diff --git a/pkg/local_object_storage/metabase/VERSION.md b/pkg/local_object_storage/metabase/VERSION.md
index 0affc36c9b..c9707c0072 100644
--- a/pkg/local_object_storage/metabase/VERSION.md
+++ b/pkg/local_object_storage/metabase/VERSION.md
@@ -84,6 +84,15 @@ This file describes changes between the metabase versions.
   - Key: split ID
   - Value: list of object IDs
 
+# History
+
+## Version 2
+
+- Container ID is encoded as 32-byte slice
+- Object ID is encoded as 32-byte slice
+- Object ID is encoded as 64-byte slice, container ID + object ID
+- Bucket naming scheme is changed:
+  - container ID + suffix -> 1-byte prefix + container ID
 
 ## Version 1
 
diff --git a/pkg/local_object_storage/metabase/containers.go b/pkg/local_object_storage/metabase/containers.go
index da7cb0f2b7..fb1b487900 100644
--- a/pkg/local_object_storage/metabase/containers.go
+++ b/pkg/local_object_storage/metabase/containers.go
@@ -1,9 +1,7 @@
 package meta
 
 import (
-	"crypto/sha256"
 	"encoding/binary"
-	"strings"
 
 	cid "github.com/nspcc-dev/neofs-sdk-go/container/id"
 	"go.etcd.io/bbolt"
@@ -30,7 +28,7 @@ func (db *DB) containers(tx *bbolt.Tx) ([]cid.ID, error) {
 	err := tx.ForEach(func(name []byte, _ *bbolt.Bucket) error {
 		if parseContainerID(&cnr, name, unique) {
 			result = append(result, cnr)
-			unique[cnr.EncodeToString()] = struct{}{}
+			unique[string(name[1:bucketKeySize])] = struct{}{}
 		}
 
 		return nil
@@ -55,24 +53,20 @@ func (db *DB) containerSize(tx *bbolt.Tx, id cid.ID) (uint64, error) {
 		return 0, err
 	}
 
-	key := make([]byte, sha256.Size)
+	key := make([]byte, cidSize)
 	id.Encode(key)
 
 	return parseContainerSize(containerVolume.Get(key)), nil
 }
 
 func parseContainerID(dst *cid.ID, name []byte, ignore map[string]struct{}) bool {
-	strContainerID := strings.Split(string(name), invalidBase58String)[0]
-
-	if _, ok := ignore[strContainerID]; ok {
+	if len(name) != bucketKeySize {
 		return false
 	}
-
-	if err := dst.DecodeString(strContainerID); err != nil {
+	if _, ok := ignore[string(name[1:bucketKeySize])]; ok {
 		return false
 	}
-
-	return true
+	return dst.Decode(name[1:bucketKeySize]) == nil
 }
 
 func parseContainerSize(v []byte) uint64 {
@@ -89,7 +83,7 @@ func changeContainerSize(tx *bbolt.Tx, id cid.ID, delta uint64, increase bool) e
 		return err
 	}
 
-	key := make([]byte, sha256.Size)
+	key := make([]byte, cidSize)
 	id.Encode(key)
 
 	size := parseContainerSize(containerVolume.Get(key))
diff --git a/pkg/local_object_storage/metabase/counter.go b/pkg/local_object_storage/metabase/counter.go
index 0ab1d98057..5085c3ee09 100644
--- a/pkg/local_object_storage/metabase/counter.go
+++ b/pkg/local_object_storage/metabase/counter.go
@@ -125,6 +125,7 @@ func syncCounter(tx *bbolt.Tx, force bool) error {
 
 	graveyardBKT := tx.Bucket(graveyardBucketName)
 	garbageBKT := tx.Bucket(garbageBucketName)
+	key := make([]byte, addressKeySize)
 
 	err = iteratePhyObjects(tx, func(cnr cid.ID, obj oid.ID) error {
 		phyCounter++
@@ -134,7 +135,7 @@ func syncCounter(tx *bbolt.Tx, force bool) error {
 
 		// check if an object is available: not with GCMark
 		// and not covered with a tombstone
-		if inGraveyardWithKey(addressKey(addr), graveyardBKT, garbageBKT) == 0 {
+		if inGraveyardWithKey(addressKey(addr, key), graveyardBKT, garbageBKT) == 0 {
 			logicCounter++
 		}
 
diff --git a/pkg/local_object_storage/metabase/db.go b/pkg/local_object_storage/metabase/db.go
index a05edf6cbb..29ba942c0e 100644
--- a/pkg/local_object_storage/metabase/db.go
+++ b/pkg/local_object_storage/metabase/db.go
@@ -11,6 +11,7 @@ import (
 	"sync"
 	"time"
 
+	"github.com/mr-tron/base58"
 	v2object "github.com/nspcc-dev/neofs-api-go/v2/object"
 	"github.com/nspcc-dev/neofs-node/pkg/local_object_storage/shard/mode"
 	"github.com/nspcc-dev/neofs-node/pkg/util/logger"
@@ -111,6 +112,8 @@ func stringifyValue(key string, objVal []byte) string {
 	switch key {
 	default:
 		return string(objVal)
+	case v2object.FilterHeaderObjectID, v2object.FilterHeaderContainerID, v2object.FilterHeaderParent:
+		return base58.Encode(objVal)
 	case v2object.FilterHeaderPayloadHash, v2object.FilterHeaderHomomorphicHash:
 		return hex.EncodeToString(objVal)
 	case v2object.FilterHeaderCreationEpoch, v2object.FilterHeaderPayloadLength:
@@ -140,6 +143,9 @@ func destringifyValue(key, value string, prefix bool) ([]byte, bool, bool) {
 	switch key {
 	default:
 		return []byte(value), false, true
+	case v2object.FilterHeaderObjectID, v2object.FilterHeaderContainerID, v2object.FilterHeaderParent:
+		v, err := base58.Decode(value)
+		return v, false, err == nil
 	case v2object.FilterHeaderPayloadHash, v2object.FilterHeaderHomomorphicHash:
 		v, err := hex.DecodeString(value)
 		if err != nil {
@@ -254,6 +260,12 @@ func unknownMatcherBucket(_ *bbolt.Bucket, _ string, _ string, _ func([]byte, []
 // in boltDB. Useful for getting filter values from unique and list indexes.
 func bucketKeyHelper(hdr string, val string) []byte {
 	switch hdr {
+	case v2object.FilterHeaderParent:
+		v, err := base58.Decode(val)
+		if err != nil {
+			return nil
+		}
+		return v
 	case v2object.FilterHeaderPayloadHash:
 		v, err := hex.DecodeString(val)
 		if err != nil {
diff --git a/pkg/local_object_storage/metabase/delete.go b/pkg/local_object_storage/metabase/delete.go
index 01398153b1..30c8049000 100644
--- a/pkg/local_object_storage/metabase/delete.go
+++ b/pkg/local_object_storage/metabase/delete.go
@@ -139,7 +139,8 @@ func (db *DB) deleteGroup(tx *bbolt.Tx, addrs []oid.Address) (uint64, uint64, er
 // object was available before the removal (for calculating the logical object
 // counter).
 func (db *DB) delete(tx *bbolt.Tx, addr oid.Address, refCounter referenceCounter, currEpoch uint64) (bool, bool, error) {
-	addrKey := addressKey(addr)
+	key := make([]byte, addressKeySize)
+	addrKey := addressKey(addr, key)
 	garbageBKT := tx.Bucket(garbageBucketName)
 	graveyardBKT := tx.Bucket(graveyardBucketName)
 
@@ -154,7 +155,7 @@ func (db *DB) delete(tx *bbolt.Tx, addr oid.Address, refCounter referenceCounter
 	}
 
 	// unmarshal object, work only with physically stored (raw == true) objects
-	obj, err := db.get(tx, addr, false, true, currEpoch)
+	obj, err := db.get(tx, addr, key, false, true, currEpoch)
 	if err != nil {
 		if errors.As(err, new(apistatus.ObjectNotFound)) {
 			return false, false, nil
@@ -166,9 +167,10 @@ func (db *DB) delete(tx *bbolt.Tx, addr oid.Address, refCounter referenceCounter
 	// if object is an only link to a parent, then remove parent
 	if parent := obj.Parent(); parent != nil {
 		parAddr := object.AddressOf(parent)
-		sParAddr := parAddr.EncodeToString()
+		sParAddr := addressKey(parAddr, key)
+		k := string(sParAddr)
 
-		nRef, ok := refCounter[sParAddr]
+		nRef, ok := refCounter[k]
 		if !ok {
 			nRef = &referenceNumber{
 				all:  parentLength(tx, parAddr),
@@ -176,7 +178,7 @@ func (db *DB) delete(tx *bbolt.Tx, addr oid.Address, refCounter referenceCounter
 				obj:  parent,
 			}
 
