diff --git a/lib/mergeset/encoding.go b/lib/mergeset/encoding.go
index 94eaa14157..9eab5ebd1f 100644
--- a/lib/mergeset/encoding.go
+++ b/lib/mergeset/encoding.go
@@ -185,7 +185,7 @@ func (ib *inmemoryBlock) marshalData(sb *storageBlock, firstItemDst, commonPrefi
 	firstItemDst = append(firstItemDst, ib.items[0]...)
 	commonPrefixDst = append(commonPrefixDst, ib.commonPrefix...)
 
-	if len(ib.data)-len(ib.commonPrefix)*len(ib.items) < 64 || len(ib.items) < 10 {
+	if len(ib.data)-len(ib.commonPrefix)*len(ib.items) < 64 || len(ib.items) < 2 {
 		// Use plain encoding form small block, since it is cheaper.
 		ib.marshalDataPlain(sb)
 		return firstItemDst, commonPrefixDst, uint32(len(ib.items)), marshalTypePlain
diff --git a/lib/mergeset/merge.go b/lib/mergeset/merge.go
index 27a6a014db..398fffa9d8 100644
--- a/lib/mergeset/merge.go
+++ b/lib/mergeset/merge.go
@@ -7,16 +7,27 @@ import (
 	"sync/atomic"
 )
 
+// PrepareBlockCallback can transform the passed items allocated at the given data.
+//
+// The callback is called during merge before flushing full block of the given items
+// to persistent storage.
+//
+// The callback must return sorted items.
+// The callback can re-use data and items for storing the result.
+type PrepareBlockCallback func(data []byte, items [][]byte) ([]byte, [][]byte)
+
 // mergeBlockStreams merges bsrs and writes result to bsw.
 //
 // It also fills ph.
 //
+// prepareBlock is optional.
+//
 // The function immediately returns when stopCh is closed.
 //
 // It also atomically adds the number of items merged to itemsMerged.
-func mergeBlockStreams(ph *partHeader, bsw *blockStreamWriter, bsrs []*blockStreamReader, stopCh <-chan struct{}, itemsMerged *uint64) error {
+func mergeBlockStreams(ph *partHeader, bsw *blockStreamWriter, bsrs []*blockStreamReader, prepareBlock PrepareBlockCallback, stopCh <-chan struct{}, itemsMerged *uint64) error {
 	bsm := bsmPool.Get().(*blockStreamMerger)
-	if err := bsm.Init(bsrs); err != nil {
+	if err := bsm.Init(bsrs, prepareBlock); err != nil {
 		return fmt.Errorf("cannot initialize blockStreamMerger: %s", err)
 	}
 	err := bsm.Merge(bsw, ph, stopCh, itemsMerged)
@@ -39,6 +50,8 @@ var bsmPool = &sync.Pool{
 }
 
 type blockStreamMerger struct {
+	prepareBlock PrepareBlockCallback
+
 	bsrHeap bsrHeap
 
 	// ib is a scratch block with pending items.
@@ -48,6 +61,8 @@ type blockStreamMerger struct {
 }
 
 func (bsm *blockStreamMerger) reset() {
+	bsm.prepareBlock = nil
+
 	for i := range bsm.bsrHeap {
 		bsm.bsrHeap[i] = nil
 	}
@@ -57,8 +72,9 @@ func (bsm *blockStreamMerger) reset() {
 	bsm.phFirstItemCaught = false
 }
 
-func (bsm *blockStreamMerger) Init(bsrs []*blockStreamReader) error {
+func (bsm *blockStreamMerger) Init(bsrs []*blockStreamReader, prepareBlock PrepareBlockCallback) error {
 	bsm.reset()
+	bsm.prepareBlock = prepareBlock
 	for _, bsr := range bsrs {
 		if bsr.Next() {
 			bsm.bsrHeap = append(bsm.bsrHeap, bsr)
@@ -134,9 +150,11 @@ func (bsm *blockStreamMerger) flushIB(bsw *blockStreamWriter, ph *partHeader, it
 		// Nothing to flush.
 		return
 	}
-	itemsCount := uint64(len(bsm.ib.items))
-	ph.itemsCount += itemsCount
-	atomic.AddUint64(itemsMerged, itemsCount)
+	atomic.AddUint64(itemsMerged, uint64(len(bsm.ib.items)))
+	if bsm.prepareBlock != nil {
+		bsm.ib.data, bsm.ib.items = bsm.prepareBlock(bsm.ib.data, bsm.ib.items)
+	}
+	ph.itemsCount += uint64(len(bsm.ib.items))
 	if !bsm.phFirstItemCaught {
 		ph.firstItem = append(ph.firstItem[:0], bsm.ib.items[0]...)
 		bsm.phFirstItemCaught = true
diff --git a/lib/mergeset/merge_test.go b/lib/mergeset/merge_test.go
index 8fb80023dc..02499ffd57 100644
--- a/lib/mergeset/merge_test.go
+++ b/lib/mergeset/merge_test.go
@@ -30,14 +30,14 @@ func TestMultilevelMerge(t *testing.T) {
 	var dstIP1 inmemoryPart
 	var bsw1 blockStreamWriter
 	bsw1.InitFromInmemoryPart(&dstIP1, 0)
-	if err := mergeBlockStreams(&dstIP1.ph, &bsw1, bsrs[:5], nil, &itemsMerged); err != nil {
+	if err := mergeBlockStreams(&dstIP1.ph, &bsw1, bsrs[:5], nil, nil, &itemsMerged); err != nil {
 		t.Fatalf("cannot merge first level part 1: %s", err)
 	}
 
 	var dstIP2 inmemoryPart
 	var bsw2 blockStreamWriter
 	bsw2.InitFromInmemoryPart(&dstIP2, 0)
-	if err := mergeBlockStreams(&dstIP2.ph, &bsw2, bsrs[5:], nil, &itemsMerged); err != nil {
+	if err := mergeBlockStreams(&dstIP2.ph, &bsw2, bsrs[5:], nil, nil, &itemsMerged); err != nil {
 		t.Fatalf("cannot merge first level part 2: %s", err)
 	}
 
@@ -54,7 +54,7 @@ func TestMultilevelMerge(t *testing.T) {
 		newTestBlockStreamReader(&dstIP2),
 	}
 	bsw.InitFromInmemoryPart(&dstIP, 0)
-	if err := mergeBlockStreams(&dstIP.ph, &bsw, bsrsTop, nil, &itemsMerged); err != nil {
+	if err := mergeBlockStreams(&dstIP.ph, &bsw, bsrsTop, nil, nil, &itemsMerged); err != nil {
 		t.Fatalf("cannot merge second level: %s", err)
 	}
 	if itemsMerged != uint64(len(items)) {
@@ -76,7 +76,7 @@ func TestMergeForciblyStop(t *testing.T) {
 	ch := make(chan struct{})
 	var itemsMerged uint64
 	close(ch)
-	if err := mergeBlockStreams(&dstIP.ph, &bsw, bsrs, ch, &itemsMerged); err != errForciblyStopped {
+	if err := mergeBlockStreams(&dstIP.ph, &bsw, bsrs, nil, ch, &itemsMerged); err != errForciblyStopped {
 		t.Fatalf("unexpected error during merge: got %v; want %v", err, errForciblyStopped)
 	}
 	if itemsMerged != 0 {
@@ -120,7 +120,7 @@ func testMergeBlockStreamsSerial(blocksToMerge, maxItemsPerBlock int) error {
 	var dstIP inmemoryPart
 	var bsw blockStreamWriter
 	bsw.InitFromInmemoryPart(&dstIP, 0)
-	if err := mergeBlockStreams(&dstIP.ph, &bsw, bsrs, nil, &itemsMerged); err != nil {
+	if err := mergeBlockStreams(&dstIP.ph, &bsw, bsrs, nil, nil, &itemsMerged); err != nil {
 		return fmt.Errorf("cannot merge block streams: %s", err)
 	}
 	if itemsMerged != uint64(len(items)) {
diff --git a/lib/mergeset/part_search_test.go b/lib/mergeset/part_search_test.go
index 3e35f54a44..29dce77093 100644
--- a/lib/mergeset/part_search_test.go
+++ b/lib/mergeset/part_search_test.go
@@ -150,7 +150,7 @@ func newTestPart(blocksCount, maxItemsPerBlock int) (*part, []string, error) {
 	var ip inmemoryPart
 	var bsw blockStreamWriter
 	bsw.InitFromInmemoryPart(&ip, 0)
-	if err := mergeBlockStreams(&ip.ph, &bsw, bsrs, nil, &itemsMerged); err != nil {
+	if err := mergeBlockStreams(&ip.ph, &bsw, bsrs, nil, nil, &itemsMerged); err != nil {
 		return nil, nil, fmt.Errorf("cannot merge blocks: %s", err)
 	}
 	if itemsMerged != uint64(len(items)) {
diff --git a/lib/mergeset/table.go b/lib/mergeset/table.go
index 458b09a7e0..77d4fb3257 100644
--- a/lib/mergeset/table.go
+++ b/lib/mergeset/table.go
@@ -74,6 +74,8 @@ type Table struct {
 
