diff --git a/api/api.go b/api/api.go
index 1ac6f35cbd..5114a28938 100644
--- a/api/api.go
+++ b/api/api.go
@@ -36,9 +36,10 @@ type VersionResponse struct {
 }
 
 type PoolPostRequest struct {
-	Name     string   `json:"name"`
-	SortKeys SortKeys `json:"layout"`
-	Thresh   int64    `json:"thresh"`
+	Name      string   `json:"name"`
+	SortKeys  SortKeys `json:"layout"`
+	ObjectCap uint64   `json:"objectcap"`
+	FrameCap  uint64   `json:"framecap"`
 }
 
 type SortKeys struct {
diff --git a/bsup/writer.go b/bsup/writer.go
index 6d321318f7..b3e45eadd6 100644
--- a/bsup/writer.go
+++ b/bsup/writer.go
@@ -14,45 +14,52 @@ import (
 	"github.com/superdb/super/vector/vio"
 )
 
-var maxFrameSize uint32 = 120_000
-
-// XXX a future PR will wire in compress / thresh options to command line.
-// XXX Rows is the key flag we need for the rows writer.
 type WriterOpts struct {
 	Compress bool
-	// FrameThresh is the minimum frame size in uncompressed bytes.
-	FrameThresh int
-	Rows        bool
+	FrameCap uint64
+	Rows     bool
 }
 
+const DefaultFrameCap = 20_000
+
 // ColumnWriter implements the vio.Pusher interface. A Pusher creates a vector
 // BSUP object from a stream of vector.Any.
 type ColumnWriter struct {
-	writer  io.WriteCloser
-	dynamic *vbuild.DynamicBuilder
-	sctx    *super.Context
-	fuser   fuser
-	size    uint64
-	ctrl    *RowWriter
+	writer   io.WriteCloser
+	dynamic  *vbuild.DynamicBuilder
+	sctx     *super.Context
+	fuser    fuser
+	size     uint64
+	ctrl     *RowWriter
+	framecap uint64
 }
 
 var _ vio.Pusher = (*ColumnWriter)(nil)
 
 func NewColumnWriter(w io.WriteCloser) *ColumnWriter {
+	return NewColumnWriterWithCap(w, DefaultFrameCap)
+}
+
+func NewColumnWriterWithCap(w io.WriteCloser, frameCap uint64) *ColumnWriter {
 	sctx := super.NewContext()
 	return &ColumnWriter{
-		writer:  w,
-		dynamic: vbuild.NewDynamicBuilder(),
-		sctx:    sctx,
-		fuser:   newFuser(sctx),
+		writer:   w,
+		dynamic:  vbuild.NewDynamicBuilder(),
+		sctx:     sctx,
+		fuser:    newFuser(sctx),
+		framecap: frameCap,
 	}
 }
 
 func NewWriterWithOpts(w io.WriteCloser, opt WriterOpts) vio.PushCloser {
 	if opt.Rows {
-		return NewRowWriter(w)
+		writer := NewRowWriter(w)
+		writer.framecap = uint32(opt.FrameCap)
+		return writer
 	}
-	return NewColumnWriter(w)
+	writer := NewColumnWriter(w)
+	writer.framecap = opt.FrameCap
+	return writer
 }
 
 func (c *ColumnWriter) Close() error {
@@ -87,7 +94,7 @@ func (c *ColumnWriter) WriteSuperFrame(vec vector.Any) (uint64, error) {
 func (c *ColumnWriter) Push(vec vector.Any) error {
 	if vec.Len() != 0 {
 		c.dynamic.Write(vec)
-		if c.dynamic.Len() >= maxFrameSize {
+		if c.dynamic.Len() >= uint32(c.framecap) {
 			return c.pushFrame()
 		}
 	}
@@ -202,6 +209,7 @@ type RowWriter struct {
 	size       uint64
 	bytes      []byte
 	len        uint32
+	framecap   uint32
 }
 
 var _ vio.Pusher = (*RowWriter)(nil)
@@ -215,6 +223,7 @@ func NewRowWriter(w io.WriteCloser) *RowWriter {
 		sctx:       sctx,
 		fuser:      newFuser(sctx),
 		superfuser: newFuser(sctx),
+		framecap:   DefaultFrameCap,
 	}
 }
 