-			refCounter[sParAddr] = nRef
+			refCounter[k] = nRef
 		}
 
 		nRef.cur++
@@ -216,12 +218,14 @@ func (db *DB) deleteObject(
 
 // parentLength returns amount of available children from parentid index.
 func parentLength(tx *bbolt.Tx, addr oid.Address) int {
-	bkt := tx.Bucket(parentBucketName(addr.Container()))
+	bucketName := make([]byte, bucketKeySize)
+
+	bkt := tx.Bucket(parentBucketName(addr.Container(), bucketName[:]))
 	if bkt == nil {
 		return 0
 	}
 
-	lst, err := decodeList(bkt.Get(objectKey(addr.Object())))
+	lst, err := decodeList(bkt.Get(objectKey(addr.Object(), bucketName[:])))
 	if err != nil {
 		return 0
 	}
@@ -291,23 +295,22 @@ func delListIndexItem(tx *bbolt.Tx, item namedBucketItem) error {
 func delUniqueIndexes(tx *bbolt.Tx, obj *objectSDK.Object, isParent bool) error {
 	addr := object.AddressOf(obj)
 
-	objKey := objectKey(addr.Object())
-	addrKey := addressKey(addr)
+	objKey := objectKey(addr.Object(), make([]byte, objectKeySize))
+	addrKey := addressKey(addr, make([]byte, addressKeySize))
 	cnr := addr.Container()
+	bucketName := make([]byte, bucketKeySize)
 
 	// add value to primary unique bucket
 	if !isParent {
-		var bucketName []byte
-
 		switch obj.Type() {
 		case objectSDK.TypeRegular:
-			bucketName = primaryBucketName(cnr)
+			bucketName = primaryBucketName(cnr, bucketName)
 		case objectSDK.TypeTombstone:
-			bucketName = tombstoneBucketName(cnr)
+			bucketName = tombstoneBucketName(cnr, bucketName)
 		case objectSDK.TypeStorageGroup:
-			bucketName = storageGroupBucketName(cnr)
+			bucketName = storageGroupBucketName(cnr, bucketName)
 		case objectSDK.TypeLock:
-			bucketName = bucketNameLockers(cnr)
+			bucketName = bucketNameLockers(cnr, bucketName)
 		default:
 			return ErrUnknownObjectType
 		}
@@ -318,17 +321,17 @@ func delUniqueIndexes(tx *bbolt.Tx, obj *objectSDK.Object, isParent bool) error
 		})
 	} else {
 		delUniqueIndexItem(tx, namedBucketItem{
-			name: parentBucketName(cnr),
+			name: parentBucketName(cnr, bucketName),
 			key:  objKey,
 		})
 	}
 
 	delUniqueIndexItem(tx, namedBucketItem{ // remove from storage id index
-		name: smallBucketName(cnr),
+		name: smallBucketName(cnr, bucketName),
 		key:  objKey,
 	})
 	delUniqueIndexItem(tx, namedBucketItem{ // remove from root index
-		name: rootBucketName(cnr),
+		name: rootBucketName(cnr, bucketName),
 		key:  objKey,
 	})
 	delUniqueIndexItem(tx, namedBucketItem{ // remove from ToMoveIt index
diff --git a/pkg/local_object_storage/metabase/exists.go b/pkg/local_object_storage/metabase/exists.go
index 00965d7583..34da4d2060 100644
--- a/pkg/local_object_storage/metabase/exists.go
+++ b/pkg/local_object_storage/metabase/exists.go
@@ -71,17 +71,18 @@ func (db *DB) exists(tx *bbolt.Tx, addr oid.Address, currEpoch uint64) (exists b
 		return false, object.ErrObjectIsExpired
 	}
 
-	objKey := objectKey(addr.Object())
+	objKey := objectKey(addr.Object(), make([]byte, objectKeySize))
 
 	cnr := addr.Container()
+	key := make([]byte, bucketKeySize)
 
 	// if graveyard is empty, then check if object exists in primary bucket
-	if inBucket(tx, primaryBucketName(cnr), objKey) {
+	if inBucket(tx, primaryBucketName(cnr, key), objKey) {
 		return true, nil
 	}
 
 	// if primary bucket is empty, then check if object exists in parent bucket
-	if inBucket(tx, parentBucketName(cnr), objKey) {
+	if inBucket(tx, parentBucketName(cnr, key), objKey) {
 		splitInfo, err := getSplitInfo(tx, cnr, objKey)
 		if err != nil {
 			return false, err
@@ -105,15 +106,16 @@ func objectStatus(tx *bbolt.Tx, addr oid.Address, currEpoch uint64) uint8 {
 	// GC is expected to collect all the objects that have
 	// expired previously for less than the one epoch duration
 
-	rawOID := []byte(addr.Object().EncodeToString())
 	var expired bool
 
 	// bucket with objects that have expiration attr
-	expirationBucket := tx.Bucket(attributeBucketName(addr.Container(), objectV2.SysAttributeExpEpoch))
+	attrKey := make([]byte, bucketKeySize+len(objectV2.SysAttributeExpEpoch))
+	expirationBucket := tx.Bucket(attributeBucketName(addr.Container(), objectV2.SysAttributeExpEpoch, attrKey))
 	if expirationBucket != nil {
 		// bucket that contains objects that expire in the current epoch
 		prevEpochBkt := expirationBucket.Bucket([]byte(strconv.FormatUint(currEpoch-1, 10)))
 		if prevEpochBkt != nil {
+			rawOID := objectKey(addr.Object(), make([]byte, objectKeySize))
 			if prevEpochBkt.Get(rawOID) != nil {
 				expired = true
 			}
@@ -126,7 +128,7 @@ func objectStatus(tx *bbolt.Tx, addr oid.Address, currEpoch uint64) uint8 {
 
 	graveyardBkt := tx.Bucket(graveyardBucketName)
 	garbageBkt := tx.Bucket(garbageBucketName)
-	addrKey := addressKey(addr)
+	addrKey := addressKey(addr, make([]byte, addressKeySize))
 	return inGraveyardWithKey(addrKey, graveyardBkt, garbageBkt)
 }
 
@@ -175,7 +177,8 @@ func inBucket(tx *bbolt.Tx, name, key []byte) bool {
 // getSplitInfo returns SplitInfo structure from root index. Returns error
 // if there is no `key` record in root index.
 func getSplitInfo(tx *bbolt.Tx, cnr cid.ID, key []byte) (*objectSDK.SplitInfo, error) {
-	rawSplitInfo := getFromBucket(tx, rootBucketName(cnr), key)
+	bucketName := rootBucketName(cnr, make([]byte, bucketKeySize))
+	rawSplitInfo := getFromBucket(tx, bucketName, key)
 	if len(rawSplitInfo) == 0 {
 		return nil, ErrLackSplitInfo
 	}
diff --git a/pkg/local_object_storage/metabase/get.go b/pkg/local_object_storage/metabase/get.go
index 2b67823dc2..1a34b398ff 100644
--- a/pkg/local_object_storage/metabase/get.go
+++ b/pkg/local_object_storage/metabase/get.go
@@ -53,7 +53,8 @@ func (db *DB) Get(prm GetPrm) (res GetRes, err error) {
 	currEpoch := db.epochState.CurrentEpoch()
 
 	err = db.boltDB.View(func(tx *bbolt.Tx) error {
-		res.hdr, err = db.get(tx, prm.addr, true, prm.raw, currEpoch)
+		key := make([]byte, addressKeySize)
+		res.hdr, err = db.get(tx, prm.addr, key, true, prm.raw, currEpoch)
 
 		return err
 	})
@@ -61,9 +62,7 @@ func (db *DB) Get(prm GetPrm) (res GetRes, err error) {
 	return
 }
 
-func (db *DB) get(tx *bbolt.Tx, addr oid.Address, checkStatus, raw bool, currEpoch uint64) (*objectSDK.Object, error) {
-	key := objectKey(addr.Object())
-
+func (db *DB) get(tx *bbolt.Tx, addr oid.Address, key []byte, checkStatus, raw bool, currEpoch uint64) (*objectSDK.Object, error) {
 	if checkStatus {
 		switch objectStatus(tx, addr, currEpoch) {
 		case 1:
@@ -79,29 +78,31 @@ func (db *DB) get(tx *bbolt.Tx, addr oid.Address, checkStatus, raw bool, currEpo
 		}
 	}
 
+	key = objectKey(addr.Object(), key)
 	cnr := addr.Container()
 	obj := objectSDK.New()
+	bucketName := make([]byte, bucketKeySize)
 
 	// check in primary index
-	data := getFromBucket(tx, primaryBucketName(cnr), key)
+	data := getFromBucket(tx, primaryBucketName(cnr, bucketName), key)
 	if len(data) != 0 {
 		return obj, obj.Unmarshal(data)
 	}
 
 	// if not found then check in tombstone index
-	data = getFromBucket(tx, tombstoneBucketName(cnr), key)
+	data = getFromBucket(tx, tombstoneBucketName(cnr, bucketName), key)
 	if len(data) != 0 {
 		return obj, obj.Unmarshal(data)
 	}
 
 	// if not found then check in storage group index
-	data = getFromBucket(tx, storageGroupBucketName(cnr), key)
+	data = getFromBucket(tx, storageGroupBucketName(cnr, bucketName), key)
 	if len(data) != 0 {
 		return obj, obj.Unmarshal(data)
 	}
 
 	// if not found then check in locker index
-	data = getFromBucket(tx, bucketNameLockers(cnr), key)
+	data = getFromBucket(tx, bucketNameLockers(cnr, bucketName), key)
 	if len(data) != 0 {
 		return obj, obj.Unmarshal(data)
 	}
@@ -124,7 +125,8 @@ func getVirtualObject(tx *bbolt.Tx, cnr cid.ID, key []byte, raw bool) (*objectSD
 		return nil, getSplitInfoError(tx, cnr, key)
 	}
 
-	parentBucket := tx.Bucket(parentBucketName(cnr))
+	bucketName := make([]byte, bucketKeySize)
+	parentBucket := tx.Bucket(parentBucketName(cnr, bucketName))
 	if parentBucket == nil {
 		var errNotFound apistatus.ObjectNotFound
 
@@ -146,7 +148,7 @@ func getVirtualObject(tx *bbolt.Tx, cnr cid.ID, key []byte, raw bool) (*objectSD
 	// but later list might be sorted so first or last value can be more
 	// prioritized to choose
 	virtualOID := relativeLst[len(relativeLst)-1]
-	data := getFromBucket(tx, primaryBucketName(cnr), virtualOID)
+	data := getFromBucket(tx, primaryBucketName(cnr, bucketName), virtualOID)
 
 	child := objectSDK.New()
 
diff --git a/pkg/local_object_storage/metabase/graveyard.go b/pkg/local_object_storage/metabase/graveyard.go
index 3b96bcb5fe..dcf7d1e831 100644
--- a/pkg/local_object_storage/metabase/graveyard.go
+++ b/pkg/local_object_storage/metabase/graveyard.go
@@ -174,7 +174,7 @@ func (db *DB) iterateDeletedObj(tx *bbolt.Tx, h kvHandler, offset *oid.Address)
 	if offset == nil {
 		k, v = c.First()
 	} else {
-		rawAddr := addressKey(*offset)
+		rawAddr := addressKey(*offset, make([]byte, addressKeySize))
 
 		k, v = c.Seek(rawAddr)
 		if bytes.Equal(k, rawAddr) {
@@ -222,6 +222,8 @@ func graveFromKV(k, v []byte) (res TombstonedObject, err error) {
 //
 // Returns any error appeared during deletion process.
 func (db *DB) DropGraves(tss []TombstonedObject) error {
+	buf := make([]byte, addressKeySize)
+
 	return db.boltDB.Update(func(tx *bbolt.Tx) error {
 		bkt := tx.Bucket(graveyardBucketName)
 		if bkt == nil {
@@ -229,7 +231,7 @@ func (db *DB) DropGraves(tss []TombstonedObject) error {
 		}
 
 		for _, ts := range tss {
-			err := bkt.Delete(addressKey(ts.Address()))
+			err := bkt.Delete(addressKey(ts.Address(), buf))
 			if err != nil {
 				return err
 			}
diff --git a/pkg/local_object_storage/metabase/inhume.go b/pkg/local_object_storage/metabase/inhume.go
index 344259b618..296e3bd6c9 100644
--- a/pkg/local_object_storage/metabase/inhume.go
+++ b/pkg/local_object_storage/metabase/inhume.go
@@ -113,7 +113,7 @@ func (db *DB) Inhume(prm InhumePrm) (res InhumeRes, err error) {
 
 		if prm.tomb != nil {
 			bkt = graveyardBKT
-			tombKey := addressKey(*prm.tomb)
+			tombKey := addressKey(*prm.tomb, make([]byte, addressKeySize))
 
 			// it is forbidden to have a tomb-on-tomb in NeoFS,
 			// so graveyard keys must not be addresses of tombstones
@@ -131,6 +131,7 @@ func (db *DB) Inhume(prm InhumePrm) (res InhumeRes, err error) {
 			value = zeroValue
 		}
 
+		buf := make([]byte, addressKeySize)
 		for i := range prm.target {
 			id := prm.target[i].Object()
 			cnr := prm.target[i].Container()
@@ -153,9 +154,8 @@ func (db *DB) Inhume(prm InhumePrm) (res InhumeRes, err error) {
 				lockWasChecked = true
 			}
 
-			targetKey := addressKey(prm.target[i])
-
-			obj, err := db.get(tx, prm.target[i], false, true, currEpoch)
+			obj, err := db.get(tx, prm.target[i], buf, false, true, currEpoch)
+			targetKey := addressKey(prm.target[i], buf)
 			if err == nil {
 				if inGraveyardWithKey(targetKey, graveyardBKT, garbageBKT) == 0 {
 					// object is available, decrement the
diff --git a/pkg/local_object_storage/metabase/iterators.go b/pkg/local_object_storage/metabase/iterators.go
index 85ed49f234..cc9b032b16 100644
--- a/pkg/local_object_storage/metabase/iterators.go
+++ b/pkg/local_object_storage/metabase/iterators.go
@@ -55,6 +55,12 @@ func (db *DB) iterateExpired(tx *bbolt.Tx, epoch uint64, h ExpiredObjectHandler)
 			return nil
 		}
 
+		var cnrID cid.ID
+		err := cnrID.Decode(cidBytes)
+		if err != nil {
+			return fmt.Errorf("could not parse container ID of expired bucket: %w", err)
+		}
+
 		return b.ForEach(func(expKey, _ []byte) error {
 			bktExpired := b.Bucket(expKey)
 			if bktExpired == nil {
@@ -71,18 +77,11 @@ func (db *DB) iterateExpired(tx *bbolt.Tx, epoch uint64, h ExpiredObjectHandler)
 			return bktExpired.ForEach(func(idKey, _ []byte) error {
 				var id oid.ID
 
-				err = id.DecodeString(string(idKey))
+				err = id.Decode(idKey)
 				if err != nil {
 					return fmt.Errorf("could not parse ID of expired object: %w", err)
 				}
 
-				var cnrID cid.ID
-
-				err = cnrID.DecodeString(string(cidBytes))
-				if err != nil {
-					return fmt.Errorf("could not parse container ID of expired bucket: %w", err)
-				}
-
 				// Ignore locked objects.
 				//
 				// To slightly optimize performance we can check only REGULAR objects
@@ -131,7 +130,11 @@ func (db *DB) iterateCoveredByTombstones(tx *bbolt.Tx, tss map[string]oid.Addres
 	}
 
 	err := bktGraveyard.ForEach(func(k, v []byte) error {
-		if _, ok := tss[string(v)]; ok {
+		var addr oid.Address
+		if err := decodeAddressFromKey(&addr, v); err != nil {
+			return err
+		}
+		if _, ok := tss[addr.EncodeToString()]; ok {
 			var addr oid.Address
 
 			err := decodeAddressFromKey(&addr, k)
@@ -161,22 +164,22 @@ func iteratePhyObjects(tx *bbolt.Tx, f func(cid.ID, oid.ID) error) error {
 	var oid oid.ID
 
 	return tx.ForEach(func(name []byte, b *bbolt.Bucket) error {
-		b58CID, postfix := parseContainerIDWithPostfix(&cid, name)
+		b58CID, postfix := parseContainerIDWithPrefix(&cid, name)
 		if len(b58CID) == 0 {
 			return nil
 		}
 
 		switch postfix {
-		case "",
-			storageGroupPostfix,
-			bucketNameSuffixLockers,
-			tombstonePostfix:
+		case primaryPrefix,
+			storageGroupPrefix,
+			lockersPrefix,
+			tombstonePrefix:
 		default:
 			return nil
 		}
 
 		return b.ForEach(func(k, v []byte) error {
-			if oid.DecodeString(string(k)) == nil {
+			if oid.Decode(k) == nil {
 				return f(cid, oid)
 			}
 
diff --git a/pkg/local_object_storage/metabase/list.go b/pkg/local_object_storage/metabase/list.go
index 1b2acc3197..5ab8830459 100644
--- a/pkg/local_object_storage/metabase/list.go
+++ b/pkg/local_object_storage/metabase/list.go
@@ -1,7 +1,6 @@
 package meta
 
 import (
-	"bytes"
 	"errors"
 
 	cid "github.com/nspcc-dev/neofs-sdk-go/container/id"
@@ -90,30 +89,28 @@ func (db *DB) listWithCursor(tx *bbolt.Tx, result []oid.Address, count int, curs
 	graveyardBkt := tx.Bucket(graveyardBucketName)
 	garbageBkt := tx.Bucket(garbageBucketName)
 
