diff --git a/backend/union/policy/all.go b/backend/union/policy/all.go
new file mode 100644
index 000000000..42f7a263b
--- /dev/null
+++ b/backend/union/policy/all.go
@@ -0,0 +1,44 @@
+package policy
+
+import (
+	"context"
+	
+	"github.com/rclone/rclone/fs"
+	"github.com/rclone/rclone/backend/union/upstream"
+)
+
+func init(){
+	registerPolicy("all", &All{})
+}
+
+// All policy behaves the same as EpAll except for the CREATE category
+// Action category: same as epall.
+// Create category: apply to all branches.
+// Search category: same as epall.
+type All struct {
+	EpAll
+}
+
+// Create category policy, governing the creation of files and directories
+func (p *All) Create(ctx context.Context, upstreams []*upstream.Fs, path string) ([]*upstream.Fs, error) {
+	if len(upstreams) == 0 {
+		return nil, fs.ErrorObjectNotFound
+	}
+	upstreams = filterNC(upstreams)
+	if len(upstreams) == 0 {
+		return nil, fs.ErrorPermissionDenied
+	}
+	return upstreams, nil
+}
+
+// CreateEntries is CREATE category policy but receving a set of candidate entries
+func (p *All) CreateEntries(entries ...upstream.Entry) ([]upstream.Entry, error) {
+	if len(entries) == 0 {
+		return nil, fs.ErrorObjectNotFound
+	}
+	entries = filterNCEntries(entries)
+	if len(entries) == 0 {
+		return nil, fs.ErrorPermissionDenied
+	}
+	return entries, nil
+}
\ No newline at end of file
diff --git a/backend/union/policy/epall.go b/backend/union/policy/epall.go
new file mode 100644
index 000000000..e1c5f9d34
--- /dev/null
+++ b/backend/union/policy/epall.go
@@ -0,0 +1,72 @@
+package policy
+
+import (
+	"context"
+	"sync"
+	
+	"github.com/rclone/rclone/fs"
+	"github.com/rclone/rclone/backend/union/upstream"
+)
+
+func init(){
+	registerPolicy("epall", &EpAll{})
+}
+
+// EpAll stands for existing path, All
+// Action category: apply to all found.
+// Create category: apply to all found.
+// Search category: same as epff.
+type EpAll struct {
+	EpFF
+}
+
+func (p *EpAll) epall(ctx context.Context, upstreams []*upstream.Fs, path string) ([]*upstream.Fs, error) {
+	var wg sync.WaitGroup
+	ufs := make([]*upstream.Fs, len(upstreams))
+	for i, u := range upstreams {
+		wg.Add(1)
+		i, u := i, u // Closure
+		go func() {
+			if exists(ctx, u, path) {
+				ufs[i] = u
+			}
+			wg.Done()
+		}()
+	}
+	wg.Wait()
+	var results []*upstream.Fs
+	for _, f := range ufs {
+		if f != nil {
+			results = append(results, f)
+		}
+	}
+	if len(results) == 0 {
+		return nil, fs.ErrorObjectNotFound
+	}
+	return results, nil
+}
+
+// Action category policy, governing the modification of files and directories
+func (p *EpAll) Action(ctx context.Context, upstreams []*upstream.Fs, path string) ([]*upstream.Fs, error) {
+	if len(upstreams) == 0 {
+		return nil, fs.ErrorObjectNotFound
+	}
+	upstreams = filterRO(upstreams)
+	if len(upstreams) == 0 {
+		return nil, fs.ErrorPermissionDenied
+	}
+	return p.epall(ctx, upstreams, path)
+}
+
+// Create category policy, governing the creation of files and directories
+func (p *EpAll) Create(ctx context.Context, upstreams []*upstream.Fs, path string) ([]*upstream.Fs, error) {
+	if len(upstreams) == 0 {
+		return nil, fs.ErrorObjectNotFound
+	}
+	upstreams = filterNC(upstreams)
+	if len(upstreams) == 0 {
+		return nil, fs.ErrorPermissionDenied
+	}
+	upstreams, err := p.epall(ctx, upstreams, path)
+	return upstreams, err
+}
\ No newline at end of file
diff --git a/backend/union/policy/epff.go b/backend/union/policy/epff.go
new file mode 100644
index 000000000..2da8b3f5a
--- /dev/null
+++ b/backend/union/policy/epff.go
@@ -0,0 +1,112 @@
+package policy
+
+import (
+	"context"
+	
+	"github.com/rclone/rclone/fs"
+	"github.com/rclone/rclone/backend/union/upstream"
+)
+
+func init(){
+	registerPolicy("epff", &EpFF{})
+}
+
+// EpFF stands for existing path, first found
+// Given the order of the candidates, act on the first one found where the relative path exists.
+type EpFF struct {}
+
+func (p *EpFF) epff(ctx context.Context, upstreams []*upstream.Fs, path string) (*upstream.Fs, error) {
+	ch := make(chan *upstream.Fs)
+	for _, r := range upstreams {
+		r := r // Closure
+		go func() {
+			if !exists(ctx, r, path) {
+				r = nil
+			}
+			ch <- r
+		}()
+	}
+	var r *upstream.Fs
+	for i := 0; i < len(upstreams); i++ {
+		r = <- ch
+		if r != nil {
+			// close remaining goroutines
+			go func(num int) {
+				defer close(ch)
+				for i := 0; i < num; i++ {
+					<- ch
+				}
+			}(len(upstreams) - 1 - i)
+		}
+	}
+	if r == nil {
+		return nil, fs.ErrorObjectNotFound
+	}
+	return r, nil
+}
+
+// Action category policy, governing the modification of files and directories
+func (p *EpFF) Action(ctx context.Context, upstreams []*upstream.Fs, path string) ([]*upstream.Fs, error) {
+	if len(upstreams) == 0 {
+		return nil, fs.ErrorObjectNotFound
+	}
+	upstreams = filterRO(upstreams)
+	if len(upstreams) == 0 {
+		return nil, fs.ErrorPermissionDenied
+	}
+	r, err := p.epff(ctx, upstreams, path)
+	return []*upstream.Fs{r}, err
+}
+
+// ActionEntries is ACTION category policy but receving a set of candidate entries
+func (p *EpFF) ActionEntries(entries ...upstream.Entry) ([]upstream.Entry, error) {
+	if len(entries) == 0 {
+		return nil, fs.ErrorObjectNotFound
+	}
+	entries = filterROEntries(entries)
+	if len(entries) == 0 {
+		return nil, fs.ErrorPermissionDenied
+	}
+	return entries[:1], nil
+}
+
+// Create category policy, governing the creation of files and directories
+func (p *EpFF) Create(ctx context.Context, upstreams []*upstream.Fs, path string) ([]*upstream.Fs, error) {
+	if len(upstreams) == 0 {
+		return nil, fs.ErrorObjectNotFound
+	}
+	upstreams = filterNC(upstreams)
+	if len(upstreams) == 0 {
+		return nil, fs.ErrorPermissionDenied
+	}
+	r, err := p.epff(ctx, upstreams, path)
+	return []*upstream.Fs{r}, err
+}
+
+// CreateEntries is CREATE category policy but receving a set of candidate entries
+func (p *EpFF) CreateEntries(entries ...upstream.Entry) ([]upstream.Entry, error) {
+	if len(entries) == 0 {
+		return nil, fs.ErrorObjectNotFound
+	}
+	entries = filterNCEntries(entries)
+	if len(entries) == 0 {
+		return nil, fs.ErrorPermissionDenied
+	}
+	return entries[:1], nil
+}
+
+// Search category policy, governing the access to files and directories
+func (p *EpFF) Search(ctx context.Context, upstreams []*upstream.Fs, path string) (*upstream.Fs, error) {
+	if len(upstreams) == 0 {
+		return nil, fs.ErrorObjectNotFound
+	}
+	return p.epff(ctx, upstreams, path)
+}
+
+// SearchEntries is SEARCH category policy but receving a set of candidate entries
+func (p *EpFF) SearchEntries(entries ...upstream.Entry) (upstream.Entry, error) {
+	if len(entries) == 0 {
+		return nil, fs.ErrorObjectNotFound
+	}
+	return entries[0], nil
+}
\ No newline at end of file
diff --git a/backend/union/policy/epmfs.go b/backend/union/policy/epmfs.go
new file mode 100644
index 000000000..4693c8e9a
--- /dev/null
+++ b/backend/union/policy/epmfs.go
@@ -0,0 +1,110 @@
+package policy
+
+import (
+	"context"
+	
+	"github.com/rclone/rclone/fs"
+	"github.com/rclone/rclone/backend/union/upstream"
+)
+
+func init(){
+	registerPolicy("epmfs", &EpMfs{})
+}
+
+// EpMfs stands for existing path, most free space
+// Of all the branches on which the path exists choose the drive with the most free space.
+type EpMfs struct {
+	EpAll
+}
+
+func (p *EpMfs) mfs(upstreams []*upstream.Fs) (*upstream.Fs, error) {
+	var maxFreeSpace int64
+	var mfsupstream *upstream.Fs
+	for _, r := range upstreams {
+		space, err := r.GetFreeSpace()
+		if err != nil {
+			return nil, err
+		}
+		if maxFreeSpace < space {
+			maxFreeSpace = space
+			mfsupstream = r
+		}
+	}
+	return mfsupstream, nil
+}
+
+func (p *EpMfs) mfsEntries(entries []upstream.Entry) (upstream.Entry, error) {
+	var maxFreeSpace int64
