diff --git a/core/state/metrics.go b/core/state/metrics.go
index 086666b7db..e702ef3a81 100644
--- a/core/state/metrics.go
+++ b/core/state/metrics.go
@@ -19,10 +19,12 @@ package state
 import "github.com/ethereum/go-ethereum/metrics"
 
 var (
-	accountUpdatedMeter    = metrics.NewRegisteredMeter("state/update/account", nil)
-	storageUpdatedMeter    = metrics.NewRegisteredMeter("state/update/storage", nil)
-	accountDeletedMeter    = metrics.NewRegisteredMeter("state/delete/account", nil)
-	storageDeletedMeter    = metrics.NewRegisteredMeter("state/delete/storage", nil)
-	accountTrieNodesMeter  = metrics.NewRegisteredMeter("state/trie/account", nil)
-	storageTriesNodesMeter = metrics.NewRegisteredMeter("state/trie/storage", nil)
+	accountUpdatedMeter      = metrics.NewRegisteredMeter("state/update/account", nil)
+	storageUpdatedMeter      = metrics.NewRegisteredMeter("state/update/storage", nil)
+	accountDeletedMeter      = metrics.NewRegisteredMeter("state/delete/account", nil)
+	storageDeletedMeter      = metrics.NewRegisteredMeter("state/delete/storage", nil)
+	accountTrieUpdatedMeter  = metrics.NewRegisteredMeter("state/update/accountnodes", nil)
+	storageTriesUpdatedMeter = metrics.NewRegisteredMeter("state/update/storagenodes", nil)
+	accountTrieDeletedMeter  = metrics.NewRegisteredMeter("state/delete/accountnodes", nil)
+	storageTriesDeletedMeter = metrics.NewRegisteredMeter("state/delete/storagenodes", nil)
 )
diff --git a/core/state/statedb.go b/core/state/statedb.go
index 4385c19769..3f4bec2392 100644
--- a/core/state/statedb.go
+++ b/core/state/statedb.go
@@ -970,11 +970,13 @@ func (s *StateDB) Commit(deleteEmptyObjects bool) (common.Hash, error) {
 
 	// Commit objects to the trie, measuring the elapsed time
 	var (
-		accountTrieNodes int
-		storageTrieNodes int
-		nodes            = trie.NewMergedNodeSet()
-		codeWriter       = s.db.DiskDB().NewBatch()
+		accountTrieNodesUpdated int
+		accountTrieNodesDeleted int
+		storageTrieNodesUpdated int
+		storageTrieNodesDeleted int
+		nodes                   = trie.NewMergedNodeSet()
 	)
+	codeWriter := s.db.DiskDB().NewBatch()
 	for addr := range s.stateObjectsDirty {
 		if obj := s.stateObjects[addr]; !obj.deleted {
 			// Write any contract code associated with the state object
@@ -992,9 +994,17 @@ func (s *StateDB) Commit(deleteEmptyObjects bool) (common.Hash, error) {
 				if err := nodes.Merge(set); err != nil {
 					return common.Hash{}, err
 				}
-				storageTrieNodes += set.Size()
+				updates, deleted := set.Size()
+				storageTrieNodesUpdated += updates
+				storageTrieNodesDeleted += deleted
 			}
 		}
+		// If the contract is destructed, the storage is still left in the
+		// database as dangling data. Theoretically it's should be wiped from
+		// database as well, but in hash-based-scheme it's extremely hard to
+		// determine that if the trie nodes are also referenced by other storage,
+		// and in path-based-scheme some technical challenges are still unsolved.
+		// Although it won't affect the correctness but please fix it TODO(rjl493456442).
 	}
 	if len(s.stateObjectsDirty) > 0 {
 		s.stateObjectsDirty = make(map[common.Address]struct{})
@@ -1015,7 +1025,7 @@ func (s *StateDB) Commit(deleteEmptyObjects bool) (common.Hash, error) {
 		if err := nodes.Merge(set); err != nil {
 			return common.Hash{}, err
 		}
-		accountTrieNodes = set.Size()
+		accountTrieNodesUpdated, accountTrieNodesDeleted = set.Size()
 	}
 	if metrics.EnabledExpensive {
 		s.AccountCommits += time.Since(start)
@@ -1024,9 +1034,10 @@ func (s *StateDB) Commit(deleteEmptyObjects bool) (common.Hash, error) {
 		storageUpdatedMeter.Mark(int64(s.StorageUpdated))
 		accountDeletedMeter.Mark(int64(s.AccountDeleted))
 		storageDeletedMeter.Mark(int64(s.StorageDeleted))
-		accountTrieNodesMeter.Mark(int64(accountTrieNodes))
-		storageTriesNodesMeter.Mark(int64(storageTrieNodes))
-
+		accountTrieUpdatedMeter.Mark(int64(accountTrieNodesUpdated))
+		accountTrieDeletedMeter.Mark(int64(accountTrieNodesDeleted))
+		storageTriesUpdatedMeter.Mark(int64(storageTrieNodesUpdated))
+		storageTriesDeletedMeter.Mark(int64(storageTrieNodesDeleted))
 		s.AccountUpdated, s.AccountDeleted = 0, 0
 		s.StorageUpdated, s.StorageDeleted = 0, 0
 	}
diff --git a/trie/committer.go b/trie/committer.go
index 745cb76839..c4957f3490 100644
--- a/trie/committer.go
+++ b/trie/committer.go
@@ -33,22 +33,29 @@ type leaf struct {
 // insertion order.
 type committer struct {
 	nodes       *NodeSet
+	tracer      *tracer
 	collectLeaf bool
 }
 