-	const idSize = 44 // size of the stringified object and container ids
-	var rawAddr = make([]byte, idSize*2+1)
+	var rawAddr = make([]byte, cidSize, addressKeySize)
 
 loop:
 	for ; name != nil; name, _ = c.Next() {
-		b58CID, postfix := parseContainerIDWithPostfix(&containerID, name)
-		if b58CID == nil {
+		cidRaw, prefix := parseContainerIDWithPrefix(&containerID, name)
+		if cidRaw == nil {
 			continue
 		}
 
-		switch postfix {
+		switch prefix {
 		case
-			"",
-			storageGroupPostfix,
-			bucketNameSuffixLockers,
-			tombstonePostfix:
+			primaryPrefix,
+			storageGroupPrefix,
+			lockersPrefix,
+			tombstonePrefix:
 		default:
 			continue
 		}
 
 		bkt := tx.Bucket(name)
 		if bkt != nil {
-			rawAddr = append(rawAddr[:0], b58CID...)
-			rawAddr = append(rawAddr, '/')
+			copy(rawAddr, cidRaw)
 			result, offset, cursor = selectNFromBucket(bkt, graveyardBkt, garbageBkt, rawAddr, containerID,
 				result, count, cursor, threshold)
 		}
@@ -150,7 +147,7 @@ loop:
 // object to start selecting from. Ignores inhumed objects.
 func selectNFromBucket(bkt *bbolt.Bucket, // main bucket
 	graveyardBkt, garbageBkt *bbolt.Bucket, // cached graveyard buckets
-	addrRaw []byte, // container ID prefix, optimization
+	cidRaw []byte, // container ID prefix, optimization
 	cnt cid.ID, // container ID
 	to []oid.Address, // listing result
 	limit int, // stop listing at `limit` items in result
@@ -178,12 +175,12 @@ func selectNFromBucket(bkt *bbolt.Bucket, // main bucket
 		}
 
 		var obj oid.ID
-		if err := obj.DecodeString(string(k)); err != nil {
+		if err := obj.Decode(k); err != nil {
 			break
 		}
 
 		offset = k
-		if inGraveyardWithKey(append(addrRaw, k...), graveyardBkt, garbageBkt) > 0 {
+		if inGraveyardWithKey(append(cidRaw, k...), graveyardBkt, garbageBkt) > 0 {
 			continue
 		}
 
@@ -197,21 +194,16 @@ func selectNFromBucket(bkt *bbolt.Bucket, // main bucket
 	return to, offset, cursor
 }
 
-func parseContainerIDWithPostfix(containerID *cid.ID, name []byte) ([]byte, string) {
-	var (
-		containerIDStr = name
-		postfix        []byte
-	)
-
-	ind := bytes.IndexByte(name, invalidBase58String[0])
-	if ind > 0 {
-		postfix = containerIDStr[ind:]
-		containerIDStr = containerIDStr[:ind]
+func parseContainerIDWithPrefix(containerID *cid.ID, name []byte) ([]byte, byte) {
+	if len(name) < bucketKeySize {
+		return nil, 0
 	}
 
-	if err := containerID.DecodeString(string(containerIDStr)); err != nil {
-		return nil, ""
+	rawID := name[1:bucketKeySize]
+
+	if err := containerID.Decode(rawID); err != nil {
+		return nil, 0
 	}
 
-	return containerIDStr, string(postfix)
+	return rawID, name[0]
 }
diff --git a/pkg/local_object_storage/metabase/lock.go b/pkg/local_object_storage/metabase/lock.go
index ab83ee59f4..c9e0ff5db7 100644
--- a/pkg/local_object_storage/metabase/lock.go
+++ b/pkg/local_object_storage/metabase/lock.go
@@ -11,15 +11,11 @@ import (
 	"go.etcd.io/bbolt"
 )
 
-// bucket name for locked objects.
-var bucketNameLocked = []byte(invalidBase58String + "Locked")
-
-// suffix for container buckets with objects of type LOCK.
-const bucketNameSuffixLockers = invalidBase58String + "LOCKER"
+var bucketNameLocked = []byte{lockedPrefix}
 
 // returns name of the bucket with objects of type LOCK for specified container.
-func bucketNameLockers(idCnr cid.ID) []byte {
-	return []byte(idCnr.EncodeToString() + bucketNameSuffixLockers)
+func bucketNameLockers(idCnr cid.ID, key []byte) []byte {
+	return bucketName(idCnr, lockersPrefix, key)
 }
 
 // Lock marks objects as locked with another object. All objects are from the
@@ -36,14 +32,14 @@ func (db *DB) Lock(cnr cid.ID, locker oid.ID, locked []oid.ID) error {
 		panic("empty locked list")
 	}
 
+	// check if all objects are regular
+	bucketKeysLocked := make([][]byte, len(locked))
+	for i := range locked {
+		bucketKeysLocked[i] = objectKey(locked[i], make([]byte, objectKeySize))
+	}
+	key := make([]byte, cidSize)
+
 	return db.boltDB.Update(func(tx *bbolt.Tx) error {
-		// check if all objects are regular
-		bucketKeysLocked := make([][]byte, len(locked))
-
-		for i := range locked {
-			bucketKeysLocked[i] = objectKey(locked[i])
-		}
-
 		if firstIrregularObjectType(tx, cnr, bucketKeysLocked...) != object.TypeRegular {
 			return apistatus.LockNonRegularObject{}
 		}
@@ -53,12 +49,13 @@ func (db *DB) Lock(cnr cid.ID, locker oid.ID, locked []oid.ID) error {
 			return fmt.Errorf("create global bucket for locked objects: %w", err)
 		}
 
-		bucketLockedContainer, err := bucketLocked.CreateBucketIfNotExists([]byte(cnr.EncodeToString()))
+		cnr.Encode(key)
+		bucketLockedContainer, err := bucketLocked.CreateBucketIfNotExists(key)
 		if err != nil {
 			return fmt.Errorf("create container bucket for locked objects %v: %w", cnr, err)
 		}
 
-		keyLocker := objectKey(locker)
+		keyLocker := objectKey(locker, key)
 		var exLockers [][]byte
 		var updLockers []byte
 
@@ -114,9 +111,11 @@ func (db *DB) FreeLockedBy(lockers []oid.Address) error {
 func objectLocked(tx *bbolt.Tx, idCnr cid.ID, idObj oid.ID) bool {
 	bucketLocked := tx.Bucket(bucketNameLocked)
 	if bucketLocked != nil {
-		bucketLockedContainer := bucketLocked.Bucket([]byte(idCnr.EncodeToString()))
+		key := make([]byte, cidSize)
+		idCnr.Encode(key)
+		bucketLockedContainer := bucketLocked.Bucket(key)
 		if bucketLockedContainer != nil {
-			return bucketLockedContainer.Get(objectKey(idObj)) != nil
+			return bucketLockedContainer.Get(objectKey(idObj, key)) != nil
 		}
 	}
 
@@ -131,9 +130,12 @@ func objectLocked(tx *bbolt.Tx, idCnr cid.ID, idObj oid.ID) bool {
 func freePotentialLocks(tx *bbolt.Tx, idCnr cid.ID, locker oid.ID) error {
 	bucketLocked := tx.Bucket(bucketNameLocked)
 	if bucketLocked != nil {
-		bucketLockedContainer := bucketLocked.Bucket([]byte(idCnr.EncodeToString()))
+		key := make([]byte, cidSize)
+		idCnr.Encode(key)
+
+		bucketLockedContainer := bucketLocked.Bucket(key)
 		if bucketLockedContainer != nil {
-			keyLocker := objectKey(locker)
+			keyLocker := objectKey(locker, key)
 			return bucketLockedContainer.ForEach(func(k, v []byte) error {
 				keyLockers, err := decodeList(v)
 				if err != nil {
diff --git a/pkg/local_object_storage/metabase/movable.go b/pkg/local_object_storage/metabase/movable.go
index 42935e391f..8011490d25 100644
--- a/pkg/local_object_storage/metabase/movable.go
+++ b/pkg/local_object_storage/metabase/movable.go
@@ -52,13 +52,16 @@ func (db *DB) ToMoveIt(prm ToMoveItPrm) (res ToMoveItRes, err error) {
 	db.modeMtx.RLock()
 	defer db.modeMtx.RUnlock()
 
+	key := make([]byte, addressKeySize)
+	key = addressKey(prm.addr, key)
+
 	err = db.boltDB.Update(func(tx *bbolt.Tx) error {
 		toMoveIt, err := tx.CreateBucketIfNotExists(toMoveItBucketName)
 		if err != nil {
 			return err
 		}
 