 	flushCallback func()
 
+	prepareBlock PrepareBlockCallback
+
 	partsLock sync.Mutex
 	parts     []*partWrapper
 
@@ -94,6 +96,8 @@ type Table struct {
 
 	rawItemsFlusherWG sync.WaitGroup
 
+	convertersWG sync.WaitGroup
+
 	// Use syncwg instead of sync, since Add/Wait may be called from concurrent goroutines.
 	rawItemsPendingFlushesWG syncwg.WaitGroup
 
@@ -139,8 +143,11 @@ func (pw *partWrapper) decRef() {
 // Optional flushCallback is called every time new data batch is flushed
 // to the underlying storage and becomes visible to search.
 //
+// Optional prepareBlock is called during merge before flushing the prepared block
+// to persistent storage.
+//
 // The table is created if it doesn't exist yet.
-func OpenTable(path string, flushCallback func()) (*Table, error) {
+func OpenTable(path string, flushCallback func(), prepareBlock PrepareBlockCallback) (*Table, error) {
 	path = filepath.Clean(path)
 	logger.Infof("opening table %q...", path)
 	startTime := time.Now()
@@ -165,6 +172,7 @@ func OpenTable(path string, flushCallback func()) (*Table, error) {
 	tb := &Table{
 		path:          path,
 		flushCallback: flushCallback,
+		prepareBlock:  prepareBlock,
 		parts:         pws,
 		mergeIdx:      uint64(time.Now().UnixNano()),
 		flockF:        flockF,
@@ -178,6 +186,12 @@ func OpenTable(path string, flushCallback func()) (*Table, error) {
 	logger.Infof("table %q has been opened in %s; partsCount: %d; blocksCount: %d, itemsCount: %d; sizeBytes: %d",
 		path, time.Since(startTime), m.PartsCount, m.BlocksCount, m.ItemsCount, m.SizeBytes)
 
+	tb.convertersWG.Add(1)
+	go func() {
+		tb.convertToV1280()
+		tb.convertersWG.Done()
+	}()
+
 	return tb, nil
 }
 
@@ -190,6 +204,11 @@ func (tb *Table) MustClose() {
 	tb.rawItemsFlusherWG.Wait()
 	logger.Infof("raw items flusher stopped in %s on %q", time.Since(startTime), tb.path)
 
+	logger.Infof("waiting for converters to stop on %q...", tb.path)
+	startTime = time.Now()
+	tb.convertersWG.Wait()
+	logger.Infof("converters stopped in %s on %q", time.Since(startTime), tb.path)
+
 	logger.Infof("waiting for part mergers to stop on %q...", tb.path)
 	startTime = time.Now()
 	tb.partMergersWG.Wait()
@@ -216,7 +235,7 @@ func (tb *Table) MustClose() {
 	}
 	tb.partsLock.Unlock()
 
-	if err := tb.mergePartsOptimal(pws); err != nil {
+	if err := tb.mergePartsOptimal(pws, nil); err != nil {
 		logger.Panicf("FATAL: cannot flush inmemory parts to files in %q: %s", tb.path, err)
 	}
 	logger.Infof("%d inmemory parts have been flushed to files in %s on %q", len(pws), time.Since(startTime), tb.path)
@@ -393,15 +412,67 @@ func (tb *Table) rawItemsFlusher() {
 	}
 }
 
-func (tb *Table) mergePartsOptimal(pws []*partWrapper) error {
+const convertToV1280FileName = "converted-to-v1.28.0"
+
+func (tb *Table) convertToV1280() {
+	// Convert tag->metricID rows into tag->metricIDs rows when upgrading to v1.28.0+.
+	flagFilePath := tb.path + "/" + convertToV1280FileName
+	if fs.IsPathExist(flagFilePath) {
+		// The conversion has been already performed.
+		return
+	}
+
+	getAllPartsForMerge := func() []*partWrapper {
+		var pws []*partWrapper
+		tb.partsLock.Lock()
+		for _, pw := range tb.parts {
+			if pw.isInMerge {
+				continue
+			}
+			pw.isInMerge = true
+			pws = append(pws, pw)
+		}
+		tb.partsLock.Unlock()
+		return pws
+	}
+	pws := getAllPartsForMerge()
+	if len(pws) > 0 {
+		logger.Infof("started round 1 of background conversion of %q to v1.28.0 format; merge %d parts", tb.path, len(pws))
+		startTime := time.Now()
+		if err := tb.mergePartsOptimal(pws, tb.stopCh); err != nil {
+			logger.Errorf("failed round 1 of background conversion of %q to v1.28.0 format: %s", tb.path, err)
+			return
+		}
+		logger.Infof("finished round 1 of background conversion of %q to v1.28.0 format in %s", tb.path, time.Since(startTime))
+
+		// The second round is needed in order to merge small blocks
+		// with tag->metricIDs rows left after the first round.
+		pws = getAllPartsForMerge()
+		logger.Infof("started round 2 of background conversion of %q to v1.28.0 format; merge %d parts", tb.path, len(pws))
+		startTime = time.Now()
+		if len(pws) > 0 {
+			if err := tb.mergePartsOptimal(pws, tb.stopCh); err != nil {
+				logger.Errorf("failed round 2 of background conversion of %q to v1.28.0 format: %s", tb.path, err)
+				return
+			}
+		}
+		logger.Infof("finished round 2 of background conversion of %q to v1.28.0 format in %s", tb.path, time.Since(startTime))
+	}
+
+	if err := fs.WriteFileAtomically(flagFilePath, []byte("ok")); err != nil {
+		logger.Panicf("FATAL: cannot create %q: %s", flagFilePath, err)
+	}
+}
+
+func (tb *Table) mergePartsOptimal(pws []*partWrapper, stopCh <-chan struct{}) error {
 	for len(pws) > defaultPartsToMerge {
-		if err := tb.mergeParts(pws[:defaultPartsToMerge], nil, false); err != nil {
+		if err := tb.mergeParts(pws[:defaultPartsToMerge], stopCh, false); err != nil {
 			return fmt.Errorf("cannot merge %d parts: %s", defaultPartsToMerge, err)
 		}
 		pws = pws[defaultPartsToMerge:]
 	}
 	if len(pws) > 0 {
-		if err := tb.mergeParts(pws, nil, false); err != nil {
+		if err := tb.mergeParts(pws, stopCh, false); err != nil {
 			return fmt.Errorf("cannot merge %d parts: %s", len(pws), err)
 		}
 	}
@@ -541,7 +612,7 @@ func (tb *Table) mergeInmemoryBlocks(blocksToMerge []*inmemoryBlock) *partWrappe
 	// Merge parts.
 	// The merge shouldn't be interrupted by stopCh,
 	// since it may be final after stopCh is closed.
-	if err := mergeBlockStreams(&mpDst.ph, bsw, bsrs, nil, &tb.itemsMerged); err != nil {
+	if err := mergeBlockStreams(&mpDst.ph, bsw, bsrs, tb.prepareBlock, nil, &tb.itemsMerged); err != nil {
 		logger.Panicf("FATAL: cannot merge inmemoryBlocks: %s", err)
 	}
 	putBlockStreamWriter(bsw)
@@ -700,7 +771,7 @@ func (tb *Table) mergeParts(pws []*partWrapper, stopCh <-chan struct{}, isOuterP
 
