[DiskBBQ] Open vector tmp files with ReadAdvice.SEQUENTIAL (#133375)

Author: iverase

server/src/main/java/org/elasticsearch/index/codec/vectors/IVFVectorsWriter.java

@@ -28,6 +28,7 @@
 import org.apache.lucene.store.IndexInput;
 import org.apache.lucene.store.IndexOutput;
 import org.apache.lucene.store.RandomAccessInput;
+import org.apache.lucene.store.ReadAdvice;
 import org.apache.lucene.util.LongValues;
 import org.apache.lucene.util.VectorUtil;
 import org.elasticsearch.core.IOUtils;
@@ -295,9 +296,17 @@ private void mergeOneFieldIVF(FieldInfo fieldInfo, MergeState mergeState) throws
                 }
             }
         }
+        // now open the temp file and build the index structures. It is expected these files to be read in sequential order.
+        // Even when the file might be sample, the reads will be always in increase order, therefore we set the ReadAdvice to SEQUENTIAL
+        // so the OS can optimize read ahead in low memory situations.
         try (
-            IndexInput vectors = mergeState.segmentInfo.dir.openInput(tempRawVectorsFileName, IOContext.DEFAULT);
-            IndexInput docs = docsFileName == null ? null : mergeState.segmentInfo.dir.openInput(docsFileName, IOContext.DEFAULT)
+            IndexInput vectors = mergeState.segmentInfo.dir.openInput(
+                tempRawVectorsFileName,
+                IOContext.DEFAULT.withReadAdvice(ReadAdvice.SEQUENTIAL)
+            );
+            IndexInput docs = docsFileName == null
+                ? null
+                : mergeState.segmentInfo.dir.openInput(docsFileName, IOContext.DEFAULT.withReadAdvice(ReadAdvice.SEQUENTIAL))
         ) {
             final FloatVectorValues floatVectorValues = getFloatVectorValues(fieldInfo, docs, vectors, numVectors);