 // newCommitter creates a new committer or picks one from the pool.
-func newCommitter(nodes *NodeSet, collectLeaf bool) *committer {
+func newCommitter(owner common.Hash, tracer *tracer, collectLeaf bool) *committer {
 	return &committer{
-		nodes:       nodes,
+		nodes:       NewNodeSet(owner),
+		tracer:      tracer,
 		collectLeaf: collectLeaf,
 	}
 }
 
 // Commit collapses a node down into a hash node and returns it along with
 // the modified nodeset.
-func (c *committer) Commit(n node) hashNode {
+func (c *committer) Commit(n node) (hashNode, *NodeSet) {
 	h := c.commit(nil, n)
-	return h.(hashNode)
+	// Some nodes can be deleted from trie which can't be captured
+	// by committer itself. Iterate all deleted nodes tracked by
+	// tracer and marked them as deleted only if they are present
+	// in database previously.
+	c.tracer.markDeletions(c.nodes)
+	return h.(hashNode), c.nodes
 }
 
 // commit collapses a node down into a hash node and returns it.
@@ -78,6 +85,12 @@ func (c *committer) commit(path []byte, n node) node {
 		if hn, ok := hashedNode.(hashNode); ok {
 			return hn
 		}
+		// The short node now is embedded in its parent. Mark the node as
+		// deleted if it's present in database previously. It's equivalent
+		// as deletion from database's perspective.
+		if prev := c.tracer.getPrev(path); len(prev) != 0 {
+			c.nodes.markDeleted(path, prev)
+		}
 		return collapsed
 	case *fullNode:
 		hashedKids := c.commitChildren(path, cn)
@@ -88,6 +101,12 @@ func (c *committer) commit(path []byte, n node) node {
 		if hn, ok := hashedNode.(hashNode); ok {
 			return hn
 		}
+		// The full node now is embedded in its parent. Mark the node as
+		// deleted if it's present in database previously. It's equivalent
+		// as deletion from database's perspective.
+		if prev := c.tracer.getPrev(path); len(prev) != 0 {
+			c.nodes.markDeleted(path, prev)
+		}
 		return collapsed
 	case hashNode:
 		return cn
@@ -150,7 +169,7 @@ func (c *committer) store(path []byte, n node) node {
 		}
 	)
 	// Collect the dirty node to nodeset for return.
-	c.nodes.markUpdated(path, mnode)
+	c.nodes.markUpdated(path, mnode, c.tracer.getPrev(path))
 
 	// Collect the corresponding leaf node if it's required. We don't check
 	// full node since it's impossible to store value in fullNode. The key
diff --git a/trie/database.go b/trie/database.go
index 5d47475e0f..74247d59c4 100644
--- a/trie/database.go
+++ b/trie/database.go
@@ -792,12 +792,13 @@ func (db *Database) Update(nodes *MergedNodeSet) error {
 	}
 	for _, owner := range order {
 		subset := nodes.sets[owner]
-		subset.forEachWithOrder(false, func(path string, n *memoryNode) {
-			if n.isDeleted() {
-				return // ignore deletion
+		for _, path := range subset.updates.order {
+			n, ok := subset.updates.nodes[path]
+			if !ok {
+				return fmt.Errorf("missing node %x %v", owner, path)
 			}
 			db.insert(n.hash, int(n.size), n.node)
-		})
+		}
 	}
 	// Link up the account trie and storage trie if the node points
 	// to an account trie leaf.
diff --git a/trie/nodeset.go b/trie/nodeset.go
index 4251eccaf6..9281723501 100644
--- a/trie/nodeset.go
+++ b/trie/nodeset.go
@@ -19,7 +19,6 @@ package trie
 import (
 	"fmt"
 	"reflect"
-	"sort"
 	"strings"
 
 	"github.com/ethereum/go-ethereum/common"
@@ -41,8 +40,8 @@ var memoryNodeSize = int(reflect.TypeOf(memoryNode{}).Size())
 
 // memorySize returns the total memory size used by this node.
 // nolint:unused
-func (n *memoryNode) memorySize(pathlen int) int {
-	return int(n.size) + memoryNodeSize + pathlen
+func (n *memoryNode) memorySize(key int) int {
+	return int(n.size) + memoryNodeSize + key
 }
 
 // rlp returns the raw rlp encoded blob of the cached trie node, either directly
@@ -65,20 +64,14 @@ func (n *memoryNode) obj() node {
 	return expandNode(n.hash[:], n.node)
 }
 
-// isDeleted returns the indicator if the node is marked as deleted.
-func (n *memoryNode) isDeleted() bool {
-	return n.hash == (common.Hash{})
-}
-
 // nodeWithPrev wraps the memoryNode with the previous node value.
-// nolint: unused
 type nodeWithPrev struct {
 	*memoryNode
 	prev []byte // RLP-encoded previous value, nil means it's non-existent
 }
 
 // unwrap returns the internal memoryNode object.
-// nolint: unused
+// nolint:unused
 func (n *nodeWithPrev) unwrap() *memoryNode {
 	return n.memoryNode
 }
@@ -86,69 +79,64 @@ func (n *nodeWithPrev) unwrap() *memoryNode {
 // memorySize returns the total memory size used by this node. It overloads
 // the function in memoryNode by counting the size of previous value as well.
 // nolint: unused
-func (n *nodeWithPrev) memorySize(pathlen int) int {
-	return n.memoryNode.memorySize(pathlen) + len(n.prev)
+func (n *nodeWithPrev) memorySize(key int) int {
+	return n.memoryNode.memorySize(key) + len(n.prev)
+}
+
+// nodesWithOrder represents a collection of dirty nodes which includes
+// newly-inserted and updated nodes. The modification order of all nodes
+// is represented by order list.
+type nodesWithOrder struct {
+	order []string                 // the path list of dirty nodes, sort by insertion order
+	nodes map[string]*nodeWithPrev // the map of dirty nodes, keyed by node path
 }
 