 	// Merge parts into a temporary location.
 	var ph partHeader
-	err := mergeBlockStreams(&ph, bsw, bsrs, stopCh, &tb.itemsMerged)
+	err := mergeBlockStreams(&ph, bsw, bsrs, tb.prepareBlock, stopCh, &tb.itemsMerged)
 	putBlockStreamWriter(bsw)
 	if err != nil {
 		if err == errForciblyStopped {
@@ -950,11 +1021,20 @@ func (tb *Table) CreateSnapshotAt(dstDir string) error {
 		return fmt.Errorf("cannot read directory: %s", err)
 	}
 	for _, fi := range fis {
+		fn := fi.Name()
 		if !fs.IsDirOrSymlink(fi) {
-			// Skip non-directories.
+			switch fn {
+			case convertToV1280FileName:
+				srcPath := srcDir + "/" + fn
+				dstPath := dstDir + "/" + fn
+				if err := os.Link(srcPath, dstPath); err != nil {
+					return fmt.Errorf("cannot hard link from %q to %q: %s", srcPath, dstPath, err)
+				}
+			default:
+				// Skip other non-directories.
+			}
 			continue
 		}
-		fn := fi.Name()
 		if isSpecialDir(fn) {
 			// Skip special dirs.
 			continue
diff --git a/lib/mergeset/table_search_test.go b/lib/mergeset/table_search_test.go
index 806e2a8878..3673ac8ab2 100644
--- a/lib/mergeset/table_search_test.go
+++ b/lib/mergeset/table_search_test.go
@@ -41,7 +41,7 @@ func TestTableSearchSerial(t *testing.T) {
 
 	func() {
 		// Re-open the table and verify the search works.
-		tb, err := OpenTable(path, nil)
+		tb, err := OpenTable(path, nil, nil)
 		if err != nil {
 			t.Fatalf("cannot open table: %s", err)
 		}
@@ -76,7 +76,7 @@ func TestTableSearchConcurrent(t *testing.T) {
 
 	// Re-open the table and verify the search works.
 	func() {
-		tb, err := OpenTable(path, nil)
+		tb, err := OpenTable(path, nil, nil)
 		if err != nil {
 			t.Fatalf("cannot open table: %s", err)
 		}
@@ -152,7 +152,7 @@ func newTestTable(path string, itemsCount int) (*Table, []string, error) {
 	flushCallback := func() {
 		atomic.AddUint64(&flushes, 1)
 	}
-	tb, err := OpenTable(path, flushCallback)
+	tb, err := OpenTable(path, flushCallback, nil)
 	if err != nil {
 		return nil, nil, fmt.Errorf("cannot open table: %s", err)
 	}
diff --git a/lib/mergeset/table_search_timing_test.go b/lib/mergeset/table_search_timing_test.go
index 16c2f6035f..f8e0c9851b 100644
--- a/lib/mergeset/table_search_timing_test.go
+++ b/lib/mergeset/table_search_timing_test.go
@@ -32,7 +32,7 @@ func benchmarkTableSearch(b *testing.B, itemsCount int) {
 
 	// Force finishing pending merges
 	tb.MustClose()
-	tb, err = OpenTable(path, nil)
+	tb, err = OpenTable(path, nil, nil)
 	if err != nil {
 		b.Fatalf("unexpected error when re-opening table %q: %s", path, err)
 	}
diff --git a/lib/mergeset/table_test.go b/lib/mergeset/table_test.go
index db063f1ebf..aa43fe9e11 100644
--- a/lib/mergeset/table_test.go
+++ b/lib/mergeset/table_test.go
@@ -21,7 +21,7 @@ func TestTableOpenClose(t *testing.T) {
 	}()
 
 	// Create a new table
-	tb, err := OpenTable(path, nil)
+	tb, err := OpenTable(path, nil, nil)
 	if err != nil {
 		t.Fatalf("cannot create new table: %s", err)
 	}
@@ -31,7 +31,7 @@ func TestTableOpenClose(t *testing.T) {
 
 	// Re-open created table multiple times.
 	for i := 0; i < 10; i++ {
-		tb, err := OpenTable(path, nil)
+		tb, err := OpenTable(path, nil, nil)
 		if err != nil {
 			t.Fatalf("cannot open created table: %s", err)
 		}
@@ -45,14 +45,14 @@ func TestTableOpenMultipleTimes(t *testing.T) {
 		_ = os.RemoveAll(path)
 	}()
 
-	tb1, err := OpenTable(path, nil)
+	tb1, err := OpenTable(path, nil, nil)
 	if err != nil {
 		t.Fatalf("cannot open table: %s", err)
 	}
 	defer tb1.MustClose()
 
 	for i := 0; i < 10; i++ {
-		tb2, err := OpenTable(path, nil)
+		tb2, err := OpenTable(path, nil, nil)
 		if err == nil {
 			tb2.MustClose()
 			t.Fatalf("expecting non-nil error when opening already opened table")
@@ -73,7 +73,7 @@ func TestTableAddItemSerial(t *testing.T) {
 	flushCallback := func() {
 		atomic.AddUint64(&flushes, 1)
 	}
-	tb, err := OpenTable(path, flushCallback)
+	tb, err := OpenTable(path, flushCallback, nil)
 	if err != nil {
 		t.Fatalf("cannot open %q: %s", path, err)
 	}
@@ -99,7 +99,7 @@ func TestTableAddItemSerial(t *testing.T) {
 	testReopenTable(t, path, itemsCount)
 
 	// Add more items in order to verify merge between inmemory parts and file-based parts.
-	tb, err = OpenTable(path, nil)
+	tb, err = OpenTable(path, nil, nil)
 	if err != nil {
 		t.Fatalf("cannot open %q: %s", path, err)
 	}
@@ -132,7 +132,7 @@ func TestTableCreateSnapshotAt(t *testing.T) {
 		_ = os.RemoveAll(path)
 	}()
 
-	tb, err := OpenTable(path, nil)
+	tb, err := OpenTable(path, nil, nil)
 	if err != nil {
 		t.Fatalf("cannot open %q: %s", path, err)
 	}
@@ -163,13 +163,13 @@ func TestTableCreateSnapshotAt(t *testing.T) {
 	}()
 
 	// Verify snapshots contain all the data.
-	tb1, err := OpenTable(snapshot1, nil)
+	tb1, err := OpenTable(snapshot1, nil, nil)
 	if err != nil {
 		t.Fatalf("cannot open %q: %s", path, err)
 	}
 	defer tb1.MustClose()
 
-	tb2, err := OpenTable(snapshot2, nil)
+	tb2, err := OpenTable(snapshot2, nil, nil)
 	if err != nil {
 		t.Fatalf("cannot open %q: %s", path, err)
 	}
@@ -217,7 +217,12 @@ func TestTableAddItemsConcurrent(t *testing.T) {
 	flushCallback := func() {
 		atomic.AddUint64(&flushes, 1)
 	}
-	tb, err := OpenTable(path, flushCallback)
+	var itemsMerged uint64
+	prepareBlock := func(data []byte, items [][]byte) ([]byte, [][]byte) {
+		atomic.AddUint64(&itemsMerged, uint64(len(items)))
+		return data, items
+	}
+	tb, err := OpenTable(path, flushCallback, prepareBlock)
 	if err != nil {
 		t.Fatalf("cannot open %q: %s", path, err)
 	}
@@ -230,6 +235,10 @@ func TestTableAddItemsConcurrent(t *testing.T) {
 	if atomic.LoadUint64(&flushes) == 0 {
 		t.Fatalf("unexpected zero flushes")
 	}
+	n := atomic.LoadUint64(&itemsMerged)
+	if n < itemsCount {
+		t.Fatalf("too low number of items merged; got %v; must be at least %v", n, itemsCount)
+	}
 
 	var m TableMetrics
 	tb.UpdateMetrics(&m)
@@ -243,7 +252,7 @@ func TestTableAddItemsConcurrent(t *testing.T) {
 	testReopenTable(t, path, itemsCount)
 