+	var mfsEntry upstream.Entry
+	for _, e := range entries {
+		space, err := e.UpstreamFs().GetFreeSpace()
+		if err != nil {
+			return nil, err
+		}
+		if maxFreeSpace < space {
+			maxFreeSpace = space
+			mfsEntry = e
+		}
+	}
+	return mfsEntry, nil
+}
+
+// Action category policy, governing the modification of files and directories
+func (p *EpMfs) Action(ctx context.Context, upstreams []*upstream.Fs, path string) ([]*upstream.Fs, error) {
+	upstreams, err := p.EpAll.Action(ctx, upstreams, path)
+	if err != nil {
+		return nil, err
+	}
+	r, err := p.mfs(upstreams)
+	return []*upstream.Fs{r}, err
+}
+
+// ActionEntries is ACTION category policy but receving a set of candidate entries
+func (p *EpMfs) ActionEntries(entries ...upstream.Entry) ([]upstream.Entry, error) {
+	entries, err := p.EpAll.ActionEntries(entries...)
+	if err != nil {
+		return nil, err
+	}
+	e, err := p.mfsEntries(entries)
+	return []upstream.Entry{e}, err
+}
+
+// Create category policy, governing the creation of files and directories
+func (p *EpMfs) Create(ctx context.Context, upstreams []*upstream.Fs, path string) ([]*upstream.Fs, error) {
+	upstreams, err := p.EpAll.Create(ctx, upstreams, path)
+	if err != nil {
+		return nil, err
+	}
+	r, err := p.mfs(upstreams)
+	return []*upstream.Fs{r}, err
+}
+
+// CreateEntries is CREATE category policy but receving a set of candidate entries
+func (p *EpMfs) CreateEntries(entries ...upstream.Entry) ([]upstream.Entry, error) {
+	entries, err := p.EpAll.CreateEntries(entries...)
+	if err != nil {
+		return nil, err
+	}
+	e, err := p.mfsEntries(entries)
+	return []upstream.Entry{e}, err
+}
+
+// Search category policy, governing the access to files and directories
+func (p *EpMfs) Search(ctx context.Context, upstreams []*upstream.Fs, path string) (*upstream.Fs, error) {
+	if len(upstreams) == 0 {
+		return nil, fs.ErrorObjectNotFound
+	}
+	upstreams, err := p.epall(ctx, upstreams, path)
+	if err != nil {
+		return nil, err
+	}
+	return p.mfs(upstreams)
+}
+
+// SearchEntries is SEARCH category policy but receving a set of candidate entries
+func (p *EpMfs) SearchEntries(entries ...upstream.Entry) (upstream.Entry, error) {
+	if len(entries) == 0 {
+		return nil, fs.ErrorObjectNotFound
+	}
+	return p.mfsEntries(entries)
+}
\ No newline at end of file
diff --git a/backend/union/policy/ff.go b/backend/union/policy/ff.go
new file mode 100644
index 000000000..87be6904a
--- /dev/null
+++ b/backend/union/policy/ff.go
@@ -0,0 +1,32 @@
+package policy
+
+import (
+	"context"
+	
+	"github.com/rclone/rclone/fs"
+	"github.com/rclone/rclone/backend/union/upstream"
+)
+
+func init(){
+	registerPolicy("ff", &FF{})
+}
+
+// FF stands for first found
+// Search category: same as epff.
+// Action category: same as epff.
+// Create category: Given the order of the candiates, act on the first one found.
+type FF struct {
+	EpFF
+}
+
+// Create category policy, governing the creation of files and directories
+func (p *FF) Create(ctx context.Context, upstreams []*upstream.Fs, path string) ([]*upstream.Fs, error) {
+	if len(upstreams) == 0 {
+		return nil, fs.ErrorObjectNotFound
+	}
+	upstreams = filterNC(upstreams)
+	if len(upstreams) == 0 {
+		return upstreams, fs.ErrorPermissionDenied
+	}
+	return upstreams[:1], nil
+}
\ No newline at end of file
diff --git a/backend/union/policy/mfs.go b/backend/union/policy/mfs.go
new file mode 100644
index 000000000..3143ca2ec
--- /dev/null
+++ b/backend/union/policy/mfs.go
@@ -0,0 +1,31 @@
+package policy
+
+import (
+	"context"
+
+	"github.com/rclone/rclone/fs"
+	"github.com/rclone/rclone/backend/union/upstream"
+)
+
+func init(){
+	registerPolicy("mfs", &Mfs{})
+}
+
+// Mfs stands for most free space
+// Of all the candidates on which the path exists choose the one with the most free space.
+type Mfs struct {
+	EpMfs
+}
+
+// Create category policy, governing the creation of files and directories
+func (p *Mfs) Create(ctx context.Context, upstreams []*upstream.Fs, path string) ([]*upstream.Fs, error) {
+	if len(upstreams) == 0 {
+		return nil, fs.ErrorObjectNotFound
+	}
+	upstreams = filterNC(upstreams)
+	if len(upstreams) == 0 {
+		return nil, fs.ErrorPermissionDenied
+	}
+	r, err := p.mfs(upstreams)
+	return []*upstream.Fs{r}, err
+}
diff --git a/backend/union/policy/policy.go b/backend/union/policy/policy.go
new file mode 100644
index 000000000..91651d497
--- /dev/null
+++ b/backend/union/policy/policy.go
@@ -0,0 +1,123 @@
+package policy
+
+import (
+	"context"
+	"strings"
+	"path"
+	"path/filepath"
+
+	"github.com/pkg/errors"
+	"github.com/rclone/rclone/fs"
+	"github.com/rclone/rclone/backend/union/upstream"
+)
+
+var policies = make(map[string]Policy)
+
+// Policy is the interface of a set of defined behavior choosing
+// the upstream Fs to operate on
+type Policy interface {
+	// Action category policy, governing the modification of files and directories
+	Action(ctx context.Context, upstreams []*upstream.Fs, path string) ([]*upstream.Fs, error)
+
+	// Create category policy, governing the creation of files and directories
+	Create(ctx context.Context, upstreams []*upstream.Fs, path string) ([]*upstream.Fs, error)
+
+	// Search category policy, governing the access to files and directories
+	Search(ctx context.Context, upstreams []*upstream.Fs, path string) (*upstream.Fs, error)
+
+	// ActionEntries is ACTION category policy but receving a set of candidate entries
+	ActionEntries(entries ...upstream.Entry) ([]upstream.Entry, error)
+
+	// CreateEntries is CREATE category policy but receving a set of candidate entries
+	CreateEntries(entries ...upstream.Entry) ([]upstream.Entry, error)
+
+	// SearchEntries is SEARCH category policy but receving a set of candidate entries
+	SearchEntries(entries ...upstream.Entry) (upstream.Entry, error)
+}
+
+func registerPolicy(name string, p Policy) {
+	policies[strings.ToLower(name)] = p
+}
+
+// Get a Policy from the list
+func Get(name string) (Policy, error) {
+	p, ok := policies[strings.ToLower(name)]
+	if !ok {
+		return nil, errors.Errorf("didn't find policy called %q", name)
+	}
+	return p, nil
+}
+
+func filterRO(ufs []*upstream.Fs) (wufs []*upstream.Fs) {
+	for _, u := range ufs {
+		if u.IsWritable() {
+			wufs = append(wufs, u)
+		}
+	}
+	return wufs
+}
+
+func filterROEntries(ue []upstream.Entry) (wue []upstream.Entry) {
+	for _, e := range ue {
+		if e.UpstreamFs().IsWritable() {
+			wue = append(wue, e)
+		}
+	}
+	return wue
+}
+
+func filterNC(ufs []*upstream.Fs) (wufs []*upstream.Fs) {
+	for _, u := range ufs {
+		if u.IsCreatable() {
+			wufs = append(wufs, u)
+		}
+	}
+	return wufs
+}
+
+func filterNCEntries(ue []upstream.Entry) (wue []upstream.Entry) {
+	for _, e := range ue {
+		if e.UpstreamFs().IsCreatable() {
+			wue = append(wue, e)
+		}
+	}
+	return wue
+}
+
+func parentDir(absPath string) string {
+	parent := path.Dir(strings.TrimRight(absPath, "/"))
+	if parent == "." {
+		parent = ""
+	}
+	return parent
+}
+
+func clean(absPath string) string {
+	cleanPath := path.Clean(filepath.ToSlash(absPath))
+	if cleanPath == "." {
+		cleanPath = ""
+	}
+	return cleanPath
+}
+
+func exists(ctx context.Context, f fs.Fs, remote string) bool {
+	remote = clean(remote)
+	dir := parentDir(remote)
+	if remote == dir {
+		return true
+	}
+	found := false
+	entries, _ := f.List(ctx, dir);
+	for _, e := range entries {
+		eRemote := e.Remote()
+		if f.Features().CaseInsensitive {
+			found = strings.EqualFold(remote, eRemote)
+		} else {
+			found = (remote == eRemote)
+		}
+		if found {
+			break
+		}
+	}
+	return found
+}
diff --git a/backend/union/union.go b/backend/union/union.go
index 8934a2526..2b7db71c2 100644
--- a/backend/union/union.go
+++ b/backend/union/union.go
@@ -2,16 +2,20 @@ package union
 