 // NodeSet contains all dirty nodes collected during the commit operation.
 // Each node is keyed by path. It's not thread-safe to use.
 type NodeSet struct {
-	owner      common.Hash            // the identifier of the trie
-	nodes      map[string]*memoryNode // the set of dirty nodes(inserted, updated, deleted)
-	leaves     []*leaf                // the list of dirty leaves
-	accessList map[string][]byte      // The list of accessed nodes, which records the original node value
+	owner   common.Hash       // the identifier of the trie
+	updates *nodesWithOrder   // the set of updated nodes(newly inserted, updated)
+	deletes map[string][]byte // the map of deleted nodes, keyed by node
+	leaves  []*leaf           // the list of dirty leaves
 }
 
 // NewNodeSet initializes an empty node set to be used for tracking dirty nodes
-// for a specific account or storage trie. The owner is zero for the account trie
-// and the owning account address hash for storage tries. The provided accessList
-// represents the original value of accessed nodes, it can be optional but would
-// be beneficial for speeding up the construction of trie history.
-func NewNodeSet(owner common.Hash, accessList map[string][]byte) *NodeSet {
-	// Don't panic for lazy users
-	if accessList == nil {
-		accessList = make(map[string][]byte)
-	}
+// from a specific account or storage trie. The owner is zero for the account
+// trie and the owning account address hash for storage tries.
+func NewNodeSet(owner common.Hash) *NodeSet {
 	return &NodeSet{
-		owner:      owner,
-		nodes:      make(map[string]*memoryNode),
-		accessList: accessList,
+		owner: owner,
+		updates: &nodesWithOrder{
+			nodes: make(map[string]*nodeWithPrev),
+		},
+		deletes: make(map[string][]byte),
 	}
 }
 
-// forEachWithOrder iterates the dirty nodes with the specified order.
-// If topToBottom is true:
-//
-//	then the order of iteration is top to bottom, left to right.
-//
-// If topToBottom is false:
-//
-//	then the order of iteration is bottom to top, right to left.
-func (set *NodeSet) forEachWithOrder(topToBottom bool, callback func(path string, n *memoryNode)) {
-	var paths sort.StringSlice
-	for path := range set.nodes {
-		paths = append(paths, path)
-	}
-	if topToBottom {
-		paths.Sort()
-	} else {
-		sort.Sort(sort.Reverse(paths))
-	}
-	for _, path := range paths {
-		callback(path, set.nodes[path])
+/*
+// NewNodeSetWithDeletion initializes the nodeset with provided deletion set.
+func NewNodeSetWithDeletion(owner common.Hash, paths [][]byte, prev [][]byte) *NodeSet {
+	set := NewNodeSet(owner)
+	for i, path := range paths {
+		set.markDeleted(path, prev[i])
 	}
+	return set
 }
+*/
 
 // markUpdated marks the node as dirty(newly-inserted or updated) with provided
 // node path, node object along with its previous value.
-func (set *NodeSet) markUpdated(path []byte, node *memoryNode) {
-	set.nodes[string(path)] = node
+func (set *NodeSet) markUpdated(path []byte, node *memoryNode, prev []byte) {
+	set.updates.order = append(set.updates.order, string(path))
+	set.updates.nodes[string(path)] = &nodeWithPrev{
+		memoryNode: node,
+		prev:       prev,
+	}
 }
 
 // markDeleted marks the node as deleted with provided path and previous value.
-// nolint: unused
-func (set *NodeSet) markDeleted(path []byte) {
-	set.nodes[string(path)] = &memoryNode{}
+func (set *NodeSet) markDeleted(path []byte, prev []byte) {
+	set.deletes[string(path)] = prev
 }
 
 // addLeaf collects the provided leaf node into set.
@@ -156,16 +144,16 @@ func (set *NodeSet) addLeaf(node *leaf) {
 	set.leaves = append(set.leaves, node)
 }
 
-// Size returns the number of dirty nodes contained in the set.
-func (set *NodeSet) Size() int {
-	return len(set.nodes)
+// Size returns the number of updated and deleted nodes contained in the set.
+func (set *NodeSet) Size() (int, int) {
+	return len(set.updates.order), len(set.deletes)
 }
 
 // Hashes returns the hashes of all updated nodes. TODO(rjl493456442) how can
 // we get rid of it?
 func (set *NodeSet) Hashes() []common.Hash {
 	var ret []common.Hash
-	for _, node := range set.nodes {
+	for _, node := range set.updates.nodes {
 		ret = append(ret, node.hash)
 	}
 	return ret
@@ -175,17 +163,19 @@ func (set *NodeSet) Hashes() []common.Hash {
 func (set *NodeSet) Summary() string {
 	var out = new(strings.Builder)
 	fmt.Fprintf(out, "nodeset owner: %v\n", set.owner)
-	if set.nodes != nil {
-		for path, n := range set.nodes {
-			// Deletion
-			if n.isDeleted() {
-				fmt.Fprintf(out, "  [-]: %x\n", path)
-				continue
+	if set.updates != nil {
+		for _, key := range set.updates.order {
+			updated := set.updates.nodes[key]
+			if updated.prev != nil {
+				fmt.Fprintf(out, "  [*]: %x -> %v prev: %x\n", key, updated.hash, updated.prev)
+			} else {
+				fmt.Fprintf(out, "  [+]: %x -> %v\n", key, updated.hash)
 			}
-			// Update
-			fmt.Fprintf(out, "  [+]: %x -> %v\n", path, n.hash)
 		}
 	}
+	for k, n := range set.deletes {
+		fmt.Fprintf(out, "  [-]: %x -> %x\n", k, n)
+	}
 	for _, n := range set.leaves {
 		fmt.Fprintf(out, "[leaf]: %v\n", n)
 	}
diff --git a/trie/trie.go b/trie/trie.go
index 65f89d91f8..c467ac4766 100644
--- a/trie/trie.go
+++ b/trie/trie.go
@@ -51,11 +51,12 @@ type Trie struct {
 	// actually unhashed nodes.
 	unhashed int
 