 	// Add more items in order to verify merge between inmemory parts and file-based parts.
-	tb, err = OpenTable(path, nil)
+	tb, err = OpenTable(path, nil, nil)
 	if err != nil {
 		t.Fatalf("cannot open %q: %s", path, err)
 	}
@@ -285,7 +294,7 @@ func testReopenTable(t *testing.T, path string, itemsCount int) {
 	t.Helper()
 
 	for i := 0; i < 10; i++ {
-		tb, err := OpenTable(path, nil)
+		tb, err := OpenTable(path, nil, nil)
 		if err != nil {
 			t.Fatalf("cannot re-open %q: %s", path, err)
 		}
diff --git a/lib/storage/index_db.go b/lib/storage/index_db.go
index bacf2c4dbf..43d5e73925 100644
--- a/lib/storage/index_db.go
+++ b/lib/storage/index_db.go
@@ -28,7 +28,7 @@ const (
 	nsPrefixMetricNameToTSID = 0
 
 	// Prefix for Tag->MetricID entries.
-	nsPrefixTagToMetricID = 1
+	nsPrefixTagToMetricIDs = 1
 
 	// Prefix for MetricID->TSID entries.
 	nsPrefixMetricIDToTSID = 2
@@ -116,7 +116,7 @@ func openIndexDB(path string, metricIDCache, metricNameCache *workingsetcache.Ca
 		logger.Panicf("BUG: prevHourMetricIDs must be non-nil")
 	}
 
-	tb, err := mergeset.OpenTable(path, invalidateTagCache)
+	tb, err := mergeset.OpenTable(path, invalidateTagCache, mergeTagToMetricIDsRows)
 	if err != nil {
 		return nil, fmt.Errorf("cannot open indexDB %q: %s", path, err)
 	}
@@ -451,6 +451,7 @@ type indexSearch struct {
 	db *indexDB
 	ts mergeset.TableSearch
 	kb bytesutil.ByteBuffer
+	mp tagToMetricIDsRowParser
 
 	// tsidByNameMisses and tsidByNameSkips is used for a performance
 	// hack in GetOrCreateTSIDByName. See the comment there.
@@ -505,6 +506,7 @@ func (db *indexDB) getIndexSearch() *indexSearch {
 func (db *indexDB) putIndexSearch(is *indexSearch) {
 	is.ts.MustClose()
 	is.kb.Reset()
+	is.mp.Reset()
 
 	// Do not reset tsidByNameMisses and tsidByNameSkips,
 	// since they are used in GetOrCreateTSIDByName across call boundaries.
@@ -548,7 +550,7 @@ func (db *indexDB) generateTSID(dst *TSID, metricName []byte, mn *MetricName) er
 		}
 	}
 
-	// The TSID wan't found in the external storage.
+	// The TSID wasn't found in the external storage.
 	// Generate it locally.
 	dst.AccountID = mn.AccountID
 	dst.ProjectID = mn.ProjectID
@@ -589,7 +591,7 @@ func (db *indexDB) createIndexes(tsid *TSID, mn *MetricName) error {
 	items.Next()
 
 	commonPrefix := kbPool.Get()
-	commonPrefix.B = marshalCommonPrefix(commonPrefix.B[:0], nsPrefixTagToMetricID, mn.AccountID, mn.ProjectID)
+	commonPrefix.B = marshalCommonPrefix(commonPrefix.B[:0], nsPrefixTagToMetricIDs, mn.AccountID, mn.ProjectID)
 
 	// Create MetricGroup -> MetricID index.
 	items.B = append(items.B, commonPrefix.B...)
@@ -680,50 +682,37 @@ func (db *indexDB) SearchTagKeys(accountID, projectID uint32, maxTagKeys int) ([
 func (is *indexSearch) searchTagKeys(accountID, projectID uint32, tks map[string]struct{}, maxTagKeys int) error {
 	ts := &is.ts
 	kb := &is.kb
+	mp := &is.mp
+	mp.Reset()
 	dmis := is.db.getDeletedMetricIDs()
-	commonPrefix := marshalCommonPrefix(nil, nsPrefixTagToMetricID, accountID, projectID)
-	ts.Seek(commonPrefix)
+	kb.B = marshalCommonPrefix(kb.B[:0], nsPrefixTagToMetricIDs, accountID, projectID)
+	prefix := kb.B
+	ts.Seek(prefix)
 	for len(tks) < maxTagKeys && ts.NextItem() {
 		item := ts.Item
-		if !bytes.HasPrefix(item, commonPrefix) {
+		if !bytes.HasPrefix(item, prefix) {
 			break
 		}
-		tail := item[len(commonPrefix):]
-
-		// Unmarshal tag key into kb.B
-		var err error
-		tail, kb.B, err = unmarshalTagValue(kb.B[:0], tail)
-		if err != nil {
-			return fmt.Errorf("cannot unmarshal tagKey from %X: %s", item, err)
+		if err := mp.Init(item); err != nil {
+			return err
 		}
-
-		// Verify that the tag key points to existing metric.
-		if len(tail) < 8 {
-			return fmt.Errorf("cannot unmarshal metricID from less than 8 bytes; got %d bytes; item=%X", len(tail), tail)
-		}
-		metricID := encoding.UnmarshalUint64(tail[len(tail)-8:])
-		if _, deleted := dmis[metricID]; deleted {
-			// The given metric is deleted. Skip it.
+		if mp.IsDeletedTag(dmis) {
 			continue
 		}
 
 		// Store tag key.
-		tks[string(kb.B)] = struct{}{}
+		tks[string(mp.Tag.Key)] = struct{}{}
 
 		// Search for the next tag key.
-		// tkp (tag key prefix) contains (commonPrefix + encoded tag key).
-		// The last char must be tagSeparatorChar. Just increment it
-		// in order to jump to the next tag key.
-		tkp := item[:len(item)-len(tail)]
-		if len(tkp) == 0 || tkp[len(tkp)-1] != tagSeparatorChar || tagSeparatorChar >= 0xff {
-			logger.Panicf("BUG: the last char in tkp=%X must be %X. Check unmarshalTagValue code", tkp, tagSeparatorChar)
-		}
-		kb.B = append(kb.B[:0], tkp...)
+		// The last char in kb.B must be tagSeparatorChar.
+		// Just increment it in order to jump to the next tag key.
+		kb.B = marshalCommonPrefix(kb.B[:0], nsPrefixTagToMetricIDs, accountID, projectID)
+		kb.B = marshalTagValue(kb.B, mp.Tag.Key)
 		kb.B[len(kb.B)-1]++
 		ts.Seek(kb.B)
 	}
 	if err := ts.Error(); err != nil {
-		return fmt.Errorf("error during search for commonPrefix %q: %s", commonPrefix, err)
+		return fmt.Errorf("error during search for prefix %q: %s", prefix, err)
 	}
 	return nil
 }
@@ -732,24 +721,18 @@ func (is *indexSearch) searchTagKeys(accountID, projectID uint32, tks map[string
 func (db *indexDB) SearchTagValues(accountID, projectID uint32, tagKey []byte, maxTagValues int) ([]string, error) {
 	// TODO: cache results?
 
