0.0.1/8
vnic1/v4          static   ok           10.0.0.2/24
lo0/v6            static   ok           ::1/128
vnic1/v6          addrconf ok           fe80::8:20ff:fe7c:9c6/10

As you can see, vnic1 has IP address 10.0.0.2.

Now, from target_zone, check that you can ping source_zone.

root@target_zone:~# ping 10.0.0.1
10.0.0.1 is alive

Note: In some environments, the Oracle Solaris 11 firewall might block network traffic. If your security policy allows it, you can disable the firewall service using the svcadm disable ipfilter command or by adding firewall rule in order to enable network traffic between the two environments. For more Oracle Solaris firewall examples, see "Securing the Network in Oracle Solaris 11.1."

Install the iperf tool using the following command:

root@target_zone:~# pkg install iperf

Starting the Network Performance Benchmark

In order to perform an iperf measurement, you must establish both a server and a client to generate the network traffic.

Therefore, on the target zone, run the iperf command in server mode using the following command, where

-s specifies server mode.
-l sets the length of the read/write buffer (128K).

root@target_zone:~# iperf -s -l 128k

After running the iperf command, you will see the following message on the terminal:

------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 125 KByte (default)
------------------------------------------------------------

The next step is to run the iperf client on the source zone.

From another terminal window, log in to the source zone:

global_zone# zlogin source_zone

Run the following iperf command in client (that is, loader) mode on the source zone, where

-c specifies the command in client mode.
-l sets the length of the read/write buffer (128K).
-P specifies the number of parallel client threads to run (4).
-i n pauses n seconds (1 second, in this case) between periodic bandwidth reports.
-t specifies the time in seconds to transmit data (180 seconds).

root@source_zone:~# iperf -c 10.0.0.2 -l 128k -P 4 -i 1 -t 180

Note: 10.0.0.2 is the IP address of target_zone.

After running the iperf command, you will start to see runtime statistics:

------------------------------------------------------------
Client connecting to 10.0.0.2, TCP port 5001
TCP window size: 48.0 KByte (default)
------------------------------------------------------------
[  6] local 10.0.0.1 port 54972 connected with 10.0.0.2 port 5001
[  4] local 10.0.0.1 port 64738 connected with 10.0.0.2 port 5001
[  5] local 10.0.0.1 port 60614 connected with 10.0.0.2 port 5001
[  3] local 10.0.0.1 port 57875 connected with 10.0.0.2 port 5001
[ ID] Interval       Transfer     Bandwidth
[  5]  0.0- 1.0 sec    316 MBytes  2.65 Gbits/sec

Displaying the Network Statistics

On target_zone, we use the dlstat(1m) command in order to display the network statistics for the vnic1 network interface. Thedlstat(1m) command reports network usage statistics during runtime about physical or virtual data links (for example, VNICs).

Open another terminal and log in to target_zone.

The following command displays statistics for vnic1 at one-second time intervals:

root@target_zone:~# dlstat -i 1 vnic1
LINK IPKTS RBYTES OPKTS OBYTES
vnic1 144.51M 217.32G 18.69M 1.23G
vnic1 806.08K 1.21G 100.83K 6.66M
vnic1 813.68K 1.22G 101.72K 6.71M
vnic1 846.54K 1.27G 105.89K 6.99M
vnic1 868.58K 1.31G 108.63K 7.17M
^C

Note: In order to stop the dlstat command press Ctrl-c.

As you can see, the dlstat command displays the following information:

The number of inbound packets (IPKTS)
How many bytes have been received (RBYTES)
The number of outbound packets (OPKTS)
How many bytes have been transmitted (OBYTES)

One of the main challenges in a multitenant cloud environment with many Oracle Solaris Zones running various network workloads is the ability to find system resources that are overloaded.

Using the Oracle Solaris 11 monitoring tools, we can pinpoint our network measurement to a specific zone and a specific VNIC in order to find a bottleneck. For example, let's display network statistics for a given zone and its associated network cards.