 import (
 	"context"
+	"bufio"
 	"fmt"
 	"io"
 	"path"
 	"path/filepath"
 	"strings"
+	"sync"
+	"sync/atomic"
 	"time"
 
 	"github.com/pkg/errors"
+	"github.com/rclone/rclone/backend/union/upstream"
+	"github.com/rclone/rclone/backend/union/policy"
 	"github.com/rclone/rclone/fs"
-	"github.com/rclone/rclone/fs/cache"
 	"github.com/rclone/rclone/fs/config/configmap"
 	"github.com/rclone/rclone/fs/config/configstruct"
 	"github.com/rclone/rclone/fs/hash"
@@ -21,12 +25,32 @@ import (
 func init() {
 	fsi := &fs.RegInfo{
 		Name:        "union",
-		Description: "Union merges the contents of several remotes",
+		Description: "Union merges the contents of several upstream fs",
 		NewFs:       NewFs,
 		Options: []fs.Option{{
-			Name:     "remotes",
-			Help:     "List of space separated remotes.\nCan be 'remotea:test/dir remoteb:', '\"remotea:test/space dir\" remoteb:', etc.\nThe last remote is used to write to.",
+			Name:     "upstreams",
+			Help:     "List of space separated upstreams.\nCan be 'upstreama:test/dir upstreamb:', '\"remotea:test/space:ro dir\" remoteb:', etc.\n",
 			Required: true,
+		}, {
+			Name:     "action_policy",
+			Help:     "Policy to choose upstream on ACTION class.",
+			Required: true,
+			Default:  "epall",
+		}, {
+			Name:     "create_policy",
+			Help:     "Policy to choose upstream on CREATE class.",
+			Required: true,
+			Default:  "epmfs",
+		}, {
+			Name:     "search_policy",
+			Help:     "Policy to choose upstream on SEARCH class.",
+			Required: true,
+			Default:  "ff",
+		}, {
+			Name:     "cache_time",
+			Help:     "Cache time of usage and free space (in seconds)",
+			Required: true,
+			Default:  120,
 		}},
 	}
 	fs.Register(fsi)
@@ -34,34 +58,75 @@ func init() {
 
 // Options defines the configuration for this backend
 type Options struct {
-	Remotes fs.SpaceSepList `config:"remotes"`
+	Upstreams    fs.SpaceSepList `config:"upstreams"`
+	ActionPolicy string          `config:"action_policy"`
+	CreatePolicy string          `config:"create_policy"`
+	SearchPolicy string          `config:"search_policy"`
+	CacheTime    int             `config:"cache_time"`
 }
 
-// Fs represents a union of remotes
+// Fs represents a union of upstreams
 type Fs struct {
-	name     string       // name of this remote
-	features *fs.Features // optional features
-	opt      Options      // options for this Fs
-	root     string       // the path we are working on
-	remotes  []fs.Fs      // slice of remotes
-	wr       fs.Fs        // writable remote
-	hashSet  hash.Set     // intersection of hash types
+	name         string            // name of this remote
+	features     *fs.Features      // optional features
+	opt          Options           // options for this Fs
+	root         string            // the path we are working on
+	upstreams    []*upstream.Fs    // slice of upstreams
+	hashSet      hash.Set          // intersection of hash types
+	actionPolicy policy.Policy     // policy for ACTION
+	createPolicy policy.Policy     // policy for CREATE
+	searchPolicy policy.Policy     // policy for SEARCH
 }
 
 // Object describes a union Object
 //
 // This is a wrapped object which returns the Union Fs as its parent
 type Object struct {
-	fs.Object
+	*upstream.Object
 	fs *Fs // what this object is part of
+	co []upstream.Entry
 }
 
-// Wrap an existing object in the union Object
-func (f *Fs) wrapObject(o fs.Object) *Object {
-	return &Object{
-		Object: o,
-		fs:     f,
+// Directory describes a union Directory
+//
+// This is a wrapped object contains all candidates
+type Directory struct {
+	*upstream.Directory
+	cd []upstream.Entry
+}
+
+type entry interface {
+	upstream.Entry
+	candidates() []upstream.Entry
+}
+
+// Wrap candidate objects in to an union Object
+func (f *Fs) wrapEntries(entries ...upstream.Entry) (entry, error) {
+	e, err := f.searchEntries(entries...)
+	if err != nil {
+		return nil, err
 	}
+	switch e.(type) {
+	case *upstream.Object:
+		return &Object {
+			Object:     e.(*upstream.Object),
+			fs:         f,
+			co: entries,
+		}, nil
+	case *upstream.Directory:
+		return &Directory {
+			Directory:     e.(*upstream.Directory),
+			cd: entries,
+		}, nil
+	default:
+		return nil, errors.Errorf("unknown object type %T", e)
+	}
+}
+
+// UnWrap returns the Object that this Object is wrapping or
+// nil if it isn't wrapping anything
+func (o *Object) UnWrap() *upstream.Object {
+	return o.Object
 }
 