-	// accessList tracks the loaded nodes from database.
-	accessList map[string][]byte
-
 	// reader is the handler trie can retrieve nodes from.
 	reader *trieReader
+
+	// tracer is the tool to track the trie changes.
+	// It will be reset after each commit operation.
+	tracer *tracer
 }
 
 // newFlag returns the cache flag value for a newly created node.
@@ -65,16 +66,12 @@ func (t *Trie) newFlag() nodeFlag {
 
 // Copy returns a copy of Trie.
 func (t *Trie) Copy() *Trie {
-	accessList := make(map[string][]byte)
-	for path, blob := range t.accessList {
-		accessList[path] = common.CopyBytes(blob)
-	}
 	return &Trie{
-		root:       t.root,
-		owner:      t.owner,
-		unhashed:   t.unhashed,
-		reader:     t.reader,
-		accessList: accessList,
+		root:     t.root,
+		owner:    t.owner,
+		unhashed: t.unhashed,
+		reader:   t.reader,
+		tracer:   t.tracer.copy(),
 	}
 }
 
@@ -90,9 +87,9 @@ func New(id *ID, db NodeReader) (*Trie, error) {
 		return nil, err
 	}
 	trie := &Trie{
-		owner:      id.Owner,
-		reader:     reader,
-		accessList: make(map[string][]byte),
+		owner:  id.Owner,
+		reader: reader,
+		//tracer: newTracer(),
 	}
 	if id.Root != (common.Hash{}) && id.Root != emptyRoot {
 		rootnode, err := trie.resolveAndTrack(id.Root[:], nil)
@@ -329,6 +326,11 @@ func (t *Trie) insert(n node, prefix, key []byte, value node) (bool, node, error
 		if matchlen == 0 {
 			return true, branch, nil
 		}
+		// New branch node is created as a child of the original short node.
+		// Track the newly inserted node in the tracer. The node identifier
+		// passed is the path from the root node.
+		t.tracer.onInsert(append(prefix, key[:matchlen]...))
+
 		// Replace it with a short node leading up to the branch.
 		return true, &shortNode{key[:matchlen], branch, t.newFlag()}, nil
 
@@ -343,6 +345,11 @@ func (t *Trie) insert(n node, prefix, key []byte, value node) (bool, node, error
 		return true, n, nil
 
 	case nil:
+		// New short node is created and track it in the tracer. The node identifier
+		// passed is the path from the root node. Note the valueNode won't be tracked
+		// since it's always embedded in its parent.
+		t.tracer.onInsert(prefix)
+
 		return true, &shortNode{key, value, t.newFlag()}, nil
 
 	case hashNode:
@@ -395,6 +402,11 @@ func (t *Trie) delete(n node, prefix, key []byte) (bool, node, error) {
 			return false, n, nil // don't replace n on mismatch
 		}
 		if matchlen == len(key) {
+			// The matched short node is deleted entirely and track
+			// it in the deletion set. The same the valueNode doesn't
+			// need to be tracked at all since it's always embedded.
+			t.tracer.onDelete(prefix)
+
 			return true, nil, nil // remove n entirely for whole matches
 		}
 		// The key is longer than n.Key. Remove the remaining suffix
@@ -407,6 +419,10 @@ func (t *Trie) delete(n node, prefix, key []byte) (bool, node, error) {
 		}
 		switch child := child.(type) {
 		case *shortNode:
+			// The child shortNode is merged into its parent, track
+			// is deleted as well.
+			t.tracer.onDelete(append(prefix, n.Key...))
+
 			// Deleting from the subtrie reduced it to another
 			// short node. Merge the nodes to avoid creating a
 			// shortNode{..., shortNode{...}}. Use concat (which
@@ -468,6 +484,11 @@ func (t *Trie) delete(n node, prefix, key []byte) (bool, node, error) {
 					return false, nil, err
 				}
 				if cnode, ok := cnode.(*shortNode); ok {
+					// Replace the entire full node with the short node.
+					// Mark the original short node as deleted since the
+					// value is embedded into the parent now.
+					t.tracer.onDelete(append(prefix, byte(pos)))
+
 					k := append([]byte{byte(pos)}, cnode.Key...)
 					return true, &shortNode{k, cnode.Val, t.newFlag()}, nil
 				}
@@ -527,7 +548,7 @@ func (t *Trie) resolveAndTrack(n hashNode, prefix []byte) (node, error) {
 	if err != nil {
 		return nil, err
 	}
-	t.accessList[string(prefix)] = blob
+	t.tracer.onRead(prefix, blob)
 	return mustDecodeNode(n, blob), nil
 }
 
