You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: docs/index.rst
+26Lines changed: 26 additions & 0 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -111,6 +111,32 @@ Apply the initialization file
111
111
112
112
Once the initialization file has been generated at ``cluster.trig``, apply this file to your empty quadstore (for example, by using its Graph Store Protocol API to insert the data) before using Flexo MMS.
113
113
114
+
Performance Considerations
115
+
--------------------------
116
+
117
+
Writing to a triplestore can impact query performance significantly. For production deployments with moderate to high load, we recommend using a dedicated write node and separate read-only nodes for the triple store.
118
+
119
+
The service configuration supports this architecture through the following settings:
120
+
121
+
- ``FLEXO_MMS_UPDATE_URL``: Points to your dedicated write node
122
+
- ``FLEXO_MMS_QUERY_URL``: Points to your read-only node(s)
123
+
- ``FLEXO_MMS_MASTER_QUERY_URL``: Optional endpoint for when you need the most up-to-date data (typically points to the write node)
124
+
125
+
With this configuration:
126
+
127
+
1. All write operations (updates, deletes) will be directed to the dedicated write node
128
+
2. Most read operations (queries) will be directed to the read-only node(s)
129
+
3. When absolute consistency is required, queries will use the master query URL
130
+
131
+
This separation provides several benefits:
132
+
133
+
- Improved query performance by offloading read traffic from the write node
134
+
- Better write throughput by dedicating resources to write operations
135
+
- Horizontal scalability by adding more read nodes as query demand increases
136
+
- Improved reliability as read operations can continue even if the write node is under heavy load
137
+
138
+
For smaller deployments with limited traffic, a single node configuration may be sufficient, but as your model size and user base grow, the separated architecture becomes increasingly important.
0 commit comments