 // Fs returns the union Fs as the parent
@@ -69,6 +134,133 @@ func (o *Object) Fs() fs.Info {
 	return o.fs
 }
 
+func (o *Object) candidates() []upstream.Entry {
+	return o.co
+}
+
+func (d *Directory) candidates() []upstream.Entry {
+	return d.cd
+}
+
+// Update in to the object with the modTime given of the given size
+//
+// When called from outside a Fs by rclone, src.Size() will always be >= 0.
+// But for unknown-sized objects (indicated by src.Size() == -1), Upload should either
+// return an error or update the object properly (rather than e.g. calling panic).
+func (o *Object) Update(ctx context.Context, in io.Reader, src fs.ObjectInfo, options ...fs.OpenOption) error {
+	f := o.Fs().(*Fs)
+	entries, err := f.actionEntries(o.candidates()...)
+	if err != nil {
+		return err
+	}
+	// Get mutliple reader
+	readers := make([]io.Reader, len(entries))
+	writers := make([]io.Writer, len(entries))
+	errs := make([]error, len(entries) + 1)
+	for i := range entries {
+		r, w := io.Pipe()
+		bw := bufio.NewWriter(w)
+		readers[i], writers[i] = r, bw
+		defer func () {
+			w.Close()
+		}()
+	}
+	go func() {
+		mw := io.MultiWriter(writers...)
+        _, errs[len(entries)] = io.Copy(mw, in)
+		for _, bw := range writers {
+			bw.(*bufio.Writer).Flush()
+		}
+	}()
+	// Multi-threading
+	multithread(len(entries), func(i int){
+		if o, ok := entries[i].(*upstream.Object); ok {
+			errs[i] = o.Update(ctx, readers[i], src, options...)
+		} else {
+			errs[i] = fs.ErrorNotAFile
+		}
+	})
+	// Get an object for future operation
+	for _, err := range errs {
+		if err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+// Remove candidate objects selected by ACTION policy
+func (o *Object) Remove(ctx context.Context) error {
+	f := o.Fs().(*Fs)
+	entries, err := f.actionEntries(o.candidates()...)
+	if err != nil {
+		return err
+	}
+	errs := make([]error, len(entries))
+	multithread(len(entries), func(i int){
+		if o, ok := entries[i].(*upstream.Object); ok {
+			errs[i] = o.Remove(ctx)
+		} else {
+			errs[i] = fs.ErrorNotAFile
+		}
+	})
+	for _, err := range errs {
+		if err != nil {
+			return err
+		}
+	}
+	return nil
+}
+// SetModTime sets the metadata on the object to set the modification date
+func (o *Object) SetModTime(ctx context.Context, t time.Time) error {
+	f := o.Fs().(*Fs)
+	entries, err := f.actionEntries(o.candidates()...)
+	if err != nil {
+		return err
+	}
+	var wg sync.WaitGroup
+	errs := make([]error, len(entries))
+	multithread(len(entries), func(i int){
+		if o, ok := entries[i].(*upstream.Object); ok {
+			errs[i] = o.SetModTime(ctx, t)
+		} else {
+			errs[i] = fs.ErrorNotAFile
+		}
+	})
+	wg.Wait()
+	for _, err := range errs {
+		if err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+// ModTime returns the modification date of the directory
+// It returns the latest ModTime of all candidates
+func (d *Directory) ModTime(ctx context.Context) (t time.Time) {
+	entries := d.candidates()
+	times := make([]time.Time, len(entries))
+	multithread(len(entries), func(i int){
+		times[i] = entries[i].ModTime(ctx)
+	})
+	for _, ti := range times {
+		if t.Before(ti) {
+			t = ti
+		}
+	}
+	return t
+}
+
+// Size returns the size of the directory
+// It returns the sum of all candidates
+func (d *Directory) Size() (s int64) {
+	for _, e := range d.candidates() {
+		s += e.Size()
+	}
+	return s
+}
+
 // Name of the remote (as passed into NewFs)
 func (f *Fs) Name() string {
 	return f.name
@@ -91,7 +283,20 @@ func (f *Fs) Features() *fs.Features {
 
 // Rmdir removes the root directory of the Fs object
 func (f *Fs) Rmdir(ctx context.Context, dir string) error {
-	return f.wr.Rmdir(ctx, dir)
+	upstreams, err := f.action(ctx, dir)
+	if err != nil {
+		return err
+	}
+	errs := make([]error, len(upstreams))
+	multithread(len(upstreams), func(i int){
+		errs[i] = upstreams[i].Rmdir(ctx, dir)
+	})
+	for _, err := range errs {
+		if err != nil {
+			return err
+		}
+	}
+	return nil
 }
 
 // Hashes returns hash.HashNone to indicate remote hashing is unavailable
@@ -101,7 +306,21 @@ func (f *Fs) Hashes() hash.Set {
 
 // Mkdir makes the root directory of the Fs object
 func (f *Fs) Mkdir(ctx context.Context, dir string) error {
-	return f.wr.Mkdir(ctx, dir)
+	parent := parentDir(dir)
+	upstreams, err := f.create(ctx, parent)
+	if err != nil {
+		return err
+	}
+	errs := make([]error, len(upstreams))
+	multithread(len(upstreams), func(i int){
+		errs[i] = upstreams[i].Mkdir(ctx, dir)
+	})
+	for _, err := range errs {
+		if err != nil {
+			return err
+		}
+	}
+	return nil
 }
 
 // Purge all files in the root and the root directory
@@ -111,7 +330,27 @@ func (f *Fs) Mkdir(ctx context.Context, dir string) error {
 //
 // Return an error if it doesn't exist
 func (f *Fs) Purge(ctx context.Context) error {
-	return f.wr.Features().Purge(ctx)
+	for _, r := range f.upstreams {
+		if r.Features().Purge == nil {
+			return fs.ErrorCantPurge
+		}
+	}
+	for _, r := range f.upstreams {
+		// Can't Purge if any read-only upstreams
+		if !r.IsWritable() {
+			return fs.ErrorPermissionDenied
+		}
+	}
+	errs := make([]error, len(f.upstreams))
+	multithread(len(f.upstreams), func(i int){
+		errs[i] = f.upstreams[i].Features().Purge(ctx)
+	})
+	for _, err := range errs {
+		if err != nil {
+			return err
+		}
+	}
+	return nil
 }
 
 // Copy src to this remote using server side copy operations.
@@ -124,15 +363,26 @@ func (f *Fs) Purge(ctx context.Context) error {
 //
 // If it isn't possible then return fs.ErrorCantCopy
 func (f *Fs) Copy(ctx context.Context, src fs.Object, remote string) (fs.Object, error) {
-	if src.Fs() != f.wr {
+	srcObj, ok := src.(*Object)
+	if !ok {
 		fs.Debugf(src, "Can't copy - not same remote type")
 		return nil, fs.ErrorCantCopy
 	}
-	o, err := f.wr.Features().Copy(ctx, src, remote)
-	if err != nil {
+	o := srcObj.UnWrap()
+	u := o.UpstreamFs()
+	do := u.Features().Copy
+	if do == nil {
+		return nil, fs.ErrorCantCopy
+	}
+	if !u.IsCreatable() {
+		return nil, fs.ErrorPermissionDenied
+	}
+	co, err := do(ctx, o, remote)
+	if err != nil || co == nil {
 		return nil, err
 	}
-	return f.wrapObject(o), nil
+	wo, err := f.wrapEntries(u.WrapObject(co))
+	return wo.(*Object), err
 }
 