-	kb := kbPool.Get()
-	kb.B = marshalCommonPrefix(kb.B[:0], nsPrefixTagToMetricID, accountID, projectID)
-	kb.B = marshalTagValue(kb.B, tagKey)
-
 	tvs := make(map[string]struct{})
 	is := db.getIndexSearch()
-	err := is.searchTagValues(tvs, kb.B, maxTagValues)
+	err := is.searchTagValues(accountID, projectID, tvs, tagKey, maxTagValues)
 	db.putIndexSearch(is)
 	if err != nil {
-		kbPool.Put(kb)
 		return nil, err
 	}
 	ok := db.doExtDB(func(extDB *indexDB) {
 		is := extDB.getIndexSearch()
-		err = is.searchTagValues(tvs, kb.B, maxTagValues)
+		err = is.searchTagValues(accountID, projectID, tvs, tagKey, maxTagValues)
 		extDB.putIndexSearch(is)
 	})
-	kbPool.Put(kb)
 	if ok && err != nil {
 		return nil, err
 	}
@@ -763,49 +746,37 @@ func (db *indexDB) SearchTagValues(accountID, projectID uint32, tagKey []byte, m
 	return tagValues, nil
 }
 
-func (is *indexSearch) searchTagValues(tvs map[string]struct{}, prefix []byte, maxTagValues int) error {
+func (is *indexSearch) searchTagValues(accountID, projectID uint32, tvs map[string]struct{}, tagKey []byte, maxTagValues int) error {
 	ts := &is.ts
 	kb := &is.kb
+	mp := &is.mp
+	mp.Reset()
 	dmis := is.db.getDeletedMetricIDs()
+	kb.B = marshalCommonPrefix(kb.B[:0], nsPrefixTagToMetricIDs, accountID, projectID)
+	kb.B = marshalTagValue(kb.B, tagKey)
+	prefix := kb.B
 	ts.Seek(prefix)
 	for len(tvs) < maxTagValues && ts.NextItem() {
-		k := ts.Item
-		if !bytes.HasPrefix(k, prefix) {
+		item := ts.Item
+		if !bytes.HasPrefix(item, prefix) {
 			break
 		}
-
-		// Get TagValue
-		k = k[len(prefix):]
-		var err error
-		k, kb.B, err = unmarshalTagValue(kb.B[:0], k)
-		if err != nil {
-			return fmt.Errorf("cannot unmarshal tagValue: %s", err)
+		if err := mp.Init(item); err != nil {
+			return err
 		}
-		if len(k) != 8 {
-			return fmt.Errorf("unexpected suffix after tag value; want %d bytes; got %d bytes", 8, len(k))
-		}
-
-		// Verify whether the corresponding metric is deleted.
-		if len(dmis) > 0 {
-			metricID := encoding.UnmarshalUint64(k)
-			if _, deleted := dmis[metricID]; deleted {
-				// The metric is deleted.
-				continue
-			}
+		if mp.IsDeletedTag(dmis) {
+			continue
 		}
 
 		// Store tag value
-		tvs[string(kb.B)] = struct{}{}
+		tvs[string(mp.Tag.Value)] = struct{}{}
 
 		// Search for the next tag value.
-		// tkp (tag key prefix) contains (commonPrefix + encoded tag value).
-		// The last char must be tagSeparatorChar. Just increment it
-		// in order to jump to the next tag key.
-		tkp := ts.Item[:len(ts.Item)-8]
-		if len(tkp) == 0 || tkp[len(tkp)-1] != tagSeparatorChar || tagSeparatorChar >= 0xff {
-			logger.Panicf("BUG: the last char in tkp=%X must be %X. Check unmarshalTagValue code", tkp, tagSeparatorChar)
-		}
-		kb.B = append(kb.B[:0], tkp...)
+		// The last char in kb.B must be tagSeparatorChar.
+		// Just increment it in order to jump to the next tag key.
+		kb.B = marshalCommonPrefix(kb.B[:0], nsPrefixTagToMetricIDs, accountID, projectID)
+		kb.B = marshalTagValue(kb.B, mp.Tag.Key)
+		kb.B = marshalTagValue(kb.B, mp.Tag.Value)
 		kb.B[len(kb.B)-1]++
 		ts.Seek(kb.B)
 	}
@@ -1460,7 +1431,7 @@ func (is *indexSearch) getTagFilterWithMinMetricIDsCount(tfs *TagFilters, maxMet
 }
 
 func matchTagFilters(mn *MetricName, tfs []*tagFilter, kb *bytesutil.ByteBuffer) (bool, error) {
-	kb.B = marshalCommonPrefix(kb.B[:0], nsPrefixTagToMetricID, mn.AccountID, mn.ProjectID)
+	kb.B = marshalCommonPrefix(kb.B[:0], nsPrefixTagToMetricIDs, mn.AccountID, mn.ProjectID)
 	for _, tf := range tfs {
 		if len(tf.key) == 0 {
 			// Match against mn.MetricGroup.
@@ -1628,7 +1599,10 @@ func (is *indexSearch) getMetricIDsForTagFilter(tf *tagFilter, maxMetrics int) (
 	if len(tf.orSuffixes) > 0 {
 		// Fast path for orSuffixes - seek for rows for each value from orSuffxies.
 		if err := is.updateMetricIDsForOrSuffixesNoFilter(tf, maxMetrics, metricIDs); err != nil {
-			return nil, err
+			if err == errFallbackToMetricNameMatch {
+				return nil, err
+			}
+			return nil, fmt.Errorf("error when searching for metricIDs for tagFilter in fast path: %s; tagFilter=%s", err, tf)
 		}
 		return metricIDs, nil
 	}
@@ -1640,7 +1614,10 @@ func (is *indexSearch) getMetricIDsForTagFilter(tf *tagFilter, maxMetrics int) (
 		return len(metricIDs) < maxMetrics
 	})
 	if err != nil {
-		return nil, err
+		if err == errFallbackToMetricNameMatch {
+			return nil, err
+		}
+		return nil, fmt.Errorf("error when searching for metricIDs for tagFilter in slow path: %s; tagFilter=%s", err, tf)
 	}
 	return metricIDs, nil
 }
@@ -1654,46 +1631,53 @@ func (is *indexSearch) getMetricIDsForTagFilterSlow(tf *tagFilter, maxLoops int,
 	loops := 0
 	ts := &is.ts
 	kb := &is.kb
-	var prevMatchingK []byte
+	mp := &is.mp
+	mp.Reset()
+	var prevMatchingSuffix []byte
 	var prevMatch bool
-	ts.Seek(tf.prefix)
+	prefix := tf.prefix
+	ts.Seek(prefix)
 	for ts.NextItem() {
-		loops++
-		if loops > maxLoops {
-			return errFallbackToMetricNameMatch
+		item := ts.Item
+		if !bytes.HasPrefix(item, prefix) {
+			return nil
 		}
-		k := ts.Item
-		if !bytes.HasPrefix(k, tf.prefix) {
-			break
+		tail := item[len(prefix):]
+		n := bytes.IndexByte(tail, tagSeparatorChar)
+		if n < 0 {
+			return fmt.Errorf("invalid tag->metricIDs line %q: cannot find tagSeparatorChar=%d", item, tagSeparatorChar)
 		}
-
-		// Get MetricID from k (the last 8 bytes).
-		k = k[len(tf.prefix):]
-		if len(k) < 8 {
-			return fmt.Errorf("invald key suffix size; want at least %d bytes; got %d bytes", 8, len(k))
+		suffix := tail[:n+1]
+		tail = tail[n+1:]
+		if err := mp.InitOnlyTail(item, tail); err != nil {
+			return err
 		}
-		v := k[len(k)-8:]
-		k = k[:len(k)-8]
-		metricID := encoding.UnmarshalUint64(v)
-
-		if prevMatch && string(k) == string(prevMatchingK) {
+		if prevMatch && string(suffix) == string(prevMatchingSuffix) {
 			// Fast path: the same tag value found.
 			// There is no need in checking it again with potentially
 			// slow tf.matchSuffix, which may call regexp.
-			if !f(metricID) {
-				break
+			mp.ParseMetricIDs()
+			loops += len(mp.MetricIDs)
+			if loops > maxLoops {
+				return errFallbackToMetricNameMatch
+			}
+			for _, metricID := range mp.MetricIDs {
+				if !f(metricID) {
+					return nil
+				}
 			}
 			continue
 		}
 