-		return toMoveIt.Put(addressKey(prm.addr), zeroValue)
+		return toMoveIt.Put(key, zeroValue)
 	})
 
 	return
@@ -69,13 +72,16 @@ func (db *DB) DoNotMove(prm DoNotMovePrm) (res DoNotMoveRes, err error) {
 	db.modeMtx.RLock()
 	defer db.modeMtx.RUnlock()
 
+	key := make([]byte, addressKeySize)
+	key = addressKey(prm.addr, key)
+
 	err = db.boltDB.Update(func(tx *bbolt.Tx) error {
 		toMoveIt := tx.Bucket(toMoveItBucketName)
 		if toMoveIt == nil {
 			return nil
 		}
 
-		return toMoveIt.Delete(addressKey(prm.addr))
+		return toMoveIt.Delete(key)
 	})
 
 	return
diff --git a/pkg/local_object_storage/metabase/put.go b/pkg/local_object_storage/metabase/put.go
index 3f439e3ba5..b390056335 100644
--- a/pkg/local_object_storage/metabase/put.go
+++ b/pkg/local_object_storage/metabase/put.go
@@ -168,21 +168,20 @@ func putUniqueIndexes(
 	isParent := si != nil
 	addr := object.AddressOf(obj)
 	cnr := addr.Container()
-	objKey := objectKey(addr.Object())
+	objKey := objectKey(addr.Object(), make([]byte, objectKeySize))
 
+	bucketName := make([]byte, bucketKeySize)
 	// add value to primary unique bucket
 	if !isParent {
-		var bucketName []byte
-
 		switch obj.Type() {
 		case objectSDK.TypeRegular:
-			bucketName = primaryBucketName(cnr)
+			bucketName = primaryBucketName(cnr, bucketName)
 		case objectSDK.TypeTombstone:
-			bucketName = tombstoneBucketName(cnr)
+			bucketName = tombstoneBucketName(cnr, bucketName)
 		case objectSDK.TypeStorageGroup:
-			bucketName = storageGroupBucketName(cnr)
+			bucketName = storageGroupBucketName(cnr, bucketName)
 		case objectSDK.TypeLock:
-			bucketName = bucketNameLockers(cnr)
+			bucketName = bucketNameLockers(cnr, bucketName)
 		default:
 			return ErrUnknownObjectType
 		}
@@ -204,7 +203,7 @@ func putUniqueIndexes(
 		// index storageID if it is present
 		if id != nil {
 			err = putUniqueIndexItem(tx, namedBucketItem{
-				name: smallBucketName(cnr),
+				name: smallBucketName(cnr, bucketName),
 				key:  objKey,
 				val:  id,
 			})
@@ -229,7 +228,7 @@ func putUniqueIndexes(
 		}
 
 		err = putUniqueIndexItem(tx, namedBucketItem{
-			name: rootBucketName(cnr),
+			name: rootBucketName(cnr, bucketName),
 			key:  objKey,
 			val:  splitInfo,
 		})
@@ -246,13 +245,14 @@ type updateIndexItemFunc = func(tx *bbolt.Tx, item namedBucketItem) error
 func updateListIndexes(tx *bbolt.Tx, obj *objectSDK.Object, f updateIndexItemFunc) error {
 	idObj, _ := obj.ID()
 	cnr, _ := obj.ContainerID()
-	objKey := objectKey(idObj)
+	objKey := objectKey(idObj, make([]byte, objectKeySize))
+	bucketName := make([]byte, bucketKeySize)
 
 	cs, _ := obj.PayloadChecksum()
 
 	// index payload hashes
 	err := f(tx, namedBucketItem{
-		name: payloadHashBucketName(cnr),
+		name: payloadHashBucketName(cnr, bucketName),
 		key:  cs.Value(),
 		val:  objKey,
 	})
@@ -265,8 +265,8 @@ func updateListIndexes(tx *bbolt.Tx, obj *objectSDK.Object, f updateIndexItemFun
 	// index parent ids
 	if ok {
 		err := f(tx, namedBucketItem{
-			name: parentBucketName(cnr),
-			key:  objectKey(idParent),
+			name: parentBucketName(cnr, bucketName),
+			key:  objectKey(idParent, make([]byte, objectKeySize)),
 			val:  objKey,
 		})
 		if err != nil {
@@ -277,7 +277,7 @@ func updateListIndexes(tx *bbolt.Tx, obj *objectSDK.Object, f updateIndexItemFun
 	// index split ids
 	if obj.SplitID() != nil {
 		err := f(tx, namedBucketItem{
-			name: splitBucketName(cnr),
+			name: splitBucketName(cnr, bucketName),
 			key:  obj.SplitID().ToV2(),
 			val:  objKey,
 		})
@@ -292,12 +292,13 @@ func updateListIndexes(tx *bbolt.Tx, obj *objectSDK.Object, f updateIndexItemFun
 func updateFKBTIndexes(tx *bbolt.Tx, obj *objectSDK.Object, f updateIndexItemFunc) error {
 	id, _ := obj.ID()
 	cnr, _ := obj.ContainerID()
-	objKey := []byte(id.EncodeToString())
+	objKey := objectKey(id, make([]byte, objectKeySize))
 
 	attrs := obj.Attributes()
 
+	key := make([]byte, bucketKeySize)
 	err := f(tx, namedBucketItem{
-		name: ownerBucketName(cnr),
+		name: ownerBucketName(cnr, key),
 		key:  []byte(obj.OwnerID().EncodeToString()),
 		val:  objKey,
 	})
@@ -307,8 +308,9 @@ func updateFKBTIndexes(tx *bbolt.Tx, obj *objectSDK.Object, f updateIndexItemFun
 
 	// user specified attributes
 	for i := range attrs {
+		key = attributeBucketName(cnr, attrs[i].Key(), key)
 		err := f(tx, namedBucketItem{
-			name: attributeBucketName(cnr, attrs[i].Key()),
+			name: key,
 			key:  []byte(attrs[i].Value()),
 			val:  objKey,
 		})
@@ -437,25 +439,27 @@ func getVarUint(data []byte) (uint64, int, error) {
 // updateStorageID for existing objects if they were moved from one
 // storage location to another.
 func updateStorageID(tx *bbolt.Tx, addr oid.Address, id []byte) error {
-	bkt, err := tx.CreateBucketIfNotExists(smallBucketName(addr.Container()))
+	key := make([]byte, bucketKeySize)
+	bkt, err := tx.CreateBucketIfNotExists(smallBucketName(addr.Container(), key))
 	if err != nil {
 		return err
 	}
 
-	return bkt.Put(objectKey(addr.Object()), id)
+	return bkt.Put(objectKey(addr.Object(), key), id)
 }
 
 // updateSpliInfo for existing objects if storage filled with extra information
 // about last object in split hierarchy or linking object.
 func updateSplitInfo(tx *bbolt.Tx, addr oid.Address, from *objectSDK.SplitInfo) error {
-	bkt := tx.Bucket(rootBucketName(addr.Container()))
+	key := make([]byte, bucketKeySize)
+	bkt := tx.Bucket(rootBucketName(addr.Container(), key))
 	if bkt == nil {
 		// if object doesn't exists and we want to update split info on it
 		// then ignore, this should never happen
 		return ErrIncorrectSplitInfoUpdate
 	}
 
-	objectKey := objectKey(addr.Object())
+	objectKey := objectKey(addr.Object(), key)
 
 	rawSplitInfo := bkt.Get(objectKey)
 	if len(rawSplitInfo) == 0 {
diff --git a/pkg/local_object_storage/metabase/select.go b/pkg/local_object_storage/metabase/select.go
index a8b40b1523..1da56b12f8 100644
--- a/pkg/local_object_storage/metabase/select.go
+++ b/pkg/local_object_storage/metabase/select.go
@@ -107,13 +107,16 @@ func (db *DB) selectObjects(tx *bbolt.Tx, cnr cid.ID, fs object.SearchFilters, c
 			continue // ignore objects with unmatched fast filters
 		}
 
-		var addr oid.Address
-
-		err = decodeAddressFromKey(&addr, []byte(a))
+		var id oid.ID
+		err = id.Decode([]byte(a))
 		if err != nil {
 			return nil, err
 		}
 
+		var addr oid.Address
+		addr.SetContainer(cnr)
+		addr.SetObject(id)
+
 		if objectStatus(tx, addr, currEpoch) > 0 {
 			continue // ignore removed objects
 		}
@@ -130,26 +133,24 @@ func (db *DB) selectObjects(tx *bbolt.Tx, cnr cid.ID, fs object.SearchFilters, c
 