@@ -546,15 +567,16 @@ func (t *Trie) Hash() common.Hash {
 // Once the trie is committed, it's not usable anymore. A new trie must
 // be created with new root and updated trie database for following usage
 func (t *Trie) Commit(collectLeaf bool) (common.Hash, *NodeSet) {
-	// Reset accessList at the end of commit
-	defer func() {
-		t.accessList = make(map[string][]byte)
-	}()
+	defer t.tracer.reset()
+
 	// Trie is empty and can be classified into two types of situations:
 	// - The trie was empty and no update happens
 	// - The trie was non-empty and all nodes are dropped
 	if t.root == nil {
-		return emptyRoot, nil
+		// Wrap tracked deletions as the return
+		set := NewNodeSet(t.owner)
+		t.tracer.markDeletions(set)
+		return emptyRoot, set
 	}
 	// Derive the hash for all dirty nodes first. We hold the assumption
 	// in the following procedure that all nodes are hashed.
@@ -568,9 +590,8 @@ func (t *Trie) Commit(collectLeaf bool) (common.Hash, *NodeSet) {
 		t.root = hashedNode
 		return rootHash, nil
 	}
-	nodes := NewNodeSet(t.owner, t.accessList)
-	h := newCommitter(nodes, collectLeaf)
-	newRoot := h.Commit(t.root)
+	h := newCommitter(t.owner, t.tracer, collectLeaf)
+	newRoot, nodes := h.Commit(t.root)
 	t.root = newRoot
 	return rootHash, nodes
 }
@@ -593,5 +614,5 @@ func (t *Trie) Reset() {
 	t.root = nil
 	t.owner = common.Hash{}
 	t.unhashed = 0
-	t.accessList = make(map[string][]byte)
+	t.tracer.reset()
 }
diff --git a/trie/trie_test.go b/trie/trie_test.go
index b87b7d3cd6..2fb97eebbf 100644
--- a/trie/trie_test.go
+++ b/trie/trie_test.go
@@ -410,6 +410,7 @@ func runRandTest(rt randTest) bool {
 		values   = make(map[string]string) // tracks content of the trie
 		origTrie = NewEmpty(triedb)
 	)
+	tr.tracer = newTracer()
 
 	for i, step := range rt {
 		// fmt.Printf("{op: %d, key: common.Hex2Bytes(\"%x\"), value: common.Hex2Bytes(\"%x\")}, // step %d\n",
@@ -448,14 +449,21 @@ func runRandTest(rt randTest) bool {
 			root, nodes := tr.Commit(true)
 			// Validity the returned nodeset
 			if nodes != nil {
-				for path := range nodes.nodes {
+				for path, node := range nodes.updates.nodes {
 					blob, _, _ := origTrie.TryGetNode(hexToCompact([]byte(path)))
-					got := nodes.accessList[path]
+					got := node.prev
 					if !bytes.Equal(blob, got) {
 						rt[i].err = fmt.Errorf("prevalue mismatch for 0x%x, got 0x%x want 0x%x", path, got, blob)
 						panic(rt[i].err)
 					}
 				}
+				for path, prev := range nodes.deletes {
+					blob, _, _ := origTrie.TryGetNode(hexToCompact([]byte(path)))
+					if !bytes.Equal(blob, prev) {
+						rt[i].err = fmt.Errorf("prevalue mismatch for 0x%x, got 0x%x want 0x%x", path, prev, blob)
+						return false
+					}
+				}
 			}
 			if nodes != nil {
 				triedb.Update(NewWithNodeSet(nodes))
@@ -469,6 +477,7 @@ func runRandTest(rt randTest) bool {
 
 			// Enable node tracing. Resolve the root node again explicitly
 			// since it's not captured at the beginning.
+			tr.tracer = newTracer()
 			tr.resolveAndTrack(root.Bytes(), nil)
 