-		ok, err := tf.matchSuffix(k)
+		// Slow path: need tf.matchSuffix call.
+		ok, err := tf.matchSuffix(suffix)
 		if err != nil {
-			return fmt.Errorf("error when matching %s: %s", tf, err)
+			return fmt.Errorf("error when matching %s against suffix %q: %s", tf, suffix, err)
 		}
 		if !ok {
 			prevMatch = false
 			// Optimization: skip all the metricIDs for the given tag value
-			kb.B = append(kb.B[:0], ts.Item[:len(ts.Item)-8]...)
+			kb.B = append(kb.B[:0], item[:len(item)-len(tail)]...)
 			// The last char in kb.B must be tagSeparatorChar. Just increment it
 			// in order to jump to the next tag value.
 			if len(kb.B) == 0 || kb.B[len(kb.B)-1] != tagSeparatorChar || tagSeparatorChar >= 0xff {
@@ -1704,13 +1688,20 @@ func (is *indexSearch) getMetricIDsForTagFilterSlow(tf *tagFilter, maxLoops int,
 			continue
 		}
 		prevMatch = true
-		prevMatchingK = append(prevMatchingK[:0], k...)
-		if !f(metricID) {
-			break
+		prevMatchingSuffix = append(prevMatchingSuffix[:0], suffix...)
+		mp.ParseMetricIDs()
+		loops += len(mp.MetricIDs)
+		if loops > maxLoops {
+			return errFallbackToMetricNameMatch
+		}
+		for _, metricID := range mp.MetricIDs {
+			if !f(metricID) {
+				return nil
+			}
 		}
 	}
 	if err := ts.Error(); err != nil {
-		return fmt.Errorf("error when searching for tag filter prefix %q: %s", tf.prefix, err)
+		return fmt.Errorf("error when searching for tag filter prefix %q: %s", prefix, err)
 	}
 	return nil
 }
@@ -1752,24 +1743,27 @@ func (is *indexSearch) updateMetricIDsForOrSuffixesWithFilter(tf *tagFilter, met
 
 func (is *indexSearch) updateMetricIDsForOrSuffixNoFilter(prefix []byte, maxMetrics int, metricIDs map[uint64]struct{}) error {
 	ts := &is.ts
+	mp := &is.mp
+	mp.Reset()
 	maxLoops := maxMetrics * maxIndexScanLoopsPerMetric
 	loops := 0
 	ts.Seek(prefix)
 	for len(metricIDs) < maxMetrics && ts.NextItem() {
-		loops++
+		item := ts.Item
+		if !bytes.HasPrefix(item, prefix) {
+			return nil
+		}
+		if err := mp.InitOnlyTail(item, item[len(prefix):]); err != nil {
+			return err
+		}
+		mp.ParseMetricIDs()
+		loops += len(mp.MetricIDs)
 		if loops > maxLoops {
 			return errFallbackToMetricNameMatch
 		}
-		if !bytes.HasPrefix(ts.Item, prefix) {
-			break
+		for _, metricID := range mp.MetricIDs {
+			metricIDs[metricID] = struct{}{}
 		}
-		// Get MetricID from ts.Item (the last 8 bytes).
-		v := ts.Item[len(prefix):]
-		if len(v) != 8 {
-			return fmt.Errorf("invalid key suffix size for prefix=%q; want %d bytes; got %d bytes; value=%q", 8, prefix, len(v), v)
-		}
-		metricID := encoding.UnmarshalUint64(v)
-		metricIDs[metricID] = struct{}{}
 	}
 	if err := ts.Error(); err != nil {
 		return fmt.Errorf("error when searching for tag filter prefix %q: %s", prefix, err)
@@ -1778,48 +1772,67 @@ func (is *indexSearch) updateMetricIDsForOrSuffixNoFilter(prefix []byte, maxMetr
 }
 
 func (is *indexSearch) updateMetricIDsForOrSuffixWithFilter(prefix []byte, metricIDs map[uint64]struct{}, sortedFilter []uint64, isNegative bool) error {
+	if len(sortedFilter) == 0 {
+		return nil
+	}
+	firstFilterMetricID := sortedFilter[0]
+	lastFilterMetricID := sortedFilter[len(sortedFilter)-1]
 	ts := &is.ts
-	kb := &is.kb
-	for {
-		// Seek for the next metricID from sortedFilter.
-		if len(sortedFilter) == 0 {
-			// All the sorteFilter entries have been searched.
-			break
+	mp := &is.mp
+	mp.Reset()
+	maxLoops := len(sortedFilter) * maxIndexScanLoopsPerMetric
+	loops := 0
+	ts.Seek(prefix)
+	var sf []uint64
+	var metricID uint64
+	for ts.NextItem() {
+		item := ts.Item
+		if !bytes.HasPrefix(item, prefix) {
+			return nil
 		}
-		nextMetricID := sortedFilter[0]
-		sortedFilter = sortedFilter[1:]
-		kb.B = append(kb.B[:0], prefix...)
-		kb.B = encoding.MarshalUint64(kb.B, nextMetricID)
-		ts.Seek(kb.B)
-		if !ts.NextItem() {
-			break
+		if err := mp.InitOnlyTail(item, item[len(prefix):]); err != nil {
+			return err
 		}
-		if !bytes.HasPrefix(ts.Item, prefix) {
-			break
+		firstMetricID, lastMetricID := mp.FirstAndLastMetricIDs()
+		if lastMetricID < firstFilterMetricID {
+			// Skip the item, since it contains metricIDs lower
+			// than metricIDs in sortedFilter.
+			continue
 		}
-		// Get MetricID from ts.Item (the last 8 bytes).
-		v := ts.Item[len(prefix):]
-		if len(v) != 8 {
-			return fmt.Errorf("invalid key suffix size for prefix=%q; want %d bytes; got %d bytes; value=%q", 8, prefix, len(v), v)
+		if firstMetricID > lastFilterMetricID {
+			// Stop searching, since the current item and all the subsequent items
+			// contain metricIDs higher than metricIDs in sortedFilter.
+			return nil
 		}
-		metricID := encoding.UnmarshalUint64(v)
-		if metricID != nextMetricID {
-			// Skip metricIDs smaller than the found metricID, since they don't
-			// match anything.
-			if len(sortedFilter) > 0 && metricID > sortedFilter[0] {
-				sortedFilter = sortedFilter[1:]
-				n := sort.Search(len(sortedFilter), func(i int) bool {
-					return metricID <= sortedFilter[i]
-				})
-				sortedFilter = sortedFilter[n:]
+		sf = sortedFilter
+		mp.ParseMetricIDs()
+		loops += len(mp.MetricIDs)
+		if loops > maxLoops {
+			return errFallbackToMetricNameMatch
+		}
+		for _, metricID = range mp.MetricIDs {
+			if len(sf) == 0 {
+				break
 			}
-			continue
+			if metricID > sf[0] {
+				n := sort.Search(len(sf), func(i int) bool {
+					return i >= 0 && i < len(sf) && sf[i] >= metricID
+				})
+				sf = sf[n:]
+				if len(sf) == 0 {
+					break
+				}
+			}
+			if metricID < sf[0] {
+				continue
+			}
+			if isNegative {
+				delete(metricIDs, metricID)
+			} else {
+				metricIDs[metricID] = struct{}{}
+			}
+			sf = sf[1:]
 		}
-		if isNegative {
-			delete(metricIDs, metricID)
-			continue
-		}
-		metricIDs[metricID] = struct{}{}
 	}
 	if err := ts.Error(); err != nil {
 		return fmt.Errorf("error when searching for tag filter prefix %q: %s", prefix, err)
@@ -2071,7 +2084,7 @@ func (is *indexSearch) updateMetricIDsAll(metricIDs map[uint64]struct{}, account
 // The maximum number of index scan loops per already found metric.
 // Bigger number of loops is slower than updateMetricIDsByMetricNameMatch
 // over the found metrics.
-const maxIndexScanLoopsPerMetric = 32
+const maxIndexScanLoopsPerMetric = 400
 
 func (is *indexSearch) intersectMetricIDsWithTagFilter(tf *tagFilter, filter map[uint64]struct{}) (map[uint64]struct{}, error) {
 	if len(filter) == 0 {
@@ -2084,7 +2097,10 @@ func (is *indexSearch) intersectMetricIDsWithTagFilter(tf *tagFilter, filter map
 	if len(tf.orSuffixes) > 0 {
 		// Fast path for orSuffixes - seek for rows for each value from orSuffixes.
 		if err := is.updateMetricIDsForOrSuffixesWithFilter(tf, metricIDs, filter); err != nil {
-			return nil, err
+			if err == errFallbackToMetricNameMatch {
+				return nil, err
+			}
+			return nil, fmt.Errorf("error when intersecting metricIDs for tagFilter in fast path: %s; tagFilter=%s", err, tf)
 		}
 		return metricIDs, nil
 	}
@@ -2103,7 +2119,10 @@ func (is *indexSearch) intersectMetricIDsWithTagFilter(tf *tagFilter, filter map
 		return true
 	})
 	if err != nil {
-		return nil, err
+		if err == errFallbackToMetricNameMatch {
+			return nil, err
+		}
+		return nil, fmt.Errorf("error when intersecting metricIDs for tagFilter in slow path: %s; tagFilter=%s", err, tf)
 	}
 	return metricIDs, nil
 }
@@ -2127,6 +2146,19 @@ func marshalCommonPrefix(dst []byte, nsPrefix byte, accountID, projectID uint32)
 	return dst
 }
 
+func unmarshalCommonPrefix(src []byte) ([]byte, byte, uint32, uint32, error) {
+	if len(src) < commonPrefixLen {
+		return nil, 0, 0, 0, fmt.Errorf("cannot unmarshal common prefix from %d bytes; need at least %d bytes; data=%X", len(src), commonPrefixLen, src)
+	}
+	prefix := src[0]
+	accountID := encoding.UnmarshalUint32(src[1:])
+	projectID := encoding.UnmarshalUint32(src[5:])
+	return src[commonPrefixLen:], prefix, accountID, projectID, nil
+}
+
+// 1 byte for prefix, 4 bytes for accountID, 4 bytes for projectID
+const commonPrefixLen = 9
+
 func getSortedMetricIDs(m map[uint64]struct{}) []uint64 {
 	a := make(uint64Sorter, len(m))
 	i := 0
@@ -2139,6 +2171,179 @@ func getSortedMetricIDs(m map[uint64]struct{}) []uint64 {
 	return a
 }
 
+type tagToMetricIDsRowParser struct {
+	// AccountID contains parsed value after Init call
+	AccountID uint32
+
+	// ProjectID contains parsed value after Init call
+	ProjectID uint32
+
+	// MetricIDs contains parsed MetricIDs after ParseMetricIDs call
+	MetricIDs []uint64
+
+	// Tag contains parsed tag after Init call
+	Tag Tag
+
+	// tail contains the remaining unparsed metricIDs
+	tail []byte
+}
+
+func (mp *tagToMetricIDsRowParser) Reset() {
+	mp.AccountID = 0
+	mp.ProjectID = 0
+	mp.MetricIDs = mp.MetricIDs[:0]
+	mp.Tag.Reset()
+	mp.tail = nil
+}
+
+// Init initializes mp from b, which should contain encoded tag->metricIDs row.
+//
+// b cannot be re-used until Reset call.
+func (mp *tagToMetricIDsRowParser) Init(b []byte) error {
+	tail, prefix, accountID, projectID, err := unmarshalCommonPrefix(b)
+	if err != nil {
+		return fmt.Errorf("invalid tag->metricIDs row %q: %s", b, err)
+	}
+	if prefix != nsPrefixTagToMetricIDs {
+		return fmt.Errorf("invalid prefix for tag->metricIDs row %q; got %d; want %d", b, prefix, nsPrefixTagToMetricIDs)
+	}
+	mp.AccountID = accountID
+	mp.ProjectID = projectID
+	tail, err = mp.Tag.Unmarshal(tail)
+	if err != nil {
+		return fmt.Errorf("cannot unmarshal tag from tag->metricIDs row %q: %s", b, err)
+	}
+	return mp.InitOnlyTail(b, tail)
+}
+
+// InitOnlyTail initializes mp.tail from tail.
+//
+// b must contain tag->metricIDs row.
+// b cannot be re-used until Reset call.
+func (mp *tagToMetricIDsRowParser) InitOnlyTail(b, tail []byte) error {
+	if len(tail) == 0 {
+		return fmt.Errorf("missing metricID in the tag->metricIDs row %q", b)
+	}
+	if len(tail)%8 != 0 {
+		return fmt.Errorf("invalid tail length in the tag->metricIDs row; got %d bytes; must be multiple of 8 bytes", len(tail))
+	}
+	mp.tail = tail
+	return nil
+}
+
+// EqualPrefix returns true if prefixes for mp and x are equal.
+//
+// Prefix contains (tag, accountID, projectID)
+func (mp *tagToMetricIDsRowParser) EqualPrefix(x *tagToMetricIDsRowParser) bool {
+	if !mp.Tag.Equal(&x.Tag) {
+		return false
+	}
+	return mp.ProjectID == x.ProjectID && mp.AccountID == x.AccountID
+}
+
+// FirstAndLastMetricIDs returns the first and the last metricIDs in the mp.tail.
+func (mp *tagToMetricIDsRowParser) FirstAndLastMetricIDs() (uint64, uint64) {
+	tail := mp.tail
+	if len(tail) < 8 {
+		logger.Panicf("BUG: cannot unmarshal metricID from %d bytes; need 8 bytes", len(tail))
+		return 0, 0
+	}
+	firstMetricID := encoding.UnmarshalUint64(tail)
+	lastMetricID := firstMetricID
+	if len(tail) > 8 {
+		lastMetricID = encoding.UnmarshalUint64(tail[len(tail)-8:])
+	}
+	return firstMetricID, lastMetricID
+}
+
+// ParseMetricIDs parses MetricIDs from mp.tail into mp.MetricIDs.
+func (mp *tagToMetricIDsRowParser) ParseMetricIDs() {
+	tail := mp.tail
+	mp.MetricIDs = mp.MetricIDs[:0]
+	n := len(tail) / 8
+	if n <= cap(mp.MetricIDs) {
+		mp.MetricIDs = mp.MetricIDs[:n]
+	} else {
+		mp.MetricIDs = append(mp.MetricIDs[:cap(mp.MetricIDs)], make([]uint64, n-cap(mp.MetricIDs))...)
+	}
+	metricIDs := mp.MetricIDs
+	_ = metricIDs[n-1]
+	for i := 0; i < n; i++ {
+		if len(tail) < 8 {
+			logger.Panicf("BUG: tail cannot be smaller than 8 bytes; got %d bytes; tail=%X", len(tail), tail)
+			return
+		}
+		metricID := encoding.UnmarshalUint64(tail)
+		metricIDs[i] = metricID
+		tail = tail[8:]
+	}
+}
+
+// IsDeletedTag verifies whether the tag from mp is deleted according to dmis.
+//
+// dmis must contain deleted MetricIDs.
+func (mp *tagToMetricIDsRowParser) IsDeletedTag(dmis map[uint64]struct{}) bool {
+	if len(dmis) == 0 {
+		return false
+	}
+	mp.ParseMetricIDs()
+	for _, metricID := range mp.MetricIDs {
+		if _, ok := dmis[metricID]; !ok {
+			return false
+		}
+	}
+	return true
+}
+
+func mergeTagToMetricIDsRows(data []byte, items [][]byte) ([]byte, [][]byte) {
+	// Perform quick checks whether items contain tag->metricIDs rows
+	// based on the fact that items are sorted.
+	if len(items) == 0 {
+		return data, items
+	}
+	firstItem := items[0]
+	if len(firstItem) > 0 && firstItem[0] > nsPrefixTagToMetricIDs {
+		return data, items
+	}
+	lastItem := items[len(items)-1]
+	if len(lastItem) > 0 && lastItem[0] < nsPrefixTagToMetricIDs {
+		return data, items
+	}
+
+	// items contain at least one tag->metricIDs row. Merge rows with common tag.
+	dstData := data[:0]
+	dstItems := items[:0]
+
+	tmm := getTagToMetricIDsRowsMerger()
+	defer putTagToMetricIDsRowsMerger(tmm)
+
+	mp := &tmm.mp
+	mpPrev := &tmm.mpPrev
+	for _, item := range items {
+		if len(item) == 0 || item[0] != nsPrefixTagToMetricIDs {
+			if len(tmm.pendingMetricIDs) > 0 {
+				dstData, dstItems = tmm.flushPendingMetricIDs(dstData, dstItems, mpPrev)
+			}
+			dstData = append(dstData, item...)
+			dstItems = append(dstItems, dstData[len(dstData)-len(item):])
+			continue
+		}
+		if err := mp.Init(item); err != nil {
+			logger.Panicf("FATAL: cannot parse tag->metricIDs row during merge: %s", err)
+		}
+		if len(tmm.pendingMetricIDs) > 0 && !mp.EqualPrefix(mpPrev) {
+			dstData, dstItems = tmm.flushPendingMetricIDs(dstData, dstItems, mpPrev)
+		}
+		mp.ParseMetricIDs()
+		tmm.pendingMetricIDs = append(tmm.pendingMetricIDs, mp.MetricIDs...)
+		mpPrev, mp = mp, mpPrev
+	}
+	if len(tmm.pendingMetricIDs) > 0 {
+		dstData, dstItems = tmm.flushPendingMetricIDs(dstData, dstItems, mpPrev)
+	}
+	return dstData, dstItems
+}
+
 type uint64Sorter []uint64
 
 func (s uint64Sorter) Len() int { return len(s) }
@@ -2148,3 +2353,43 @@ func (s uint64Sorter) Less(i, j int) bool {
 func (s uint64Sorter) Swap(i, j int) {
 	s[i], s[j] = s[j], s[i]
 }
+
+type tagToMetricIDsRowsMerger struct {
+	pendingMetricIDs uint64Sorter
+	mp               tagToMetricIDsRowParser
+	mpPrev           tagToMetricIDsRowParser
+}
+
+func (tmm *tagToMetricIDsRowsMerger) flushPendingMetricIDs(dstData []byte, dstItems [][]byte, mp *tagToMetricIDsRowParser) ([]byte, [][]byte) {
+	if len(tmm.pendingMetricIDs) == 0 {
+		logger.Panicf("BUG: pendingMetricIDs must be non-empty")
+	}
+	dstDataLen := len(dstData)
+	dstData = marshalCommonPrefix(dstData, nsPrefixTagToMetricIDs, mp.AccountID, mp.ProjectID)
+	dstData = mp.Tag.Marshal(dstData)
+	// Use sort.Sort instead of sort.Slice in order to reduce memory allocations
+	sort.Sort(&tmm.pendingMetricIDs)
+	for _, metricID := range tmm.pendingMetricIDs {
+		dstData = encoding.MarshalUint64(dstData, metricID)
+	}
+	tmm.pendingMetricIDs = tmm.pendingMetricIDs[:0]
+	dstItems = append(dstItems, dstData[dstDataLen:])
+	return dstData, dstItems
+}
+
+func getTagToMetricIDsRowsMerger() *tagToMetricIDsRowsMerger {
+	v := tmmPool.Get()
+	if v == nil {
+		return &tagToMetricIDsRowsMerger{}
+	}
+	return v.(*tagToMetricIDsRowsMerger)
+}
+
+func putTagToMetricIDsRowsMerger(tmm *tagToMetricIDsRowsMerger) {
+	tmm.pendingMetricIDs = tmm.pendingMetricIDs[:0]
+	tmm.mp.Reset()
+	tmm.mpPrev.Reset()
+	tmmPool.Put(tmm)
+}
+
+var tmmPool sync.Pool
diff --git a/lib/storage/metric_name.go b/lib/storage/metric_name.go
index b416aff80d..bc24a6ec63 100644
--- a/lib/storage/metric_name.go
+++ b/lib/storage/metric_name.go
@@ -25,6 +25,17 @@ type Tag struct {
 	Value []byte
 }
 
+// Reset resets the tag.
+func (tag *Tag) Reset() {
+	tag.Key = tag.Key[:0]
+	tag.Value = tag.Value[:0]
+}
+
+// Equal returns true if tag equals t
+func (tag *Tag) Equal(t *Tag) bool {
+	return string(tag.Key) == string(t.Key) && string(tag.Value) == string(t.Value)
+}
+
 // Marshal appends marshaled tag to dst and returns the result.
 func (tag *Tag) Marshal(dst []byte) []byte {
 	dst = marshalTagValue(dst, tag.Key)
diff --git a/lib/storage/partition.go b/lib/storage/partition.go
index 3f77782a37..b9fe8f6cbf 100644
--- a/lib/storage/partition.go
+++ b/lib/storage/partition.go
@@ -881,7 +881,7 @@ func (pt *partition) mergeSmallParts(isFinal bool) error {
 	maxRows := maxRowsByPath(pt.smallPartsPath)
 	if maxRows > maxRowsPerSmallPart() {
 		// The output part may go to big part,
-		// so make sure it as enough space.
+		// so make sure it has enough space.
 		maxBigPartRows := maxRowsByPath(pt.bigPartsPath)
 		if maxRows > maxBigPartRows {
 			maxRows = maxBigPartRows
diff --git a/lib/storage/tag_filters.go b/lib/storage/tag_filters.go
index d576cfe439..da03186481 100644
--- a/lib/storage/tag_filters.go
+++ b/lib/storage/tag_filters.go
@@ -23,7 +23,7 @@ type TagFilters struct {
 	tfs []tagFilter
 
 	// Common prefix for all the tag filters.
-	// Contains encoded nsPrefixTagToMetricID + accountID + projectID
+	// Contains encoded nsPrefixTagToMetricIDs + accountID + projectID.
 	commonPrefix []byte
 }
 
@@ -32,7 +32,7 @@ func NewTagFilters(accountID, projectID uint32) *TagFilters {
 	return &TagFilters{
 		accountID:    accountID,
 		projectID:    projectID,
-		commonPrefix: marshalCommonPrefix(nil, nsPrefixTagToMetricID, accountID, projectID),
+		commonPrefix: marshalCommonPrefix(nil, nsPrefixTagToMetricIDs, accountID, projectID),
 	}
 }
 
@@ -87,7 +87,7 @@ func (tfs *TagFilters) Reset(accountID, projectID uint32) {
 	tfs.accountID = accountID
 	tfs.projectID = projectID
 	tfs.tfs = tfs.tfs[:0]
-	tfs.commonPrefix = marshalCommonPrefix(tfs.commonPrefix[:0], nsPrefixTagToMetricID, accountID, projectID)
+	tfs.commonPrefix = marshalCommonPrefix(tfs.commonPrefix[:0], nsPrefixTagToMetricIDs, accountID, projectID)
 }
 
 func (tfs *TagFilters) marshal(dst []byte) []byte {
@@ -106,7 +106,7 @@ type tagFilter struct {
 	isNegative bool
 	isRegexp   bool
 
-	// Prefix always contains {nsPrefixTagToMetricID, AccountID, ProjectID, key}.
+	// Prefix always contains {nsPrefixTagToMetricIDs, AccountID, ProjectID, key}.
 	// Additionally it contains:
 	//  - value ending with tagSeparatorChar if !isRegexp.
 	//  - non-regexp prefix if isRegexp.