 // Move src to this remote using server side move operations.
@@ -145,15 +395,52 @@ func (f *Fs) Copy(ctx context.Context, src fs.Object, remote string) (fs.Object,
 //
 // If it isn't possible then return fs.ErrorCantMove
 func (f *Fs) Move(ctx context.Context, src fs.Object, remote string) (fs.Object, error) {
-	if src.Fs() != f.wr {
+	o, ok := src.(*Object)
+	if !ok {
 		fs.Debugf(src, "Can't move - not same remote type")
 		return nil, fs.ErrorCantMove
 	}
-	o, err := f.wr.Features().Move(ctx, src, remote)
+	entries, err := f.actionEntries(o.candidates()...)
 	if err != nil {
 		return nil, err
 	}
-	return f.wrapObject(o), err
+	for _, e := range entries {
+		if e.UpstreamFs().Features().Move == nil {
+			return nil, fs.ErrorCantMove
+		}
+	}
+	objs := make([]*upstream.Object, len(entries))
+	errs := make([]error, len(entries))
+	multithread(len(entries), func(i int){
+		u := entries[i].UpstreamFs()
+        o, ok := entries[i].(*upstream.Object);
+		if !ok {
+			errs[i] = fs.ErrorNotAFile
+			return
+		}
+		mo, err := u.Features().Move(ctx, o, remote)
+		if err != nil || mo == nil {
+			errs[i] = err
+			return
+		}
+		objs[i] = u.WrapObject(mo)
+	})
+	var en []upstream.Entry
+	for _, o := range objs {
+		if o != nil {
+			en = append(en, o)
+		}
+	}
+	e, err := f.wrapEntries(en...)
+	if err != nil {
+		return nil, err
+	}
+	for _, err := range errs {
+		if err != nil {
+			return e.(*Object), err
+		}
+	}
+	return e.(*Object), nil
 }
 
 // DirMove moves src, srcRemote to this remote at dstRemote
@@ -165,12 +452,31 @@ func (f *Fs) Move(ctx context.Context, src fs.Object, remote string) (fs.Object,
 //
 // If destination exists then return fs.ErrorDirExists
 func (f *Fs) DirMove(ctx context.Context, src fs.Fs, srcRemote, dstRemote string) error {
-	srcFs, ok := src.(*Fs)
+	_, ok := src.(*Fs)
 	if !ok {
-		fs.Debugf(srcFs, "Can't move directory - not same remote type")
+		fs.Debugf(src, "Can't move directory - not same remote type")
 		return fs.ErrorCantDirMove
 	}
-	return f.wr.Features().DirMove(ctx, srcFs.wr, srcRemote, dstRemote)
+	upstreams, err := f.action(ctx, srcRemote)
+	if err != nil {
+		return err
+	}
+	for _, u := range upstreams {
+		if u.Features().DirMove == nil {
+			return fs.ErrorCantDirMove
+		}
+	}
+	errs := make([]error, len(upstreams))
+	multithread(len(upstreams), func(i int){
+		u := upstreams[i]
+		errs[i] = u.Features().DirMove(ctx, u, srcRemote, dstRemote)
+	})
+	for _, err := range errs {
+		if err != nil {
+			return err
+		}
+	}
+	return nil
 }
 
 // ChangeNotify calls the passed function with a path
@@ -183,23 +489,23 @@ func (f *Fs) DirMove(ctx context.Context, src fs.Fs, srcRemote, dstRemote string
 // regularly. When the channel gets closed, the implementation
 // should stop polling and release resources.
 func (f *Fs) ChangeNotify(ctx context.Context, fn func(string, fs.EntryType), ch <-chan time.Duration) {
-	var remoteChans []chan time.Duration
+	var uChans []chan time.Duration
 
-	for _, remote := range f.remotes {
-		if ChangeNotify := remote.Features().ChangeNotify; ChangeNotify != nil {
+	for _, u := range f.upstreams {
+		if ChangeNotify := u.Features().ChangeNotify; ChangeNotify != nil {
 			ch := make(chan time.Duration)
-			remoteChans = append(remoteChans, ch)
+			uChans = append(uChans, ch)
 			ChangeNotify(ctx, fn, ch)
 		}
 	}
 
 	go func() {
 		for i := range ch {
-			for _, c := range remoteChans {
+			for _, c := range uChans {
 				c <- i
 			}
 		}
-		for _, c := range remoteChans {
+		for _, c := range uChans {
 			close(c)
 		}
 	}()
@@ -208,29 +514,81 @@ func (f *Fs) ChangeNotify(ctx context.Context, fn func(string, fs.EntryType), ch
 // DirCacheFlush resets the directory cache - used in testing
 // as an optional interface
 func (f *Fs) DirCacheFlush() {
-	for _, remote := range f.remotes {
-		if DirCacheFlush := remote.Features().DirCacheFlush; DirCacheFlush != nil {
-			DirCacheFlush()
+	multithread(len(f.upstreams), func(i int){
+		do := f.upstreams[i].Features().DirCacheFlush;
+		if do != nil {
+			do()
+		}
+	})
+}
+
+func (f *Fs) put(ctx context.Context, in io.Reader, src fs.ObjectInfo, stream bool, options ...fs.OpenOption) (fs.Object, error) {
+	srcPath := src.Remote()
+	u, err := f.search(ctx, srcPath)
+	upstreams := []*upstream.Fs{u}
+	if err != nil || len(upstreams) == 0 {
+		upstreams, err = f.create(ctx, parentDir(srcPath))
+		if err != nil {
+			return nil, err
 		}
 	}
-}
-
-// PutStream uploads to the remote path with the modTime given of indeterminate size
-//
-// May create the object even if it returns an error - if so
-// will return the object and the error, otherwise will return
-// nil and the error
-func (f *Fs) PutStream(ctx context.Context, in io.Reader, src fs.ObjectInfo, options ...fs.OpenOption) (fs.Object, error) {
-	o, err := f.wr.Features().PutStream(ctx, in, src, options...)
-	if err != nil {
-		return nil, err
+	// Get mutliple reader
+	readers := make([]*io.PipeReader, len(upstreams))
+	writers := make([]*io.PipeWriter, len(upstreams))
+	errs := make([]error, len(upstreams) + 1)
+	for i := range upstreams {
+		r, w := io.Pipe()
+		readers[i], writers[i] = r, w
 	}
-	return f.wrapObject(o), err
-}
-
-// About gets quota information from the Fs
-func (f *Fs) About(ctx context.Context) (*fs.Usage, error) {
-	return f.wr.Features().About(ctx)
+	go func() {
+		bufw := make([]io.Writer, len(writers))
+		for i, w :=  range writers {
+			bufw[i] = bufio.NewWriter(w)
+		}
+		mw := io.MultiWriter(bufw...)
+        _, err := io.Copy(mw, in)
+        if err != nil {
+            errs[len(upstreams)] = err
+		}
+		for _, bw := range bufw {
+			bw.(*bufio.Writer).Flush()
+		}
+	}()
+	// Multi-threading
+	var wg sync.WaitGroup
+	objs := make([]upstream.Entry, len(upstreams))
+	for i, u := range upstreams {
+		wg.Add(1)
+		i, u := i, u // Closure
+		go func() {
+			defer wg.Done()
+			var o fs.Object
+			var err error
+			if stream {
+				o, err = u.Features().PutStream(ctx, readers[i], src, options...)
+			} else {
+				o, err = u.Put(ctx, readers[i], src, options...)
+			}
+			if err != nil {
+				errs[i] = err
+				return
+			}
+			objs[i] = u.WrapObject(o)
+		}()
+	}
+	wg.Wait()
+	for _, w := range writers {
+		w.Close()
+	}
+	
+	// Get an object for future operation
+	e, _ := f.wrapEntries(objs...)
+	for _, err := range errs {
+		if err != nil {
+			return e.(*Object), err
+		}
+	}
+	return e.(*Object), nil
 }
 