Before we start our observation, we need to rerun the iperf network workload from source_zone:

root@source_zone:~# iperf -c 10.0.0.2 -l 128k -P 4 -i 1 -t 180

To display network statistics for target_zone and source_zone, run following command from the global zone:

root@global_zone:~# dlstat -i 1 -z target_zone,source_zone
LINK IPKTS RBYTES OPKTS OBYTES
source_zone/vnic0 16.07M 1.06G 4.62M 177.86G
target_zone/vnic1 123.48M 185.71G 16.06M 1.06G
^C

Note: To stop the dlstat command, press Ctrl-c.

As you can see, we were able to monitor both zones and their VNICs at the same time!

We can observe how many bytes each zone and its associated VNIC is transmitting (OBYTES) or receiving (RBYTES).

Another useful command that we can use in order to monitor the network in a multitenant cloud environment is the zonestat(1) command .

For example, we can monitor per zone network statistics in addition to CPU and memory utilization:

root@global_zone:~# zonestat -z source_zone -z target_zone 1
Collecting data for first interval...
Interval: 1, Duration: 0:00:01
SUMMARY                   Cpus/Online: 4/4   PhysMem: 15.9G  VirtMem: 17.9G
                    ---CPU----  --PhysMem-- --VirtMem-- --PhysNet--
               ZONE  USED %PART  USED %USED  USED %USED PBYTE %PUSE
            [total]  3.15 78.9% 1838M 11.2% 3857M 20.9%  484K 0.04%
           [system]  1.86 46.6% 1224M 7.47% 3430M 18.6%     -     -
        source_zone  0.59 14.7% 62.1M 0.37% 51.3M 0.27%  483K 0.04%
        target_zone  0.47 11.9% 62.4M 0.38% 51.5M 0.27%   132 0.00%
^C

Note: To stop the zonestat command, press Ctrl-c.

We can see utility reports for the CPU, memory, and networking for the source_zone and target_zone zones. Each zone's utilization is reported both as a percentage of system resources and of the zone's configured limits.

We can use the -r option in order to provide detailed output for specific resources (for example, networking). In addition, we can use the-x option to provide an extended view with more detailed information, as shown in Listing 2.

root@global_zone:~# zonestat -z source_zone -z target_zone -r network -x 1
Collecting data for first interval...
Interval: 1, Duration: 0:00:01
NETWORK-DEVICE                  SPEED        STATE        TYPE
net0                         1000mbps           up        phys
        ZONE       LINK           TOBYTE  MAXBW %MAXBW PRBYTE %PRBYTE POBYTE %POBYTE
      [total]      net0            926M    -      -    300   0.00%   144K   0.11%
       global      net0            1368    -      -    300   0.00%    888   0.00%
     target_zone   vnic1            463M   -      -      0   0.00%      0   0.00%
     source_zone   vnic0            463M   -      -      0   0.00%   143K   0.11%  
     source_zone  source_zone/net0  120    -      -      0   0.00%      0   0.00% 
     target_zone  target_zone/net0  120    -      -      0   0.00%      0   0.00%

Listing 2

Note: To stop the zonestat command, press Ctrl-c.

The output in Listing 2 shows the following:

The name of a datalink (LINK)
The number of bytes transmitted and received by datalinks or virtual links (TOBYTE)
The maximum bandwidth configured on a datalink (MAXBW)
The sum of all transmitted and received bytes as a percentage of the configured maximum bandwidth (%MAXBW)
The number of received bytes that consumes the physical bandwidth (PRBYTE)
The percentage of available physical bandwidth used to receive PRBYTE (%PRBYE)
The number of transmitted bytes that consumes the physical bandwidth (POBYTE)
The percentage of available physical bandwidth used to transmit POBYTE (%POBYE)

Benchmark Summary

Let's summarize the iperf benchmark results.

Once the network load is finished, iperf will print the benchmark results summary.

On target_zone, we can see the benchmark results summary:

...
...
[ ID] Interval Transfer Bandwidth
[SUM] 0.0-180.0 sec 32.2 GBytes 9.21 Gbits/sec

As you can see from the iperf benchmark results, the maximum network bandwidth between the two Oracle Solaris Zones is 9.21 Gbits/sec.

An interesting side point is that this value demonstrates one of the big benefits of installing applications in Oracle Solaris Zones on a single system (rather than having the applications on separate systems). There is no need for the network traffic to go off-system into the network, meaning you can achieve very high throughput with very low latency compared to systems with, say, 1 Gb physical network connections. For more examples, see "How to Control Your Application's Network Bandwidth."