 			origTrie = tr.Copy()
@@ -481,6 +490,59 @@ func runRandTest(rt randTest) bool {
 			if tr.Hash() != checktr.Hash() {
 				rt[i].err = fmt.Errorf("hash mismatch in opItercheckhash")
 			}
+		case opNodeDiff:
+			var (
+				inserted = tr.tracer.insertList()
+				deleted  = tr.tracer.deleteList()
+				origIter = origTrie.NodeIterator(nil)
+				curIter  = tr.NodeIterator(nil)
+				origSeen = make(map[string]struct{})
+				curSeen  = make(map[string]struct{})
+			)
+			for origIter.Next(true) {
+				if origIter.Leaf() {
+					continue
+				}
+				origSeen[string(origIter.Path())] = struct{}{}
+			}
+			for curIter.Next(true) {
+				if curIter.Leaf() {
+					continue
+				}
+				curSeen[string(curIter.Path())] = struct{}{}
+			}
+			var (
+				insertExp = make(map[string]struct{})
+				deleteExp = make(map[string]struct{})
+			)
+			for path := range curSeen {
+				_, present := origSeen[path]
+				if !present {
+					insertExp[path] = struct{}{}
+				}
+			}
+			for path := range origSeen {
+				_, present := curSeen[path]
+				if !present {
+					deleteExp[path] = struct{}{}
+				}
+			}
+			if len(insertExp) != len(inserted) {
+				rt[i].err = fmt.Errorf("insert set mismatch")
+			}
+			if len(deleteExp) != len(deleted) {
+				rt[i].err = fmt.Errorf("delete set mismatch")
+			}
+			for _, insert := range inserted {
+				if _, present := insertExp[string(insert)]; !present {
+					rt[i].err = fmt.Errorf("missing inserted node")
+				}
+			}
+			for _, del := range deleted {
+				if _, present := deleteExp[string(del)]; !present {
+					rt[i].err = fmt.Errorf("missing deleted node")
+				}
+			}
 		}
 		// Abort the test on error.
 		if rt[i].err != nil {
diff --git a/trie/util_test.go b/trie/util_test.go
new file mode 100644
index 0000000000..01a46553aa
--- /dev/null
+++ b/trie/util_test.go
@@ -0,0 +1,304 @@
+// Copyright 2022 The go-ethereum Authors
+// This file is part of the go-ethereum library.
+//
+// The go-ethereum library is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Lesser General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// The go-ethereum library is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Lesser General Public License for more details.
+//
+// You should have received a copy of the GNU Lesser General Public License
+// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
+
+package trie
+
+import (
+	"bytes"
+	"testing"
+
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/core/rawdb"
+)
+
+// Tests if the trie diffs are tracked correctly.
+func TestTrieTracer(t *testing.T) {
+	db := NewDatabase(rawdb.NewMemoryDatabase())
+	trie := NewEmpty(db)
+	trie.tracer = newTracer()
+
+	// Insert a batch of entries, all the nodes should be marked as inserted
+	vals := []struct{ k, v string }{
+		{"do", "verb"},
+		{"ether", "wookiedoo"},
+		{"horse", "stallion"},
+		{"shaman", "horse"},
+		{"doge", "coin"},
+		{"dog", "puppy"},
+		{"somethingveryoddindeedthis is", "myothernodedata"},
+	}
+	for _, val := range vals {
+		trie.Update([]byte(val.k), []byte(val.v))
+	}
+	trie.Hash()
+
+	seen := make(map[string]struct{})
+	it := trie.NodeIterator(nil)
+	for it.Next(true) {
+		if it.Leaf() {
+			continue
+		}
+		seen[string(it.Path())] = struct{}{}
+	}
+	inserted := trie.tracer.insertList()
+	if len(inserted) != len(seen) {
+		t.Fatalf("Unexpected inserted node tracked want %d got %d", len(seen), len(inserted))
+	}
+	for _, k := range inserted {
+		_, ok := seen[string(k)]
+		if !ok {
+			t.Fatalf("Unexpected inserted node")
+		}
+	}
+	deleted := trie.tracer.deleteList()
+	if len(deleted) != 0 {
+		t.Fatalf("Unexpected deleted node tracked %d", len(deleted))
+	}
+
+	// Commit the changes and re-create with new root
+	root, nodes := trie.Commit(false)
+	if err := db.Update(NewWithNodeSet(nodes)); err != nil {
+		t.Fatal(err)
+	}
+	trie, _ = New(TrieID(root), db)
+	trie.tracer = newTracer()
+
+	// Delete all the elements, check deletion set
+	for _, val := range vals {
+		trie.Delete([]byte(val.k))
+	}
+	trie.Hash()
+
+	inserted = trie.tracer.insertList()
+	if len(inserted) != 0 {
+		t.Fatalf("Unexpected inserted node tracked %d", len(inserted))
+	}
+	deleted = trie.tracer.deleteList()
+	if len(deleted) != len(seen) {
+		t.Fatalf("Unexpected deleted node tracked want %d got %d", len(seen), len(deleted))
+	}
+	for _, k := range deleted {
+		_, ok := seen[string(k)]
+		if !ok {
+			t.Fatalf("Unexpected inserted node")
+		}
+	}
+}
+
+func TestTrieTracerNoop(t *testing.T) {
+	trie := NewEmpty(NewDatabase(rawdb.NewMemoryDatabase()))
+	trie.tracer = newTracer()
+
+	// Insert a batch of entries, all the nodes should be marked as inserted
+	vals := []struct{ k, v string }{
+		{"do", "verb"},
+		{"ether", "wookiedoo"},
+		{"horse", "stallion"},
+		{"shaman", "horse"},
+		{"doge", "coin"},
+		{"dog", "puppy"},
+		{"somethingveryoddindeedthis is", "myothernodedata"},
+	}
+	for _, val := range vals {
+		trie.Update([]byte(val.k), []byte(val.v))
+	}
+	for _, val := range vals {
+		trie.Delete([]byte(val.k))
+	}
+	if len(trie.tracer.insertList()) != 0 {
+		t.Fatalf("Unexpected inserted node tracked %d", len(trie.tracer.insertList()))
+	}
+	if len(trie.tracer.deleteList()) != 0 {
+		t.Fatalf("Unexpected deleted node tracked %d", len(trie.tracer.deleteList()))
+	}
+}
+
+func TestTrieTracePrevValue(t *testing.T) {
+	db := NewDatabase(rawdb.NewMemoryDatabase())
+	trie := NewEmpty(db)
+	trie.tracer = newTracer()
+