 // Put in to the remote path with the modTime given of the given size
@@ -239,11 +597,65 @@ func (f *Fs) About(ctx context.Context) (*fs.Usage, error) {
 // will return the object and the error, otherwise will return
 // nil and the error
 func (f *Fs) Put(ctx context.Context, in io.Reader, src fs.ObjectInfo, options ...fs.OpenOption) (fs.Object, error) {
-	o, err := f.wr.Put(ctx, in, src, options...)
-	if err != nil {
-		return nil, err
+	return f.put(ctx, in, src, false, options...)
+}
+
+// PutStream uploads to the remote path with the modTime given of indeterminate size
+//
+// May create the object even if it returns an error - if so
+// will return the object and the error, otherwise will return
+// nil and the error
+func (f *Fs) PutStream(ctx context.Context, in io.Reader, src fs.ObjectInfo, options ...fs.OpenOption) (fs.Object, error) {
+	return f.put(ctx, in, src, true, options...)
+}
+
+// About gets quota information from the Fs
+func (f *Fs) About(ctx context.Context) (*fs.Usage, error) {
+	usage := &fs.Usage{
+		Total:   new(int64),
+		Used:    new(int64),
+		Trashed: new(int64),
+		Other:   new(int64),
+		Free:    new(int64),
+		Objects: new(int64),
 	}
-	return f.wrapObject(o), err
+	for _, u := range f.upstreams {
+		usg, err := u.About(ctx)
+		if err != nil {
+			return nil, err
+		}
+		if usg.Total != nil && usage.Total != nil {
+			*usage.Total += *usg.Total
+		} else {
+			usage.Total = nil
+		}
+		if usg.Used != nil && usage.Used != nil {
+			*usage.Used += *usg.Used
+		} else {
+			usage.Used = nil
+		}
+		if usg.Trashed != nil && usage.Trashed != nil {
+			*usage.Trashed += *usg.Trashed
+		} else {
+			usage.Trashed = nil
+		}
+		if usg.Other != nil && usage.Other != nil {
+			*usage.Other += *usg.Other
+		} else {
+			usage.Other = nil
+		}
+		if usg.Free != nil && usage.Free != nil {
+			*usage.Free += *usg.Free
+		} else {
+			usage.Free = nil
+		}
+		if usg.Objects != nil && usage.Objects != nil {
+			*usage.Objects += *usg.Objects
+		} else {
+			usage.Objects = nil
+		}
+	}
+	return usage, nil
 }
 
 // List the objects and directories in dir into entries.  The
@@ -256,60 +668,132 @@ func (f *Fs) Put(ctx context.Context, in io.Reader, src fs.ObjectInfo, options .
 // This should return ErrDirNotFound if the directory isn't
 // found.
 func (f *Fs) List(ctx context.Context, dir string) (entries fs.DirEntries, err error) {
-	set := make(map[string]fs.DirEntry)
-	found := false
-	for _, remote := range f.remotes {
-		var remoteEntries, err = remote.List(ctx, dir)
-		if err == fs.ErrorDirNotFound {
-			continue
+	var found int32
+	entriess := make([][]upstream.Entry, len(f.upstreams))
+	errs := make([]error, len(f.upstreams))
+	multithread(len(f.upstreams), func(i int){
+		u := f.upstreams[i]
+		entries, err := u.List(ctx, dir)
+		if err != nil && err != fs.ErrorDirNotFound {
+			errs[i] = err
+			return
 		}
-		if err != nil {
-			return nil, errors.Wrapf(err, "List failed on %v", remote)
+		atomic.StoreInt32(&found, 1)
+		uEntries := make([]upstream.Entry, len(entries))
+		for j, e := range entries {
+			uEntries[j], _ = u.WrapEntry(e)
 		}
-		found = true
-		for _, remoteEntry := range remoteEntries {
-			set[remoteEntry.Remote()] = remoteEntry
-		}
-	}
-	if !found {
+		entriess[i] = uEntries
+	})
+	if found == 0 {
 		return nil, fs.ErrorDirNotFound
 	}
-	for _, entry := range set {
-		if o, ok := entry.(fs.Object); ok {
-			entry = f.wrapObject(o)
+	entries, err = f.mergeDirEntries(entriess)
+	if err != nil {
+		return entries, err
+	}
+	for _, err := range errs {
+		if err != nil {
+			return entries, err
 		}
-		entries = append(entries, entry)
 	}
 	return entries, nil
 }
 
-// NewObject creates a new remote union file object based on the first Object it finds (reverse remote order)
-func (f *Fs) NewObject(ctx context.Context, path string) (fs.Object, error) {
-	for i := range f.remotes {
-		var remote = f.remotes[len(f.remotes)-i-1]
-		var obj, err = remote.NewObject(ctx, path)
-		if err == fs.ErrorObjectNotFound {
-			continue
+// NewObject creates a new remote union file object
+func (f *Fs) NewObject(ctx context.Context, remote string) (fs.Object, error) {
+	objs := make([]*upstream.Object, len(f.upstreams))
+	errs := make([]error, len(f.upstreams))
+	multithread(len(f.upstreams), func(i int){
+		u := f.upstreams[i]
+		o, err := u.NewObject(ctx, remote)
+		if err != nil && err != fs.ErrorObjectNotFound {
+			errs[i] = err
+			return
 		}
-		if err != nil {
-			return nil, errors.Wrapf(err, "NewObject failed on %v", remote)
+		objs[i] = u.WrapObject(o)
+	})
+	var entries []upstream.Entry
+	for _, o := range objs {
+		if o != nil {
+			entries = append(entries, o)
 		}
-		return f.wrapObject(obj), nil
 	}
-	return nil, fs.ErrorObjectNotFound
+	if len(entries) == 0 {
+		return nil, fs.ErrorObjectNotFound
+	}
+	e, err := f.wrapEntries(entries...)
+	if err != nil {
+		return nil, err
+	}
+	for _, err := range errs {
+		if err != nil {
+			return e.(*Object), err
+		}
+	}
+	return e.(*Object), nil
 }
 
-// Precision is the greatest Precision of all remotes
+// Precision is the greatest Precision of all upstreams
 func (f *Fs) Precision() time.Duration {
 	var greatestPrecision time.Duration
-	for _, remote := range f.remotes {
-		if remote.Precision() > greatestPrecision {
-			greatestPrecision = remote.Precision()
+	for _, u := range f.upstreams {
+		if u.Precision() > greatestPrecision {
+			greatestPrecision = u.Precision()
 		}
 	}
 	return greatestPrecision
 }
 
+func (f *Fs) action(ctx context.Context, path string) ([]*upstream.Fs, error) {
+	return f.actionPolicy.Action(ctx, f.upstreams, path)
+}
+
+func (f *Fs) actionEntries(entries ...upstream.Entry) ([]upstream.Entry, error) {
+	return f.actionPolicy.ActionEntries(entries...)
+}
+
+func (f *Fs) create(ctx context.Context, path string) ([]*upstream.Fs, error) {
+	return f.createPolicy.Create(ctx, f.upstreams, path)
+}
+
+func (f *Fs) createEntries(entries ...upstream.Entry) ([]upstream.Entry, error) {
+	return f.createPolicy.CreateEntries(entries...)
+}
+
+func (f *Fs) search(ctx context.Context, path string) (*upstream.Fs, error) {
+	return f.searchPolicy.Search(ctx, f.upstreams, path)
+}
+
+func (f *Fs) searchEntries(entries ...upstream.Entry) (upstream.Entry, error) {
+	return f.searchPolicy.SearchEntries(entries...)
+}
+
+func (f *Fs) mergeDirEntries(entriess [][]upstream.Entry) (fs.DirEntries, error) {
+	entryMap := make(map[string]([]upstream.Entry))
+	for _, en := range entriess {
+		if en == nil {
+			continue
+		}
+		for _, entry := range en {
+			remote := entry.Remote()
+			if f.Features().CaseInsensitive {
+				remote = strings.ToLower(remote)
+			}
+			entryMap[remote] = append(entryMap[remote], entry)
+		}
+	}
+	var entries fs.DirEntries
+	for path := range entryMap {
+		e, err := f.wrapEntries(entryMap[path]...)
+		if err != nil {
+			return nil, err
+		}
+		entries = append(entries, e)
+	}
+	return entries, nil
+}
+
 // NewFs constructs an Fs from the path.
 //
 // The returned Fs is the actual Fs, referenced by remote in the config
@@ -320,51 +804,46 @@ func NewFs(name, root string, m configmap.Mapper) (fs.Fs, error) {
 	if err != nil {
 		return nil, err
 	}
-	if len(opt.Remotes) == 0 {
-		return nil, errors.New("union can't point to an empty remote - check the value of the remotes setting")
+	if len(opt.Upstreams) == 0 {
+		return nil, errors.New("union can't point to an empty upstream - check the value of the upstreams setting")
 	}
-	if len(opt.Remotes) == 1 {
-		return nil, errors.New("union can't point to a single remote - check the value of the remotes setting")
+	if len(opt.Upstreams) == 1 {
+		return nil, errors.New("union can't point to a single upstream - check the value of the upstreams setting")
 	}
-	for _, remote := range opt.Remotes {
-		if strings.HasPrefix(remote, name+":") {
-			return nil, errors.New("can't point union remote at itself - check the value of the remote setting")
+	for _, u := range opt.Upstreams {
+		if strings.HasPrefix(u, name+":") {
+			return nil, errors.New("can't point union remote at itself - check the value of the upstreams setting")
 		}
 	}
 
-	var remotes []fs.Fs
-	for i := range opt.Remotes {
+	var upstreams []*upstream.Fs
+	for i := range opt.Upstreams {
 		// Last remote first so we return the correct (last) matching fs in case of fs.ErrorIsFile
-		var remote = opt.Remotes[len(opt.Remotes)-i-1]
-		_, configName, fsPath, err := fs.ParseRemote(remote)
+		var u = opt.Upstreams[len(opt.Upstreams)-i-1]
+		uFs, err := upstream.New(u, root, time.Duration(opt.CacheTime) * time.Second)
 		if err != nil {
 			return nil, err
 		}
-		var rootString = path.Join(fsPath, filepath.ToSlash(root))
-		if configName != "local" {
-			rootString = configName + ":" + rootString
-		}
-		myFs, err := cache.Get(rootString)
-		if err != nil {
-			if err == fs.ErrorIsFile {
-				return myFs, err
-			}
-			return nil, err
-		}
-		remotes = append(remotes, myFs)
-	}
-
-	// Reverse the remotes again so they are in the order as before
-	for i, j := 0, len(remotes)-1; i < j; i, j = i+1, j-1 {
-		remotes[i], remotes[j] = remotes[j], remotes[i]
+		upstreams = append(upstreams, uFs)
 	}
 
 	f := &Fs{
-		name:    name,
-		root:    root,
-		opt:     *opt,
-		remotes: remotes,
-		wr:      remotes[len(remotes)-1],
+		name:         name,
+		root:         root,
+		opt:          *opt,
+		upstreams:    upstreams,
+	}
+	f.actionPolicy, err = policy.Get(opt.ActionPolicy)
+	if err != nil {
+		return nil, err
+	}
+	f.createPolicy, err = policy.Get(opt.CreatePolicy)
+	if err != nil {
+		return nil, err
+	}
+	f.searchPolicy, err = policy.Get(opt.SearchPolicy)
+	if err != nil {
+		return nil, err
 	}
 	var features = (&fs.Features{
 		CaseInsensitive:         true,
@@ -376,9 +855,14 @@ func NewFs(name, root string, m configmap.Mapper) (fs.Fs, error) {
 		SetTier:                 true,
 		GetTier:                 true,
 	}).Fill(f)
-	features = features.Mask(f.wr) // mask the features just on the writable fs
+	for _, f := range upstreams {
+		if !f.IsWritable() {
+			continue
+		}
+		features = features.Mask(f) // Mask all writable upstream fs
+	}
 
-	// Really need the union of all remotes for these, so
+	// Really need the union of all upstreams for these, so
 	// re-instate and calculate separately.
 	features.ChangeNotify = f.ChangeNotify
 	features.DirCacheFlush = f.DirCacheFlush
@@ -388,12 +872,12 @@ func NewFs(name, root string, m configmap.Mapper) (fs.Fs, error) {
 	// Clear ChangeNotify and DirCacheFlush if all are nil
 	clearChangeNotify := true
 	clearDirCacheFlush := true
-	for _, remote := range f.remotes {
-		remoteFeatures := remote.Features()
-		if remoteFeatures.ChangeNotify != nil {
+	for _, u := range f.upstreams {
+		uFeatures := u.Features()
+		if uFeatures.ChangeNotify != nil {
 			clearChangeNotify = false
 		}
-		if remoteFeatures.DirCacheFlush != nil {
+		if uFeatures.DirCacheFlush != nil {
 			clearDirCacheFlush = false
 		}
 	}
@@ -407,15 +891,36 @@ func NewFs(name, root string, m configmap.Mapper) (fs.Fs, error) {
 	f.features = features
 
 	// Get common intersection of hashes
-	hashSet := f.remotes[0].Hashes()
-	for _, remote := range f.remotes[1:] {
-		hashSet = hashSet.Overlap(remote.Hashes())
+	hashSet := f.upstreams[0].Hashes()
+	for _, u := range f.upstreams[1:] {
+		hashSet = hashSet.Overlap(u.Hashes())
 	}
 	f.hashSet = hashSet
 