Note: The network benchmark results can be different; it depends on the server model, the network card, and the network infrastructure equipment.

Monitoring Network Traffic Between Two Systems on a Specific TCP or UDP Network Port

In the second use case, we want to measure the network traffic between two systems (they can be physical or virtual) on a specific TCP or UDP network port.

Although the dlstat(1M) and zonestat(1) commands are useful for bandwidth measurement, they lack the ability to monitor network traffic on a specific network port.

The ability to monitor specific TCP or UDP ports can be useful when you need to analyze performance issues between two systems running the same application, for example web server replication between two data centers.

In this use case, we will measure the network bandwidth between two Oracle Solaris Zones on TCP port 443, as shown in Figure 3; the first step is to set up a flow on the network card that we want to measure.

Flow is a sophisticated quality of service mechanism built into the new Oracle Solaris 11 network virtualization architecture, and it allows us to measure or limit the network bandwidth for a specific network port on a specific network interface.

In addition, flows can be created, modified, and removed in both global and non-global zones.

Figure 3

In the following example, we will set up a flow that is associated with the TCP 443 network port on the vnic1 network interface.

On target_zone, enable the flow:

root@target_zone:~# flowadm add-flow -l vnic1 -a transport=TCP,local_port=443 https-flow

Note: You don't need to reboot the zone in order to enable or disable the flow. This is very useful when you need to debug network performance issues on a production system!

Verify the flow creation:

root@target_zone:~# flowadm show-flow
FLOW LINK IPADDR PROTO LPORT RPORT DSFLD
https-flow vnic1 -- tcp 443 -- --

You can see the https-flow flow for the associated VNIC (vnic1) and the TCP port number (443).

On target_zone, stop the iperf server, which you can do by pressing Ctrl-c.

On target_zone, let's start the iperf server again, where

-s specifies the server mode.
-l sets the length of the read/write buffer (128K).
-p specifies the network port (443).

root@target_zone:~# iperf -s -l 128k -p 443
------------------------------------------------------------
Server listening on TCP port 443
TCP window size: 125 KByte (default)
------------------------------------------------------------

You can see that this time, the server is listening on TCP port number 443.

In order to start the loader, on source_zone, run the iperf command, where

-c specifies the client mode.
-l sets the length of the read/write buffer (128K).
-P specifies the number of parallel client threads to run (4).
-i n pauses n seconds (1 second, in this case) between periodic bandwidth reports.
-t specifies the time in seconds to transmit data (180 seconds).
-p specifies the port number (443).

source_zone:~# iperf -c 10.0.0.2 -l 128k -P 4 -i 1 -t 180 -p 443

On target_zone, we can use the flowstat(1M) command, which reports runtime statistics about user-defined flows. This command can report the receive-side or transmit-side statistics only. In addition, it can display statistics for all flows on the specified link or statistics for the specified flow.

Report the bandwidth on the https-flow flow, which monitors TCP port443:

root@target_zone:~# flowstat -i 1
FLOW IPKTS RBYTES IDROPS OPKTS OBYTES ODROPS
https-flow 24.72M 37.17G 0 3.09M 204.08M 0
https-flow 749.28K 1.13G 0 93.73K 6.19M 0
https-flow 783.68K 1.18G 0 98.03K 6.47M 0
https-flow 668.83K 1.01G 0 83.66K 5.52M 0
https-flow 783.87K 1.18G 0 98.07K 6.47M 0
https-flow 775.34K 1.17G 0 96.98K 6.40M 0
https-flow 777.15K 1.17G 0 97.21K 6.42M 0
^C

Note: To stop the flowstat command, press Ctrl-c.

As you can see, the flowstat command displays the network statistics for TCP port 443. In addition, enabling the flow on the vnic1network interface did not degrade the network performance!

Let's wait a few minutes until the load on source_zone is finished. The iperf tool will stop printing output messages when it's done.

You can observe the network statistics from the global zone. This is useful if you define multiple flows and you want to monitor them from a central location.