+	paths, blobs := trie.tracer.prevList()
+	if len(paths) != 0 || len(blobs) != 0 {
+		t.Fatalf("Nothing should be tracked")
+	}
+	// Insert a batch of entries, all the nodes should be marked as inserted
+	vals := []struct{ k, v string }{
+		{"do", "verb"},
+		{"ether", "wookiedoo"},
+		{"horse", "stallion"},
+		{"shaman", "horse"},
+		{"doge", "coin"},
+		{"dog", "puppy"},
+		{"somethingveryoddindeedthis is", "myothernodedata"},
+	}
+	for _, val := range vals {
+		trie.Update([]byte(val.k), []byte(val.v))
+	}
+	paths, blobs = trie.tracer.prevList()
+	if len(paths) != 0 || len(blobs) != 0 {
+		t.Fatalf("Nothing should be tracked")
+	}
+
+	// Commit the changes and re-create with new root
+	root, nodes := trie.Commit(false)
+	if err := db.Update(NewWithNodeSet(nodes)); err != nil {
+		t.Fatal(err)
+	}
+	trie, _ = New(TrieID(root), db)
+	trie.tracer = newTracer()
+	trie.resolveAndTrack(root.Bytes(), nil)
+
+	// Load all nodes in trie
+	for _, val := range vals {
+		trie.TryGet([]byte(val.k))
+	}
+
+	// Ensure all nodes are tracked by tracer with correct prev-values
+	iter := trie.NodeIterator(nil)
+	seen := make(map[string][]byte)
+	for iter.Next(true) {
+		// Embedded nodes are ignored since they are not present in
+		// database.
+		if iter.Hash() == (common.Hash{}) {
+			continue
+		}
+		seen[string(iter.Path())] = common.CopyBytes(iter.NodeBlob())
+	}
+
+	paths, blobs = trie.tracer.prevList()
+	if len(paths) != len(seen) || len(blobs) != len(seen) {
+		t.Fatalf("Unexpected tracked values")
+	}
+	for i, path := range paths {
+		blob := blobs[i]
+		prev, ok := seen[string(path)]
+		if !ok {
+			t.Fatalf("Missing node %v", path)
+		}
+		if !bytes.Equal(blob, prev) {
+			t.Fatalf("Unexpected value path: %v, want: %v, got: %v", path, prev, blob)
+		}
+	}
+
+	// Re-open the trie and iterate the trie, ensure nothing will be tracked.
+	// Iterator will not link any loaded nodes to trie.
+	trie, _ = New(TrieID(root), db)
+	trie.tracer = newTracer()
+
+	iter = trie.NodeIterator(nil)
+	for iter.Next(true) {
+	}
+	paths, blobs = trie.tracer.prevList()
+	if len(paths) != 0 || len(blobs) != 0 {
+		t.Fatalf("Nothing should be tracked")
+	}
+
+	// Re-open the trie and generate proof for entries, ensure nothing will
+	// be tracked. Prover will not link any loaded nodes to trie.
+	trie, _ = New(TrieID(root), db)
+	trie.tracer = newTracer()
+	for _, val := range vals {
+		trie.Prove([]byte(val.k), 0, rawdb.NewMemoryDatabase())
+	}
+	paths, blobs = trie.tracer.prevList()
+	if len(paths) != 0 || len(blobs) != 0 {
+		t.Fatalf("Nothing should be tracked")
+	}
+
+	// Delete entries from trie, ensure all previous values are correct.
+	trie, _ = New(TrieID(root), db)
+	trie.tracer = newTracer()
+	trie.resolveAndTrack(root.Bytes(), nil)
+
+	for _, val := range vals {
+		trie.TryDelete([]byte(val.k))
+	}
+	paths, blobs = trie.tracer.prevList()
+	if len(paths) != len(seen) || len(blobs) != len(seen) {
+		t.Fatalf("Unexpected tracked values")
+	}
+	for i, path := range paths {
+		blob := blobs[i]
+		prev, ok := seen[string(path)]
+		if !ok {
+			t.Fatalf("Missing node %v", path)
+		}
+		if !bytes.Equal(blob, prev) {
+			t.Fatalf("Unexpected value path: %v, want: %v, got: %v", path, prev, blob)
+		}
+	}
+}
+
+func TestDeleteAll(t *testing.T) {
+	db := NewDatabase(rawdb.NewMemoryDatabase())
+	trie := NewEmpty(db)
+	trie.tracer = newTracer()
+
+	// Insert a batch of entries, all the nodes should be marked as inserted
+	vals := []struct{ k, v string }{
+		{"do", "verb"},
+		{"ether", "wookiedoo"},
+		{"horse", "stallion"},
+		{"shaman", "horse"},
+		{"doge", "coin"},
+		{"dog", "puppy"},
+		{"somethingveryoddindeedthis is", "myothernodedata"},
+	}
+	for _, val := range vals {
+		trie.Update([]byte(val.k), []byte(val.v))
+	}
+	root, set := trie.Commit(false)
+	if err := db.Update(NewWithNodeSet(set)); err != nil {
+		t.Fatal(err)
+	}
+	// Delete entries from trie, ensure all values are detected
+	trie, _ = New(TrieID(root), db)
+	trie.tracer = newTracer()
+	trie.resolveAndTrack(root.Bytes(), nil)
+
+	// Iterate all existent nodes
+	var (
+		it    = trie.NodeIterator(nil)
+		nodes = make(map[string][]byte)
+	)
+	for it.Next(true) {
+		if it.Hash() != (common.Hash{}) {
+			nodes[string(it.Path())] = common.CopyBytes(it.NodeBlob())
+		}
+	}
+
+	// Perform deletion to purge the entire trie
+	for _, val := range vals {
+		trie.Delete([]byte(val.k))
+	}
+	root, set = trie.Commit(false)
+	if root != emptyRoot {
+		t.Fatalf("Invalid trie root %v", root)
+	}
+	for path, blob := range set.deletes {
+		prev, ok := nodes[path]
+		if !ok {
+			t.Fatalf("Extra node deleted %v", []byte(path))
+		}
+		if !bytes.Equal(prev, blob) {
+			t.Fatalf("Unexpected previous value %v", []byte(path))
+		}
+	}
+	if len(set.deletes) != len(nodes) {
+		t.Fatalf("Unexpected deletion set")
+	}
+}
diff --git a/trie/utils.go b/trie/utils.go
new file mode 100644
index 0000000000..5dce65cd29
--- /dev/null
+++ b/trie/utils.go
@@ -0,0 +1,199 @@
+// Copyright 2022 The go-ethereum Authors
+// This file is part of the go-ethereum library.
+//
+// The go-ethereum library is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Lesser General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// The go-ethereum library is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Lesser General Public License for more details.
+//
+// You should have received a copy of the GNU Lesser General Public License
+// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
+
+package trie
+
+// tracer tracks the changes of trie nodes. During the trie operations,
+// some nodes can be deleted from the trie, while these deleted nodes
+// won't be captured by trie.Hasher or trie.Committer. Thus, these deleted
+// nodes won't be removed from the disk at all. Tracer is an auxiliary tool
+// used to track all insert and delete operations of trie and capture all
+// deleted nodes eventually.
+//
+// The changed nodes can be mainly divided into two categories: the leaf
+// node and intermediate node. The former is inserted/deleted by callers
+// while the latter is inserted/deleted in order to follow the rule of trie.
+// This tool can track all of them no matter the node is embedded in its
+// parent or not, but valueNode is never tracked.
+//
+// Besides, it's also used for recording the original value of the nodes
+// when they are resolved from the disk. The pre-value of the nodes will
+// be used to construct reverse-diffs in the future.
+//
+// Note tracer is not thread-safe, callers should be responsible for handling
+// the concurrency issues by themselves.
+type tracer struct {
+	insert map[string]struct{}
+	delete map[string]struct{}
+	origin map[string][]byte
+}
+
+// newTracer initializes the tracer for capturing trie changes.
+func newTracer() *tracer {
+	return &tracer{
+		insert: make(map[string]struct{}),
+		delete: make(map[string]struct{}),
+		origin: make(map[string][]byte),
+	}
+}
+
+// onRead tracks the newly loaded trie node and caches the rlp-encoded blob internally.
+// Don't change the value outside of function since it's not deep-copied.
+func (t *tracer) onRead(path []byte, val []byte) {
+	// Tracer isn't used right now, remove this check later.
+	if t == nil {
+		return
+	}
+	t.origin[string(path)] = val
+}
+
+// onInsert tracks the newly inserted trie node. If it's already in the deletion set
+// (resurrected node), then just wipe it from the deletion set as the "untouched".
+func (t *tracer) onInsert(path []byte) {
+	// Tracer isn't used right now, remove this check later.
+	if t == nil {
+		return
+	}
+	if _, present := t.delete[string(path)]; present {
+		delete(t.delete, string(path))
+		return
+	}
+	t.insert[string(path)] = struct{}{}
+}
+
+// onDelete tracks the newly deleted trie node. If it's already
+// in the addition set, then just wipe it from the addition set
+// as it's untouched.
+func (t *tracer) onDelete(path []byte) {
+	// Tracer isn't used right now, remove this check later.
+	if t == nil {
+		return
+	}
+	if _, present := t.insert[string(path)]; present {
+		delete(t.insert, string(path))
+		return
+	}
+	t.delete[string(path)] = struct{}{}
+}
+
+// insertList returns the tracked inserted trie nodes in list format.
+func (t *tracer) insertList() [][]byte {
+	// Tracer isn't used right now, remove this check later.
+	if t == nil {
+		return nil
+	}
+	var ret [][]byte
+	for path := range t.insert {
+		ret = append(ret, []byte(path))
+	}
+	return ret
+}
+
+// deleteList returns the tracked deleted trie nodes in list format.
+func (t *tracer) deleteList() [][]byte {
+	// Tracer isn't used right now, remove this check later.
+	if t == nil {
+		return nil
+	}
+	var ret [][]byte
+	for path := range t.delete {
+		ret = append(ret, []byte(path))
+	}
+	return ret
+}
+
+// prevList returns the tracked node blobs in list format.
+func (t *tracer) prevList() ([][]byte, [][]byte) {
+	// Tracer isn't used right now, remove this check later.
+	if t == nil {
+		return nil, nil
+	}
+	var (
+		paths [][]byte
+		blobs [][]byte
+	)
+	for path, blob := range t.origin {
+		paths = append(paths, []byte(path))
+		blobs = append(blobs, blob)
+	}
+	return paths, blobs
+}
+
+// getPrev returns the cached original value of the specified node.
+func (t *tracer) getPrev(path []byte) []byte {
+	// Tracer isn't used right now, remove this check later.
+	if t == nil {
+		return nil
+	}
+	return t.origin[string(path)]
+}
+
+// reset clears the content tracked by tracer.
+func (t *tracer) reset() {
+	// Tracer isn't used right now, remove this check later.
+	if t == nil {
+		return
+	}
+	t.insert = make(map[string]struct{})
+	t.delete = make(map[string]struct{})
+	t.origin = make(map[string][]byte)
+}
+
+// copy returns a deep copied tracer instance.
+func (t *tracer) copy() *tracer {
+	// Tracer isn't used right now, remove this check later.
+	if t == nil {
+		return nil
+	}
+	var (
+		insert = make(map[string]struct{})
+		delete = make(map[string]struct{})
+		origin = make(map[string][]byte)
+	)
+	for key := range t.insert {
+		insert[key] = struct{}{}
+	}
+	for key := range t.delete {
+		delete[key] = struct{}{}
+	}
+	for key, val := range t.origin {
+		origin[key] = val
+	}
+	return &tracer{
+		insert: insert,
+		delete: delete,
+		origin: origin,
+	}
+}
+
+// markDeletions puts all tracked deletions into the provided nodeset.
+func (t *tracer) markDeletions(set *NodeSet) {
+	// Tracer isn't used right now, remove this check later.
+	if t == nil {
+		return
+	}
+	for _, path := range t.deleteList() {
+		// There are a few possibilities for this scenario(the node is deleted
+		// but not present in database previously), for example the node was
+		// embedded in the parent and now deleted from the trie. In this case
+		// it's noop from database's perspective.
+		val := t.getPrev(path)
+		if len(val) == 0 {
+			continue
+		}
+		set.markDeleted(path, val)
+	}
+}