07-11-2017 08:58 AM - edited 03-01-2019 03:55 AM
Hi, I am trying to find information about how to design a NSO solution in order to try to figure out this questions that one of my clients ask me.
Thanks.
Solved! Go to Solution.
07-12-2017 01:42 AM
How many services and devices can support with just one NSO box
There is a sizing guide which gives a starting point. NSO stores the configs for all devices and service instances in the in-memory database (CDB). So an assessment needs to be made of the current size of the device+service configs and also the projected growth. Add some buffer for metadata (backpointers, refcounts etc).
With regards to request throughput, almost always commit-queues need to be used to scale. Some simulations that were done, a rate of 10-15 requests per second seems to be very possible with a single NSO node. Obviously depends on the CPU specs.
How should be the migration from legacy networks to NSO solution
NSO will "sync-from" the device configuration out of the box. In terms of service activation, an incremental approach is preferred. Take a greenfield service/use-case if possible. Otherwise pick some high value high frequency use-case[s] and start there. Migration of all brownfield service instances (a.k.a. service reconciliation) can be a very tedious and expensive. Depends on the current processes that were used to provision the services in the first place. Number of variations etc
In a Service Provider environment, in which part of the network should be installed the NSO
No hard and fast rule here. NSO can be installed on bare-metal or in a VM form-factor. The Linux server it runs on needs to be locked down (system install - admin guide) and NSO needs to have connectivity/access to all the devices that it needs to manage.
07-12-2017 01:42 AM
How many services and devices can support with just one NSO box
There is a sizing guide which gives a starting point. NSO stores the configs for all devices and service instances in the in-memory database (CDB). So an assessment needs to be made of the current size of the device+service configs and also the projected growth. Add some buffer for metadata (backpointers, refcounts etc).
With regards to request throughput, almost always commit-queues need to be used to scale. Some simulations that were done, a rate of 10-15 requests per second seems to be very possible with a single NSO node. Obviously depends on the CPU specs.
How should be the migration from legacy networks to NSO solution
NSO will "sync-from" the device configuration out of the box. In terms of service activation, an incremental approach is preferred. Take a greenfield service/use-case if possible. Otherwise pick some high value high frequency use-case[s] and start there. Migration of all brownfield service instances (a.k.a. service reconciliation) can be a very tedious and expensive. Depends on the current processes that were used to provision the services in the first place. Number of variations etc
In a Service Provider environment, in which part of the network should be installed the NSO
No hard and fast rule here. NSO can be installed on bare-metal or in a VM form-factor. The Linux server it runs on needs to be locked down (system install - admin guide) and NSO needs to have connectivity/access to all the devices that it needs to manage.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide