Merge aefdced into 86cadcb

README.md

@@ -20,6 +20,17 @@ It currently also experimentally supports the injection of
 a backward compatible way with NGSI-v2 API. I.e. you can retrieve NGSI-LD stored data via NGSI v2
 API and retrieve data will be describe following NGSI v2 format.
 
+QuantumLeap is not a [real time](https://en.wikipedia.org/wiki/Real-time_computing)
+API, its purpose is to process notifications received from the Context Broker
+and to create temporal records for them. In general, the whole FIWARE stack,
+being based on a micro-service architecture, cannot be regarded as real time
+in case you have requirements on guaranteed delivery in a given amount of time.
+
+However, even though hard real time may not be FIWARE's forte, in our experience
+a properly tuned FIWARE stack can perform extremely well and
+handle very demanding IoT workloads without a glitch - you just need to
+configure your infrastructure to handle that :)
+
 Want to know more? Refer to the [docs](https://quantumleap.readthedocs.io/en/latest/)
 or checkout the Extra Resources below.
 

docs/manuals/user/troubleshooting.md

@@ -55,6 +55,24 @@ crate.client.exceptions.ProgrammingError: SQLActionException[ColumnValidationExc
 This related to the fact that CrateDB does not support 3D coordinates,
 as documented in [admin documentation](../admin/crate.md).
 
+### Sometimes it takes more than 500 msec to read the data sent to Orion in QuantumLeap
+
+QuantumLeap is a Timeseries API that stores values forwarded by the Context Broker,
+and due to the nature of its backends as well, there is always some synch
+latency between the data writing and the time the data is available for reading.
+
+1. Orion takes some msec to process a request and trigger a notification.
+
+1. The QL takes some msec to process a single message and store it in the database.
+
+1. Especially in the case of CrateDB, indexing of inserted data may take a bit,
+    so this means that there is additional latency between when the a message
+    is stored in crate, and when it is actually available for querying. In case
+    of multi-node CrateDB deployment this can take even more because QL writes
+    on Crate node A, first the data is indexed in node A, and then replicated
+    on node B. So if you issue a query right after writing a message and QL
+    picks node B, probability to find the data you just pushed is even lower.
+
 ## Bug reporting
 
 Bugs should be reported in the form of