 // selectAll adds to resulting cache all available objects in metabase.
 func (db *DB) selectAll(tx *bbolt.Tx, cnr cid.ID, to map[string]int) {
-	prefix := cnr.EncodeToString() + "/"
-
-	selectAllFromBucket(tx, primaryBucketName(cnr), prefix, to, 0)
-	selectAllFromBucket(tx, tombstoneBucketName(cnr), prefix, to, 0)
-	selectAllFromBucket(tx, storageGroupBucketName(cnr), prefix, to, 0)
-	selectAllFromBucket(tx, parentBucketName(cnr), prefix, to, 0)
-	selectAllFromBucket(tx, bucketNameLockers(cnr), prefix, to, 0)
+	bucketName := make([]byte, bucketKeySize)
+	selectAllFromBucket(tx, primaryBucketName(cnr, bucketName), to, 0)
+	selectAllFromBucket(tx, tombstoneBucketName(cnr, bucketName), to, 0)
+	selectAllFromBucket(tx, storageGroupBucketName(cnr, bucketName), to, 0)
+	selectAllFromBucket(tx, parentBucketName(cnr, bucketName), to, 0)
+	selectAllFromBucket(tx, bucketNameLockers(cnr, bucketName), to, 0)
 }
 
 // selectAllFromBucket goes through all keys in bucket and adds them in a
 // resulting cache. Keys should be stringed object ids.
-func selectAllFromBucket(tx *bbolt.Tx, name []byte, prefix string, to map[string]int, fNum int) {
+func selectAllFromBucket(tx *bbolt.Tx, name []byte, to map[string]int, fNum int) {
 	bkt := tx.Bucket(name)
 	if bkt == nil {
 		return
 	}
 
 	_ = bkt.ForEach(func(k, v []byte) error {
-		key := prefix + string(k) // consider using string builders from sync.Pool
-		markAddressInCache(to, fNum, key)
+		markAddressInCache(to, fNum, string(k))
 
 		return nil
 	})
@@ -164,47 +165,46 @@ func (db *DB) selectFastFilter(
 	to map[string]int, // resulting cache
 	fNum int, // index of filter
 ) {
-	prefix := cnr.EncodeToString() + "/"
 	currEpoch := db.epochState.CurrentEpoch()
-
+	bucketName := make([]byte, bucketKeySize)
 	switch f.Header() {
 	case v2object.FilterHeaderObjectID:
 		db.selectObjectID(tx, f, cnr, to, fNum, currEpoch)
 	case v2object.FilterHeaderOwnerID:
-		bucketName := ownerBucketName(cnr)
-		db.selectFromFKBT(tx, bucketName, f, prefix, to, fNum)
+		bucketName := ownerBucketName(cnr, bucketName)
+		db.selectFromFKBT(tx, bucketName, f, to, fNum)
 	case v2object.FilterHeaderPayloadHash:
-		bucketName := payloadHashBucketName(cnr)
-		db.selectFromList(tx, bucketName, f, prefix, to, fNum)
+		bucketName := payloadHashBucketName(cnr, bucketName)
+		db.selectFromList(tx, bucketName, f, to, fNum)
 	case v2object.FilterHeaderObjectType:
 		for _, bucketName := range bucketNamesForType(cnr, f.Operation(), f.Value()) {
-			selectAllFromBucket(tx, bucketName, prefix, to, fNum)
+			selectAllFromBucket(tx, bucketName, to, fNum)
 		}
 	case v2object.FilterHeaderParent:
-		bucketName := parentBucketName(cnr)
-		db.selectFromList(tx, bucketName, f, prefix, to, fNum)
+		bucketName := parentBucketName(cnr, bucketName)
+		db.selectFromList(tx, bucketName, f, to, fNum)
 	case v2object.FilterHeaderSplitID:
-		bucketName := splitBucketName(cnr)
-		db.selectFromList(tx, bucketName, f, prefix, to, fNum)
+		bucketName := splitBucketName(cnr, bucketName)
+		db.selectFromList(tx, bucketName, f, to, fNum)
 	case v2object.FilterPropertyRoot:
-		selectAllFromBucket(tx, rootBucketName(cnr), prefix, to, fNum)
+		selectAllFromBucket(tx, rootBucketName(cnr, bucketName), to, fNum)
 	case v2object.FilterPropertyPhy:
-		selectAllFromBucket(tx, primaryBucketName(cnr), prefix, to, fNum)
-		selectAllFromBucket(tx, tombstoneBucketName(cnr), prefix, to, fNum)
-		selectAllFromBucket(tx, storageGroupBucketName(cnr), prefix, to, fNum)
-		selectAllFromBucket(tx, bucketNameLockers(cnr), prefix, to, fNum)
+		selectAllFromBucket(tx, primaryBucketName(cnr, bucketName), to, fNum)
+		selectAllFromBucket(tx, tombstoneBucketName(cnr, bucketName), to, fNum)
+		selectAllFromBucket(tx, storageGroupBucketName(cnr, bucketName), to, fNum)
+		selectAllFromBucket(tx, bucketNameLockers(cnr, bucketName), to, fNum)
 	default: // user attribute
-		bucketName := attributeBucketName(cnr, f.Header())
+		bucketName := attributeBucketName(cnr, f.Header(), bucketName)
 
 		if f.Operation() == object.MatchNotPresent {
-			selectOutsideFKBT(tx, allBucketNames(cnr), bucketName, f, prefix, to, fNum)
+			selectOutsideFKBT(tx, allBucketNames(cnr), bucketName, to, fNum)
 		} else {
-			db.selectFromFKBT(tx, bucketName, f, prefix, to, fNum)
+			db.selectFromFKBT(tx, bucketName, f, to, fNum)
 		}
 	}
 }
 
-var mBucketNaming = map[string][]func(cid.ID) []byte{
+var mBucketNaming = map[string][]func(cid.ID, []byte) []byte{
 	v2object.TypeRegular.String():      {primaryBucketName, parentBucketName},
 	v2object.TypeTombstone.String():    {tombstoneBucketName},
 	v2object.TypeStorageGroup.String(): {storageGroupBucketName},
@@ -214,7 +214,7 @@ var mBucketNaming = map[string][]func(cid.ID) []byte{
 func allBucketNames(cnr cid.ID) (names [][]byte) {
 	for _, fns := range mBucketNaming {
 		for _, fn := range fns {
-			names = append(names, fn(cnr))
+			names = append(names, fn(cnr, make([]byte, bucketKeySize)))
 		}
 	}
 
@@ -226,7 +226,7 @@ func bucketNamesForType(cnr cid.ID, mType object.SearchMatchType, typeVal string
 		fns, ok := mBucketNaming[key]
 		if ok {
 			for _, fn := range fns {
-				names = append(names, fn(cnr))
+				names = append(names, fn(cnr, make([]byte, bucketKeySize)))
 			}
 		}
 	}
@@ -258,7 +258,6 @@ func (db *DB) selectFromFKBT(
 	tx *bbolt.Tx,
 	name []byte, // fkbt root bucket name
 	f object.SearchFilter, // filter for operation and value
-	prefix string, // prefix to create addr from oid in index
 	to map[string]int, // resulting cache
 	fNum int, // index of filter
 ) { //
@@ -281,8 +280,7 @@ func (db *DB) selectFromFKBT(
 		}
 
 		return fkbtLeaf.ForEach(func(k, _ []byte) error {
-			addr := prefix + string(k)
-			markAddressInCache(to, fNum, addr)
+			markAddressInCache(to, fNum, string(k))
 
 			return nil
 		})
@@ -298,8 +296,6 @@ func selectOutsideFKBT(
 	tx *bbolt.Tx,
 	incl [][]byte, // buckets
 	name []byte, // fkbt root bucket name
-	f object.SearchFilter, // filter for operation and value
-	prefix string, // prefix to create addr from oid in index
 	to map[string]int, // resulting cache
 	fNum int, // index of filter
 ) {
@@ -314,8 +310,7 @@ func selectOutsideFKBT(
 			}
 
 			return exclBktLeaf.ForEach(func(k, _ []byte) error {
-				addr := prefix + string(k)
-				mExcl[addr] = struct{}{}
+				mExcl[string(k)] = struct{}{}
 
 				return nil
 			})
@@ -329,10 +324,8 @@ func selectOutsideFKBT(
 		}
 
 		_ = bktIncl.ForEach(func(k, _ []byte) error {
-			addr := prefix + string(k)
-
-			if _, ok := mExcl[addr]; !ok {
-				markAddressInCache(to, fNum, addr)
+			if _, ok := mExcl[string(k)]; !ok {
+				markAddressInCache(to, fNum, string(k))
 			}
 
 			return nil
@@ -346,7 +339,6 @@ func (db *DB) selectFromList(
 	tx *bbolt.Tx,
 	name []byte, // list root bucket name
 	f object.SearchFilter, // filter for operation and value
-	prefix string, // prefix to create addr from oid in index
 	to map[string]int, // resulting cache
 	fNum int, // index of filter
 ) { //
@@ -398,8 +390,7 @@ func (db *DB) selectFromList(
 	}
 
 	for i := range lst {
-		addr := prefix + string(lst[i])
-		markAddressInCache(to, fNum, addr)
+		markAddressInCache(to, fNum, string(lst[i]))
 	}
 }
 