 	return f, nil
 }
 
+func parentDir(absPath string) string {
+	parent := path.Dir(strings.TrimRight(filepath.ToSlash(absPath), "/"))
+	if parent == "." {
+		parent = ""
+	}
+	return parent
+}
+
+func multithread(num int, fn func(int)) {
+	var wg sync.WaitGroup
+	for i := 0; i < num; i++ {
+		wg.Add(1)
+		i := i
+		go func() {
+			defer wg.Done()
+			fn(i)
+		}()
+	}
+	wg.Wait()
+}
+
 // Check the interfaces are satisfied
 var (
 	_ fs.Fs              = (*Fs)(nil)
diff --git a/backend/union/upstream/upstream.go b/backend/union/upstream/upstream.go
new file mode 100644
index 000000000..fa2cc2ab7
--- /dev/null
+++ b/backend/union/upstream/upstream.go
@@ -0,0 +1,311 @@
+package upstream
+
+import (
+	"context"
+	"io"
+	"path"
+	"path/filepath"
+	"strings"
+	"sync"
+	"sync/atomic"
+	"time"
+
+	"github.com/pkg/errors"
+	"github.com/rclone/rclone/fs"
+	"github.com/rclone/rclone/fs/cache"
+)
+
+var (
+	// ErrUsageFieldNotSupported stats the usage field is not supported by the backend
+	ErrUsageFieldNotSupported = errors.New("this usage field is not supported")
+)
+
+const (
+    unInitilized uint32 = iota
+    initilizing
+    normal
+    updating
+)
+
+// Fs is a wrap of any fs and its configs
+type Fs struct {
+	fs.Fs
+	writable    bool
+	creatable   bool
+	usage       *fs.Usage     // Cache the usage
+	cacheMutex  sync.RWMutex
+	cacheExpiry int64     // usage cache expiry time
+	cacheTime   time.Duration // cache duration
+	cacheState  uint32        // if the cache is updating
+}
+
+// Directory describes a wrapped Directory
+//
+// This is a wrapped Directory which contains the upstream Fs
+type Directory struct {
+	fs.Directory
+	f *Fs
+}
+
+// Object describes a wrapped Object
+//
+// This is a wrapped Object which contains the upstream Fs
+type Object struct {
+	fs.Object
+	f *Fs
+}
+
+// Entry describe a warpped fs.DirEntry interface with the
+// information of upstream Fs
+type Entry interface {
+	fs.DirEntry
+	UpstreamFs() *Fs
+}
+
+// New creates a new Fs based on the
+// string formatted `type:root_path(:ro/:nc)`
+func New(remote, root string, cacheTime time.Duration) (*Fs, error) {
+	_, configName, fsPath, err := fs.ParseRemote(remote)
+	if err != nil {
+		return nil, err
+	}
+	rFs := &Fs{
+		writable:    true,
+		creatable:   true,
+		cacheExpiry: time.Now().Unix(),
+		cacheTime:   cacheTime,
+		usage:       &fs.Usage{},
+	}
+	if strings.HasSuffix(fsPath, ":ro") {
+		rFs.writable = false
+		rFs.creatable = false
+		fsPath = fsPath[0 : len(fsPath)-3]
+	} else if strings.HasSuffix(fsPath, ":nc") {
+		rFs.writable = true
+		rFs.creatable = false
+		fsPath = fsPath[0 : len(fsPath)-3]
+	}
+	var rootString = path.Join(fsPath, filepath.ToSlash(root))
+	if configName != "local" {
+		rootString = configName + ":" + rootString
+	}
+	myFs, err := cache.Get(rootString)
+	if err != nil {
+		return nil, err
+	}
+	rFs.Fs = myFs
+	return rFs, nil
+}
+
+// WrapDirectory wraps a fs.Directory to include the info
+// of the upstream Fs
+func (f *Fs) WrapDirectory(e fs.Directory) *Directory {
+	if e == nil {
+		return nil
+	}
+	return &Directory{
+		Directory: e,
+		f:         f,
+	}
+}
+
+// WrapObject wraps a fs.Object to include the info
+// of the upstream Fs
+func (f *Fs) WrapObject(o fs.Object) *Object {
+	if o == nil {
+		return nil
+	}
+	return &Object{
+		Object: o,
+		f:      f,
+	}
+}
+
+// WrapEntry wraps a fs.DirEntry to include the info
+// of the upstream Fs
+func (f *Fs) WrapEntry(e fs.DirEntry) (Entry, error) {
+	switch e.(type) {
+	case fs.Object:
+		return f.WrapObject(e.(fs.Object)), nil
+	case fs.Directory:
+		return f.WrapDirectory(e.(fs.Directory)), nil
+	default:
+		return nil, errors.Errorf("unknown object type %T", e)
+	}
+}
+
+// UpstreamFs get the upstream Fs the entry is stored in
+func (e *Directory) UpstreamFs() *Fs {
+	return e.f
+}
+
+// UpstreamFs get the upstream Fs the entry is stored in
+func (o *Object) UpstreamFs() *Fs {
+	return o.f
+}
+
+// IsCreatable return if the fs is allowed to create new objects
+func (f *Fs) IsCreatable() bool {
+	return f.creatable
+}
+
+// IsWritable return if the fs is allowed to write
+func (f *Fs) IsWritable() bool {
+	return f.writable
+}
+
+// Put in to the remote path with the modTime given of the given size
+//
+// May create the object even if it returns an error - if so
+// will return the object and the error, otherwise will return
+// nil and the error
+func (f *Fs) Put(ctx context.Context, in io.Reader, src fs.ObjectInfo, options ...fs.OpenOption) (fs.Object, error) {
+	o, err := f.Fs.Put(ctx, in, src, options...)
+	if err != nil {
+		return o, err
+	}
+	f.cacheMutex.Lock()
+	defer f.cacheMutex.Unlock()
+	size := src.Size()
+	if f.usage.Used != nil {
+		*f.usage.Used += size
+	}
+	if f.usage.Free != nil {
+		*f.usage.Free -= size
+	}
+	return o, nil
+}
+
+// PutStream uploads to the remote path with the modTime given of indeterminate size
+//
+// May create the object even if it returns an error - if so
+// will return the object and the error, otherwise will return
+// nil and the error
+func (f *Fs) PutStream(ctx context.Context, in io.Reader, src fs.ObjectInfo, options ...fs.OpenOption) (fs.Object, error) {
+	do := f.Features().PutStream
+	if do == nil {
+		return nil, fs.ErrorNotImplemented
+	}
+	o, err := do(ctx, in, src, options...)
+	if err != nil {
+		return o, err
+	}
+	f.cacheMutex.Lock()
+	defer f.cacheMutex.Unlock()
+	size := o.Size()
+	if f.usage.Used != nil {
+		*f.usage.Used += size
+	}
+	if f.usage.Free != nil {
+		*f.usage.Free -= size
+	}
+	return o, nil
+}
+
+// Update in to the object with the modTime given of the given size
+//
+// When called from outside a Fs by rclone, src.Size() will always be >= 0.
+// But for unknown-sized objects (indicated by src.Size() == -1), Upload should either
+// return an error or update the object properly (rather than e.g. calling panic).
+func (o *Object) Update(ctx context.Context, in io.Reader, src fs.ObjectInfo, options ...fs.OpenOption) error {
+	size := o.Size()
+	err := o.Object.Update(ctx, in, src, options...)
+	if err != nil {
+		return err
+	}
+	o.f.cacheMutex.Lock()
+	defer o.f.cacheMutex.Unlock()
+	delta := o.Size() - size
+	if delta <= 0 {
+		return nil
+	}
+	if o.f.usage.Used != nil {
+		*o.f.usage.Used += size
+	}
+	if o.f.usage.Free != nil {
+		*o.f.usage.Free -= size
+	}
+	return nil
+}
+
+// About gets quota information from the Fs
+func (f *Fs) About(ctx context.Context) (*fs.Usage, error) {
+	if atomic.LoadInt64(&f.cacheExpiry) <= time.Now().Unix() {
+		err := f.updateUsage()
+		if err != nil {
+			return nil, ErrUsageFieldNotSupported
+		}
+	}
+	f.cacheMutex.RLock()
+	defer f.cacheMutex.RUnlock()
+	return f.usage, nil
+}
+
+// GetFreeSpace get the free space of the fs
+func (f *Fs) GetFreeSpace() (int64, error) {
+	if atomic.LoadInt64(&f.cacheExpiry) <= time.Now().Unix() {
+		err := f.updateUsage()
+		if err != nil {
+			return 0, ErrUsageFieldNotSupported
+		}
+	}
+	f.cacheMutex.RLock()
+	defer f.cacheMutex.RUnlock()
+	if f.usage.Free == nil {
+		return 0, ErrUsageFieldNotSupported
+	}
+	return *f.usage.Free, nil
+}
+
+// GetUsedSpace get the used space of the fs
+func (f *Fs) GetUsedSpace() (int64, error) {
+	if atomic.LoadInt64(&f.cacheExpiry) <= time.Now().Unix() {
+		err := f.updateUsage()
+		if err != nil {
+			return 0, ErrUsageFieldNotSupported
+		}
+	}
+	f.cacheMutex.RLock()
+	defer f.cacheMutex.RUnlock()
+	if f.usage.Used == nil {
+		return 0, ErrUsageFieldNotSupported
+	}
+	return *f.usage.Used, nil
+}
+
+func (f *Fs) updateUsage() error {
+	if do := f.Fs.Features().About; do == nil {
+		return ErrUsageFieldNotSupported
+	}
+	if atomic.LoadUint32(&f.cacheState) == unInitilized {
+		f.cacheMutex.Lock()
+		defer f.cacheMutex.Unlock()
+		if !atomic.CompareAndSwapUint32(&f.cacheState, unInitilized, initilizing) {
+			return f.updateUsage()
+		}
+		return f.updateUsageCore(false)
+	}
+	if atomic.CompareAndSwapUint32(&f.cacheState, normal, updating) {
+		go f.updateUsageCore(true)
+	}
+	return nil
+}
+
+func (f *Fs) updateUsageCore(lock bool) error {
+	defer func() {
+		atomic.StoreInt64(&f.cacheExpiry, time.Now().Add(f.cacheTime).Unix())
+		atomic.StoreUint32(&f.cacheState, normal)
+	}()
+	if (lock) {
+		f.cacheMutex.Lock()
+		defer f.cacheMutex.Unlock()
+	}
+	// Run in background, should not be cancelled by user
+	ctx, _ := context.WithTimeout(context.Background(), 15*time.Second)
+	usage, err := f.Features().About(ctx)
+	if err != nil {
+		return err
+	}
+	f.usage = usage
+	return nil
+}
\ No newline at end of file