high notification thread latency during RESTful queries

Document ID : KB000113584
Last Modified Date : 13/09/2018
Show Technical Document Details
Issue:
REST queries used for Discoveries to seed a CMDB database is causing high impact on performance to SpectroSERVER. During the period of the seed, SpectroSERVER gets so busy it cannot even poll. This last for some time and occurs consistently. What is happening? 
Cause:
In this case the cause is the attribute 0x129fa used for the seeding tool. In older 9.4 environment, attribute 0x129fa Model Type ID was querying a range of model types, but is an inefficient query. Now in Spectrum 10 with a much larger environment, this attribute is heavily impacting the performance during the seeding tool running. 
Resolution:
This attribute query is no longer needed for Spectrum 10.2.3 environment. Remove the rest calls using this attribute and the performance of SpectroSERVER should improve. 

While this issue is caused by a specific attribute query, this type of issue can occur on other inefficient searches as well. Typically searches against external attributes which need to poll devices will cause much higher latency than querying internal attributes. 

Spectrum has several internal attributes which store external attributes polled. For example,

External: ifAlias - 0x117fe
Internal: X_ifAlias - 0x13224

A query search on 0x13224 would be much more efficient than 0x117f3 due to the fact that Spectrum stores value of ifAlias in X_ifAlias for such a purpose here.