@@ -412,30 +403,25 @@ func (db *DB) selectObjectID(
 	fNum int, // index of filter
 	currEpoch uint64,
 ) {
-	prefix := cnr.EncodeToString() + "/"
-
-	appendOID := func(strObj string) {
-		addrStr := prefix + strObj
+	appendOID := func(id oid.ID) {
 		var addr oid.Address
-
-		err := decodeAddressFromKey(&addr, []byte(addrStr))
-		if err != nil {
-			db.log.Debug("can't decode object id address",
-				zap.String("addr", addrStr),
-				zap.String("error", err.Error()))
-
-			return
-		}
+		addr.SetContainer(cnr)
+		addr.SetObject(id)
 
 		ok, err := db.exists(tx, addr, currEpoch)
 		if (err == nil && ok) || errors.As(err, &splitInfoError) {
-			markAddressInCache(to, fNum, addrStr)
+			raw := make([]byte, objectKeySize)
+			id.Encode(raw)
+			markAddressInCache(to, fNum, string(raw))
 		}
 	}
 
 	switch op := f.Operation(); op {
 	case object.MatchStringEqual:
-		appendOID(f.Value())
+		var id oid.ID
+		if err := id.DecodeString(f.Value()); err == nil {
+			appendOID(id)
+		}
 	default:
 		fMatch, ok := db.matchers[op]
 		if !ok {
@@ -454,7 +440,10 @@ func (db *DB) selectObjectID(
 			}
 
 			err := fMatch.matchBucket(bkt, f.Header(), f.Value(), func(k, v []byte) error {
-				appendOID(string(k))
+				var id oid.ID
+				if err := id.Decode(k); err == nil {
+					appendOID(id)
+				}
 				return nil
 			})
 			if err != nil {
@@ -472,7 +461,8 @@ func (db *DB) matchSlowFilters(tx *bbolt.Tx, addr oid.Address, f object.SearchFi
 		return true
 	}
 
-	obj, err := db.get(tx, addr, true, false, currEpoch)
+	buf := make([]byte, addressKeySize)
+	obj, err := db.get(tx, addr, buf, true, false, currEpoch)
 	if err != nil {
 		return false
 	}
diff --git a/pkg/local_object_storage/metabase/shard_id.go b/pkg/local_object_storage/metabase/shard_id.go
index 93313032cf..312f3900ce 100644
--- a/pkg/local_object_storage/metabase/shard_id.go
+++ b/pkg/local_object_storage/metabase/shard_id.go
@@ -6,7 +6,7 @@ import (
 )
 
 var (
-	shardInfoBucket = []byte(invalidBase58String + "i")
+	shardInfoBucket = []byte{shardInfoPrefix}
 	shardIDKey      = []byte("id")
 )
 
diff --git a/pkg/local_object_storage/metabase/storage_id.go b/pkg/local_object_storage/metabase/storage_id.go
index ce893cffd5..e8cca5b260 100644
--- a/pkg/local_object_storage/metabase/storage_id.go
+++ b/pkg/local_object_storage/metabase/storage_id.go
@@ -39,12 +39,13 @@ func (db *DB) StorageID(prm StorageIDPrm) (res StorageIDRes, err error) {
 }
 
 func (db *DB) storageID(tx *bbolt.Tx, addr oid.Address) ([]byte, error) {
-	smallBucket := tx.Bucket(smallBucketName(addr.Container()))
+	key := make([]byte, bucketKeySize)
+	smallBucket := tx.Bucket(smallBucketName(addr.Container(), key))
 	if smallBucket == nil {
 		return nil, nil
 	}
 
-	storageID := smallBucket.Get(objectKey(addr.Object()))
+	storageID := smallBucket.Get(objectKey(addr.Object(), key))
 	if storageID == nil {
 		return nil, nil
 	}
diff --git a/pkg/local_object_storage/metabase/util.go b/pkg/local_object_storage/metabase/util.go
index b18549b905..eae4f49f45 100644
--- a/pkg/local_object_storage/metabase/util.go
+++ b/pkg/local_object_storage/metabase/util.go
@@ -2,8 +2,8 @@ package meta
 
 import (
 	"bytes"
+	"crypto/sha256"
 	"fmt"
-	"strings"
 
 	cid "github.com/nspcc-dev/neofs-sdk-go/container/id"
 	"github.com/nspcc-dev/neofs-sdk-go/object"
@@ -11,126 +11,214 @@ import (
 	"go.etcd.io/bbolt"
 )
 
-/*
-We might increase performance by not using string representation of
-identities and addresses. String representation require base58 encoding that
-slows execution. Instead we can try to marshal these structures directly into
-bytes. Check it later.
-*/
-
-const invalidBase58String = "_"
-
 var (
 	// graveyardBucketName stores rows with the objects that have been
 	// covered with Tombstone objects. That objects should not be returned
 	// from the node and should not be accepted by the node from other
 	// nodes.
-	graveyardBucketName = []byte(invalidBase58String + "Graveyard")
+	graveyardBucketName = []byte{graveyardPrefix}
 	// garbageBucketName stores rows with the objects that should be physically
 	// deleted by the node (Garbage Collector routine).
-	garbageBucketName         = []byte(invalidBase58String + "Garbage")
-	toMoveItBucketName        = []byte(invalidBase58String + "ToMoveIt")
-	containerVolumeBucketName = []byte(invalidBase58String + "ContainerSize")
+	garbageBucketName         = []byte{garbagePrefix}
+	toMoveItBucketName        = []byte{toMoveItPrefix}
+	containerVolumeBucketName = []byte{containerVolumePrefix}
 
 	zeroValue = []byte{0xFF}
-
-	smallPostfix        = invalidBase58String + "small"
-	storageGroupPostfix = invalidBase58String + "SG"
-	tombstonePostfix    = invalidBase58String + "TS"
-	ownerPostfix        = invalidBase58String + "ownerid"
-	payloadHashPostfix  = invalidBase58String + "payloadhash"
-	rootPostfix         = invalidBase58String + "root"
-	parentPostfix       = invalidBase58String + "parent"
-	splitPostfix        = invalidBase58String + "splitid"
-
-	userAttributePostfix = invalidBase58String + "attr_"
-
-	splitInfoError *object.SplitInfoError // for errors.As comparisons
 )
 
+// Prefix bytes for database keys. All ids and addresses are encoded in binary
+// unless specified otherwise.
+const (
+	// graveyardPrefix is used for the graveyard bucket.
+	// 	Key: object address
+	// 	Value: tombstone address
+	graveyardPrefix = iota
+	// garbagePrefix is used for the garbage bucket.
+	// 	Key: object address
+	// 	Value: dummy value
+	garbagePrefix
+	// toMoveItPrefix is used for bucket containing IDs of objects that are candidates for moving
+	// to another shard.
+	toMoveItPrefix
+	// containerVolumePrefix is used for storing container size estimations.
+	//	Key: container ID
+	//  Value: container size in bytes as little-endian uint64
+	containerVolumePrefix
+	// lockedPrefix is used for storing locked objects information.
+	//  Key: container ID
+	//  Value: bucket mapping objects locked to the list of corresponding LOCK objects.
+	lockedPrefix
+	// shardInfoPrefix is used for storing shard ID. All keys are custom and are not connected to the container.
+	shardInfoPrefix
+
+	//======================
+	// Unique index buckets.
+	//======================
+
+	// primaryPrefix is used for prefixing buckets containing objects of REGULAR type.
+	//  Key: object ID
+	//  Value: marshalled object
+	primaryPrefix
+	// lockersPrefix is used for prefixing buckets containing objects of LOCK type.
+	//  Key: object ID
+	//  Value: marshalled object
+	lockersPrefix
+	// storageGroupPrefix is used for prefixing buckets containing objects of STORAGEGROUP type.
+	//  Key: object ID
+	//  Value: marshaled object
+	storageGroupPrefix
+	// tombstonePrefix is used for prefixing buckets containing objects of TOMBSTONE type.
+	//  Key: object ID
+	//  Value: marshaled object
+	tombstonePrefix
+	// smallPrefix is used for prefixing buckets mapping objects to the blobovniczas they are stored in.
+	//  Key: object ID
+	//  Value: blobovnicza ID
+	smallPrefix
+	// rootPrefix is used for prefixing buckets mapping parent object to the split info.
+	//  Key: object ID
+	//  Value: split info
+	rootPrefix
+
+	//====================
+	// FKBT index buckets.
+	//====================
+
+	// ownerPrefix is used for prefixing FKBT index buckets mapping owner to object IDs.
+	// Key: owner ID
+	// Value: bucket containing object IDs as keys
+	ownerPrefix
+	// userAttributePrefix is used for prefixing FKBT index buckets containing objects.
+	// Key: attribute value
+	// Value: bucket containing object IDs as keys
+	userAttributePrefix
+
+	//====================
+	// List index buckets.
+	//====================
+
+	// payloadHashPrefix is used for prefixing List index buckets mapping payload hash to a list of object IDs.
+	//  Key: payload hash
+	//  Value: list of object IDs
+	payloadHashPrefix
+	// parentPrefix is used for prefixing List index buckets mapping parent ID to a list of children IDs.
+	//  Key: parent ID
+	//  Value: list of object IDs
+	parentPrefix
+	// splitPrefix is used for prefixing List index buckets mapping split ID to a list of object IDs.
+	//  Key: split ID
+	//  Value: list of object IDs
+	splitPrefix
+)
+
+const (
+	cidSize        = sha256.Size
+	bucketKeySize  = 1 + cidSize
+	objectKeySize  = sha256.Size
+	addressKeySize = cidSize + objectKeySize
+)
+
+var splitInfoError *object.SplitInfoError // for errors.As comparisons
+
+func bucketName(cnr cid.ID, prefix byte, key []byte) []byte {
+	key[0] = prefix
+	cnr.Encode(key[1:])
+	return key[:bucketKeySize]
+}
+
 // primaryBucketName returns <CID>.
-func primaryBucketName(cnr cid.ID) []byte {
-	return []byte(cnr.EncodeToString())
+func primaryBucketName(cnr cid.ID, key []byte) []byte {
+	return bucketName(cnr, primaryPrefix, key)
 }
 
