Let us take an example of XYZ company having the following requirements for a Carrier Class application and explain how to go about deploying in their site.
There are 50 networks to be managed and all the networks are connected by links of different line speed.
Each network consists of 200 devices, making a total of 10,000 devices.
20 Managed Objects per device, making a total of 200,000 MOs
Two PolledData per MO, making a total of 400,000 PolledData.
Traps received at the rate of 100 per second.
There are 50 clients monitoring the device status.
All the clients would connect from different geographical locations and the server locale will be in English.
The clients have to support Japanese, Chinese, and German.
The following are the considerations made, based on the deployment scenario stated above.
Placing of Web NMS Server: Since the number of devices in each of the networks is equal, the Web NMS server can be placed in any network. But, since each network's line speed is different, place the Web NMS server in the network that has a high speed link, compared to others.
Time Zone Considerations: The clients would connect from the countries such as Chine, Japan and Germany, and hence ensure that the time zone related configuration are done in the Web NMS server before deployment.
Internationalization Considerations: Since the clients use the locales Chinese, Japanese, and German, the necessary resource bundles have to be created. As the resource bundles are already available for Chinese and Japanese, you need to create the resource bundle only for the German language.
Security Considerations: Since the servers and the database are distributed, there is every chance that the servers might get hacked. Hence, make sure that the systems are hardened properly. This would ensure that the systems are not affected through the known security vulnerabilities.
Choosing the Deployment Model
If the number of Managed Objects and the PolledData are heavier, and the number of clients connecting to the server would be around 50, choose the "Type III Deployment" model. In this model, the BE, FE, and the database are run in different machines with very high hardware configurations. High Availability Deployment Model is chosen, as a small amount of downtime will cause huge loss of revenues.
Choosing the Database
Considering the number of Managed Objects and PolledData, the number of read/write operations that would be involved in such cases, etc., it is suggested to use "Oracle" as the database.
Tuning the Hardware
As the performance requirements are on the heavier side, with more number of Managed Objects and PolledData and high data collection rates required, tune the hardware for optimal performance. Tuning the hardware helps in optimizing the I/O activities of the system, thus provides you higher performance.