Let's define another network flow in source_zone. We will define the flow on vnic0 using the TCP port 443:

root@source_zone # flowadm add-flow -l vnic0 -a transport=TCP,local_port=443 https-flow

From the global zone, verify the flows that we created in the previous steps:

root@global_zone # flowadm show-flow
FLOW        LINK                IPADDR                   PROTO  LPORT   RPORT   DSFLD
source_zone/https-flow vnic0    --                       tcp    443     --      --
target_zone/https-flow vnic1    --                       tcp    443     --      --

Start the iperf benchmark again on source_zone:

root@source_zone:~# iperf -c 10.0.0.2 -l 128k -P 4 -i 1 -t 180 -p 443

From the global zone, start the network flows observation:

root@global_zone:~# flowstat -i 1
FLOW IPKTS RBYTES IDROPS OPKTS OBYTES ODROPS
source_zone/https-flow 0 0 0 0 0 0
target_zone/https-flow 132.33M 198.97G 0 17.31M 1.14G 0
source_zone/https-flow 0 0 0 0 0 0
target_zone/https-flow 848.25K 1.28G 0 109.92K 7.25M 0
source_zone/https-flow 0 0 0 0 0 0
target_zone/https-flow 848.26K 1.28G 0 110.05K 7.26M 0
^C

Note: To stop the flowstat command, press Ctrl-c.

You can see the network statistics for the two flows that we created in the previous steps (source_zone/https-flow andtarget_zone/https-flow). Thus, we can monitor both the zones and their associated flows!

Optional: Once you finish your network measurements, you can remove the flow:

root@target_zone:~# flowadm remove-flow https-flow

Verify the flow removal:

root@target_zone:~# flowadm show-flow

Note: You don't need to reboot the zone in order to remove the flow; this is very useful in production environments.

Collecting and Analyzing Historic Network Statistics

In our final use case, we will use the Oracle Solaris 11 ability to store historic network statistics data. This is convenient if we are running our benchmark for a long period and we want to see network trends. Network usage statistics can be stored by using the Oracle Solaris extended accounting facility. For more information see acctadm(1M).

In addition, if we're suspicious that we are having network performance degeneration, we can compare the current network statistics to the historical network statistics.

The extended accounting facility allows you to collect various operating systems statistics—for example, statistics for tasks, processes, and flows—and store them in order to perform performance analyses for long intervals such as weeks, months, and years.

Note: The Oracle Solaris extended accounting facility can be enabled both in global and non-global zones.

In this use case, we will use the extended accounting facility in order to collect networking statistics.

From the global zone, enable the extended accounting facility:

root@global_zone:~# acctadm -e extended -f /var/log/net.log net

Verify the activation of the extended accounting facility:

root@global_zone:~# acctadm net
Net accounting: active
Net accounting file: /var/log/net.log
Tracked net resources: extended
Untracked net resources: none

From now on, the network usage history statistics are stored in the /var/log/net.log file.

The following example shows historical statistics about network traffic and its use of resources on the vnic0 datalink:

root@global_zone:~# dlstat show-link -h -f /var/log/net.log vnic0
LINK START END RBYTES OBYTES BANDWIDTH
vnic0 10:45:52 10:46:12 1502 0 0 Mbps
vnic0 10:46:12 10:46:32 1918 0 0 Mbps
vnic0 10:46:32 10:46:52 2676218 4668466 1878.091 Mbp
vnic0 10:46:52 10:47:12 1313205 2057617 8282.997 Mbp

If we used this facility to collect network statistics for a long duration, we can show historical network statistics.

To display historical statistics of resource usage by traffic on vnic0 over a given date and time range (for example, from January 1, 2013 until June 24, 2013), run the command shown in Listing 3:

root@global_zone:~# dlstat show-link -h -s 01/01/2013,00:00:00 -e 06/24/2013,10:50:00 -f /var/log/net.log vnic0

LINK START END RBYTES OBYTES BANDWIDTH
vnic0 10:45:52 10:46:12 1502 0 0 Mbps
vnic0 10:46:12 10:46:32 1918 0 0 Mbps
vnic0 10:46:32 10:46:52 2676218 4668466 1878.091 Mbp
vnic0 10:46:52 10:47:12 1313205 2057617 8282.997 Mbp
vnic0 10:47:12 10:47:32 3790824 5786125 2329.613 Mbp
vnic0 10:47:32 10:47:52 860 126 0 Mbps
vnic0 10:47:52 10:48:12 890 0 0 Mbps
vnic0 10:48:12 10:48:32 330 0 0 Mbps
vnic0 10:48:32 10:48:52 212 0 0 Mbps
vnic0 10:48:52 10:49:12 243 0 0 Mbps
vnic0 10:49:12 10:49:32 0 0 0 Mbps
vnic0 10:49:32 10:49:52 274 126 0 Mbps