 // tombstoneBucketName returns <CID>_TS.
-func tombstoneBucketName(cnr cid.ID) []byte {
-	return []byte(cnr.EncodeToString() + tombstonePostfix)
+func tombstoneBucketName(cnr cid.ID, key []byte) []byte {
+	return bucketName(cnr, tombstonePrefix, key)
 }
 
 // storageGroupBucketName returns <CID>_SG.
-func storageGroupBucketName(cnr cid.ID) []byte {
-	return []byte(cnr.EncodeToString() + storageGroupPostfix)
+func storageGroupBucketName(cnr cid.ID, key []byte) []byte {
+	return bucketName(cnr, storageGroupPrefix, key)
 }
 
 // smallBucketName returns <CID>_small.
-func smallBucketName(cnr cid.ID) []byte {
-	return []byte(cnr.EncodeToString() + smallPostfix) // consider caching output values
+func smallBucketName(cnr cid.ID, key []byte) []byte {
+	return bucketName(cnr, smallPrefix, key)
 }
 
 // attributeBucketName returns <CID>_attr_<attributeKey>.
-func attributeBucketName(cnr cid.ID, attributeKey string) []byte {
-	sb := strings.Builder{} // consider getting string builders from sync.Pool
-	sb.WriteString(cnr.EncodeToString())
-	sb.WriteString(userAttributePostfix)
-	sb.WriteString(attributeKey)
-
-	return []byte(sb.String())
+func attributeBucketName(cnr cid.ID, attributeKey string, key []byte) []byte {
+	key[0] = userAttributePrefix
+	cnr.Encode(key[1:])
+	return append(key[:bucketKeySize], attributeKey...)
 }
 
 // returns <CID> from attributeBucketName result, nil otherwise.
 func cidFromAttributeBucket(val []byte, attributeKey string) []byte {
-	suffix := []byte(userAttributePostfix + attributeKey)
-	if !bytes.HasSuffix(val, suffix) {
+	if len(val) < bucketKeySize || val[0] != userAttributePrefix || !bytes.Equal(val[bucketKeySize:], []byte(attributeKey)) {
 		return nil
 	}
 
-	return val[:len(val)-len(suffix)]
+	return val[1:bucketKeySize]
 }
 
 // payloadHashBucketName returns <CID>_payloadhash.
-func payloadHashBucketName(cnr cid.ID) []byte {
-	return []byte(cnr.EncodeToString() + payloadHashPostfix)
+func payloadHashBucketName(cnr cid.ID, key []byte) []byte {
+	return bucketName(cnr, payloadHashPrefix, key)
 }
 
 // rootBucketName returns <CID>_root.
-func rootBucketName(cnr cid.ID) []byte {
-	return []byte(cnr.EncodeToString() + rootPostfix)
+func rootBucketName(cnr cid.ID, key []byte) []byte {
+	return bucketName(cnr, rootPrefix, key)
 }
 
 // ownerBucketName returns <CID>_ownerid.
-func ownerBucketName(cnr cid.ID) []byte {
-	return []byte(cnr.EncodeToString() + ownerPostfix)
+func ownerBucketName(cnr cid.ID, key []byte) []byte {
+	return bucketName(cnr, ownerPrefix, key)
 }
 
 // parentBucketName returns <CID>_parent.
-func parentBucketName(cnr cid.ID) []byte {
-	return []byte(cnr.EncodeToString() + parentPostfix)
+func parentBucketName(cnr cid.ID, key []byte) []byte {
+	return bucketName(cnr, parentPrefix, key)
 }
 
 // splitBucketName returns <CID>_splitid.
-func splitBucketName(cnr cid.ID) []byte {
-	return []byte(cnr.EncodeToString() + splitPostfix)
+func splitBucketName(cnr cid.ID, key []byte) []byte {
+	return bucketName(cnr, splitPrefix, key)
 }
 
 // addressKey returns key for K-V tables when key is a whole address.
-func addressKey(addr oid.Address) []byte {
-	return []byte(addr.EncodeToString())
+func addressKey(addr oid.Address, key []byte) []byte {
+	addr.Container().Encode(key)
+	addr.Object().Encode(key[cidSize:])
+	return key[:addressKeySize]
 }
 
 // parses object address formed by addressKey.
 func decodeAddressFromKey(dst *oid.Address, k []byte) error {
-	err := dst.DecodeString(string(k))
-	if err != nil {
-		return fmt.Errorf("decode object address from db key: %w", err)
+	if len(k) != addressKeySize {
+		return fmt.Errorf("invalid length")
 	}
 
+	var cnr cid.ID
+	if err := cnr.Decode(k[:cidSize]); err != nil {
+		return err
+	}
+
+	var obj oid.ID
+	if err := obj.Decode(k[cidSize:]); err != nil {
+		return err
+	}
+
+	dst.SetObject(obj)
+	dst.SetContainer(cnr)
 	return nil
 }
 
 // objectKey returns key for K-V tables when key is an object id.
-func objectKey(obj oid.ID) []byte {
-	return []byte(obj.EncodeToString())
+func objectKey(obj oid.ID, key []byte) []byte {
+	obj.Encode(key)
+	return key[:objectKeySize]
 }
 
 // removes all bucket elements.
@@ -152,13 +240,15 @@ func firstIrregularObjectType(tx *bbolt.Tx, idCnr cid.ID, objs ...[]byte) object
 		panic("empty object list in firstIrregularObjectType")
 	}
 
+	var keys [3][1 + cidSize]byte
+
 	irregularTypeBuckets := [...]struct {
 		typ  object.Type
 		name []byte
 	}{
-		{object.TypeTombstone, tombstoneBucketName(idCnr)},
-		{object.TypeStorageGroup, storageGroupBucketName(idCnr)},
-		{object.TypeLock, bucketNameLockers(idCnr)},
+		{object.TypeTombstone, tombstoneBucketName(idCnr, keys[0][:])},
+		{object.TypeStorageGroup, storageGroupBucketName(idCnr, keys[1][:])},
+		{object.TypeLock, bucketNameLockers(idCnr, keys[2][:])},
 	}
 
 	for i := range objs {
@@ -174,5 +264,7 @@ func firstIrregularObjectType(tx *bbolt.Tx, idCnr cid.ID, objs ...[]byte) object
 
 // return true if provided object is of LOCK type.
 func isLockObject(tx *bbolt.Tx, idCnr cid.ID, obj oid.ID) bool {
-	return inBucket(tx, bucketNameLockers(idCnr), objectKey(obj))
+	return inBucket(tx,
+		bucketNameLockers(idCnr, make([]byte, bucketKeySize)),
+		objectKey(obj, make([]byte, objectKeySize)))
 }
diff --git a/pkg/local_object_storage/metabase/version.go b/pkg/local_object_storage/metabase/version.go
index 78b4ed3f4c..f5af18c126 100644
--- a/pkg/local_object_storage/metabase/version.go
+++ b/pkg/local_object_storage/metabase/version.go
@@ -8,7 +8,7 @@ import (
 )
 
 // version contains current metabase version.
-const version = 1
+const version = 2
 
 var versionKey = []byte("version")