Cloud Computing Lab Performance

Bandwidth Requirements

Presented below are the bandwidth requirements for NetLab and/or a similar setup:

  • Network Connectivity Requirements (NETLAB+ Host Site)
  • 5.0 Mb/s or higher Internet connection, with higher outbound burst capability

What we have setup in our current proposed lab environment are 20 workstations with 1Gb/s NIC cards in each as well as a vCenter server which also has a 1 Gb/s NIC card. The middle man between the two are two 2960 switches that are trunked together with a 1 GE connection. This will work, however this will not allow all of our host machines to operate at their full networking capabilities.

A proposed solution would be to eliminate one of the 2960 switches. From here we would use the 2960 switch that remains and take advantage of it’s two 10G ports. These 10 ports would then connect directly to the two 10G ports on the vCenter server. This is turn will allow for better throughput for the host machines to operate at their full networking capabilities.
 

Stress Testing

Network testing specific

What has been done:

  • Initial setup of PRTG and Kiwi (decided not to use this software since it duplicated abilities found in the others used with no additional benefit) in personal test environment.
  • Additional configuration and testing needed to determine the best way gather network data. It was decided through this testing to use vCenters built in monitoring software and PRTG for the switch data.
  • Initial data seems to be good, hard to tell without the ability to simulate large amounts of network traffic.
     

PRTG Network Monitor

  • PTRG web page
  • Allows up to 10 “sensors” (data collector sets) on the freeware version
  • Uses SNMP polling to get its data

In this case the data will be provided by the Cisco switches in the classroom. Will be able to determine network load over time and compare real numbers coming from devices.

  • Requires the SNMP settings be placed on the switches and configured to talk to the SNMP server
  • Software uses a web interface to provide management of the SNMP server
  • Alarms and reporting handled in the software
     

vCenter Monitoring

vCenter’s built in monitoring of its attached hardware

Allows direct view of hardware resources being accessed by Vcenter

Creates no extra overhead from additional software or services having to be run on a VM or physical hardware. Allows for historical monitoring and information can be pushed into excel spreadsheet format to further work with. We will need actual VM’s running to further create the correct metrics to verify we are tracking the needed data. This will allow us to track hard drive performance and CPU usage.

Initial Setup of PRTG walk through

Startup configuration guru (their version of a wizard)

  1. Start at the Welcome screen and choose the "Start Guru" button.

  2. Enable SSL encryption for PRTG Web Interface: Choose the "Yes" button to enable encryption

  3. Enter Administrator credentials for Windows Systems in your Network: Enter the domain or computer name, username, and password then save

  4. Enter Credentials for SNMP Based Network Devices: Slect your version of SNMP to use. I went with v2 since it was the most commonly supported with Cisco devices. Then enter the community string, SNMP port, and SNMP timeout (seconds).

  5. Enter Credentials for VMware/XenServer: There are built in sensors for monitoring ESXi (something that maybe useful for additional tracking during testing).

  6. There are additional questions about premade sensors like Mac servers and the like that will not apply to us.

  7. After the several questions about other network devices, you have the option of setting up auto discovery to locate other devices on your network. This only works well on a very small network, I was able to exceed the counters allowed in this version just on my home network. Manual setup for counters will likely be the most reasonable plan for implementation in our test environment.

  8. Once initial configuration is completed, you will be returned to the monitoring page where you can then track and modify the sensors you have set up for the network.

The addition of a sensor that tracks the bandwidth the switch is using is fairly straightforward. Then view traffic generation information.


Switch configuration for the SNMP monitoring

From the switch the configuration is pretty simple. The following screen capture shows all the commands that are required to setup the monitoring.


switch configuration for SNMP monitoring screen shot. View the commands below the image.
 

The commands are entered in global configuration:

User Access Verification

PasswordL

3550>en

Password:3550#conf t

Enter configuration commands, one per line. End with CNTL/Z.

3550(config)#snmp-server community netlabtest1 ro

3550(config)#access-list 10 permit 192.168.20.1 0.0.0.0

3550(config)#snmp-server community netlabtest1 ro 10

3550(config)#snmp-server enable traps

% Cannot enable both sham-link state-change interface traps.

% New sham link interface trap not enabled.

3550(config)#^Z

3550#wr

Building configureation...

[OK]

3550#


The first command snmp-server community netlabtest1 ro does several things. The snmp-server community is used to set the snmp access string that will be used by the PRTG server for polling the switch. The netlabtest1 portion of the command creates the string name and ro makes it as a read only string. There are large security issues if the string write enabled instead of read, like the total loss of control of your device.

The next command access-list 10 permit 192.168.20.1 0.0.0.0 creates an access list that allows only the PRTG server to poll this switch. This command just adds an additional layer of security.

That command is followed with the snmp-server community netlabtest1 ro 10 command which assigns the access list that was just created to the snmp-community string of netlabtest1.

The final command snmp-server enable traps just activates all the traps that are built into the switch by default. (This is likely not necessary for what we will be using this feature for).

VCENTER BUILT-IN PERFORMANCE MONITORING:

Example of what the  tracking page will monitor:

screenshot shows line charts for the top 10: Average Device Latency per Host, Max Queue Dept per Host, Read IOPs per Host, and Write IOPs per Host

Cisco’s documentation on SNMP setup