Listing 3

Visualizing the Network Statistics

We can format the output in gnuplot format, which is a command-driven interactive function plotting tool, and later we can use it in order to generate performance graphs. For more information, see gnuplot(1).

Let's display historical statistics of resource usage by traffic on vnic0 over a given date and time range (for example, from June 24, 2013 until June 25, 2013) using the gnuplot format:

root@global_zone:~# dlstat show-link -h -s 06/24/2013 -e 06/25/2013 -F gnuplot
-f /var/log/net.log vnic0 > /var/tmp/netplot.dat

Now, let's format our output file for better readability (we will convert the bandwidth measurement to the 10-GB range):

root@global_zone:~# cat /var/tmp/netplot.dat | awk '{print $1,$2/1000}' > /var/tmp/netplot_format.dat

Then we will generate our graph using gnuplot.

In order to use the gnuplot tool, you need to install the Oracle Solaris desktop package. The first command below installs the gnuplottool. The command below that installs the Oracle Solaris desktop package.

root@global_zone:~# pkg install gnuplot
root@global_zone:~# pkg install solaris-desktop

Then, we need to prepare the plot file in order to set up the plot properties (later we will load the file into the gnuplot tool).

Edit the plot file using your favorite editor, as shown in Listing 4.

root@global_zone:~# vi /var/tmp/plot_file

set autoscale 
unset log
unset label 
set xtic auto 
set ytic auto 
set title "Network statistics"
set xlabel "Time"
set ylabel "Bandwidth (Mb/sec)"
set xdata time
set timefmt "%H:%M:%S"
plot "/var/tmp/netplot_format.dat" using 1:2 title "Bandwidth (Mb/sec)" with lines

Listing 4

In order to access our system using a GUI desktop, we need to start the VNC server, as shown below.

root@global_zone:~# vncserver :1
You will require a password to access your desktops.

Password:
Verify:

New 'global_zone:1 (root)' desktop is global_zone:1

Note: Remember the password; you will need later it to connect to the VNC server. For more VNC server setup examples, see the Oracle Solaris 11.1 Desktop Administrator's Guide.

You can now connect remotely via VNC viewer/client software (such as the freeware TightVNC viewer).

Note: You should connect to the address that showed in the desktop is: message.

$ vncviewer global_zone:1

After you log in using the VNC viewer/client software, open a terminal window by right-clicking any point in the background of the desktop and selecting Open Terminal in the pop-up menu, as shown in Figure 4.

Figure 4

At the terminal window, run the gnuplot tool:

root@global_zone:~# gnuplot

From the gnuplot prompt, load the plot file, as shown in Figure 5:

gnuplot> load '/var/tmp/plot_file'

Figure 5

Based on your network statistics collection, gnuplot will display a graph, as shown in Figure 6:

Figure 6

You can see that we had a network traffic spike at 11:00, 11:50, and 12:45.

For more Oracle Solaris performance graph examples with gnuplot, see Graphing Solaris Performance Stats with gnuplot.

Optional: You can disable the network extended accounting facility using the following command:

root@global_zone:~# acctadm -D net

Verify that the network extended accounting facility is inactive:

root@global_zone:~# acctadm net
            Net accounting: inactive
       Net accounting file: /var/log/net.log
     Tracked net resources: none
   Untracked net resources: extended

Conclusion

In this article, we saw how we can leverage the built in Oracle Solaris 11 iperf, dlstat, flowadm, and flowstat commands in order to observe and gather network statistics.

About the Author

Orgad Kimchi is a principal software engineer on the ISV Engineering team at Oracle (formerly Sun Microsystems). For 6 years he has specialized in virtualization and cloud computing technologies.

Revision 1.0, 09/09/2013

Revision 1.0, 09/09/2013

UNIX-LINUX SERVER

Buscar este blog