@@ -237,7 +246,7 @@ func (r *RowWriter) Push(vec vector.Any) error {
 		}
 	}
 	r.len += vec.Len()
-	if r.len >= maxFrameSize {
+	if r.len >= r.framecap {
 		return r.pushFrame(false)
 	}
 	return nil
@@ -253,7 +262,7 @@ func (r *RowWriter) write(val super.Value, ctrl bool) error {
 	r.fuser.fuse(typ)
 	r.superfuser.fuse(typ)
 	r.len++
-	if r.len >= maxFrameSize || ctrl {
+	if r.len >= r.framecap || ctrl {
 		return r.pushFrame(ctrl)
 	}
 	return nil
diff --git a/bsup/ztests/framecap.yaml b/bsup/ztests/framecap.yaml
new file mode 100644
index 0000000000..30add89cc4
--- /dev/null
+++ b/bsup/ztests/framecap.yaml
@@ -0,0 +1,9 @@
+script: |
+  seq 100 | super -inputcap 5 -framecap 5 -o out.bsup -
+  super dev bsup out.bsup | super -s -c "count() by nameof(typeof(this)) | ? Header or Footer | sort count" -
+
+outputs:
+  - name: stdout
+    data: |
+      {nameof:"SuperFooter",count:1}
+      {nameof:"ColumnHeader",count:20}
diff --git a/cli/inputflags/flags.go b/cli/inputflags/flags.go
index f37100eb7f..203f83c8f6 100644
--- a/cli/inputflags/flags.go
+++ b/cli/inputflags/flags.go
@@ -26,6 +26,7 @@ func (f *Flags) SetFlags(fs *flag.FlagSet) {
 	})
 	fs.BoolVar(&f.Dynamic, "dynamic", false, "disable static type checking of inputs")
 	fs.StringVar(&opts.Format, "i", "auto", "format of input data [auto,arrows,bsup,csv,json,line,parquet,sup,tsv,zeek]")
+	fs.IntVar(&opts.InputCap, "inputcap", 0, "limit size of batched units of input")
 	fs.BoolVar(&f.Static, "static", false, "force static type checking of inputs")
 }
 
diff --git a/cli/outputflags/flags.go b/cli/outputflags/flags.go
index 34eece7312..1dbc46b576 100644
--- a/cli/outputflags/flags.go
+++ b/cli/outputflags/flags.go
@@ -41,6 +41,7 @@ func (f *Flags) Options() anyio.WriterOpts {
 
 func (f *Flags) setFlags(fs *flag.FlagSet) {
 	fs.BoolVar(&f.color, "color", true, "enable/disable color formatting for -S and db text output")
+	fs.Uint64Var(&f.BSUP.FrameCap, "framecap", 10000, "number of values per BSUP frame")
 	fs.BoolVar(&f.CSV.NoHeader, "noheader", false, "omit header for CSV and TSV output")
 	fs.IntVar(&f.pretty, "pretty", 2,
 		"tab size to pretty print JSON and Super JSON output (0 for newline-delimited output")
diff --git a/cmd/super/db/create/command.go b/cmd/super/db/create/command.go
index 7ddedc958a..96b912e2b1 100644
--- a/cmd/super/db/create/command.go
+++ b/cmd/super/db/create/command.go
@@ -5,12 +5,12 @@ import (
 	"flag"
 	"fmt"
 
+	"github.com/superdb/super/bsup"
 	"github.com/superdb/super/cli/poolflags"
 	"github.com/superdb/super/cmd/super/db"
 	"github.com/superdb/super/db/data"
 	"github.com/superdb/super/order"
 	"github.com/superdb/super/pkg/charm"
-	"github.com/superdb/super/pkg/units"
 )
 
 var spec = &charm.Spec{
@@ -25,9 +25,10 @@ See https://superdb.org/command/db.html#super-db-create
 
 type Command struct {
 	*db.Command
-	sortKey string
-	thresh  units.Bytes
-	use     bool
+	sortKey   string
+	frameCap  uint64
+	objectCap uint64
+	use       bool
 }
 
 func init() {
@@ -36,8 +37,8 @@ func init() {
 
 func New(parent charm.Command, f *flag.FlagSet) (charm.Command, error) {
 	c := &Command{Command: parent.(*db.Command)}
-	c.thresh = data.DefaultThreshold
-	f.Var(&c.thresh, "S", "target size of pool data objects, as '10MB' or '4GiB', etc.")
+	f.Uint64Var(&c.frameCap, "framecap", bsup.DefaultFrameCap, "target number of values BSUP frames")
+	f.Uint64Var(&c.objectCap, "objectcap", data.DefaultObjectCap, "target number of values in pool data objects")
 	f.BoolVar(&c.use, "use", false, "set created pool as the current pool")
 	f.StringVar(&c.sortKey, "orderby", "ts:desc", "pool key with optional :asc or :desc suffix to organize data in pool (cannot be changed)")
 	return c, nil
@@ -61,7 +62,7 @@ func (c *Command) Run(args []string) error {
 		return err
 	}
 	poolName := args[0]
-	id, err := db.CreatePool(ctx, poolName, sortKey, int64(c.thresh))
+	id, err := db.CreatePool(ctx, poolName, sortKey, c.objectCap, c.frameCap)
 	if err != nil {
 		return err
 	}
diff --git a/cmd/super/db/internal/dbmanage/scan.go b/cmd/super/db/internal/dbmanage/scan.go
index 35b7221bd3..6c3e60ec57 100644
--- a/cmd/super/db/internal/dbmanage/scan.go
+++ b/cmd/super/db/internal/dbmanage/scan.go
@@ -48,7 +48,7 @@ func scan(ctx context.Context, it *objectIterator, pool *pools.Config, runCh cha
 		if err != nil {
 			return err
 		}
-		if run.overlaps(o.Min, o.Max) || run.size+o.Size < pool.Threshold {
+		if run.overlaps(o.Min, o.Max) || run.len+o.Count < pool.ObjectCap {
 			run.add(o)
 			continue
 		}
@@ -119,7 +119,7 @@ type runBuilder struct {
 	span    extent.Span
 	cmp     expr.CompareFn
 	objects []*object
-	size    int64
+	len     uint64
 }
 
 func newRunBuilder() *runBuilder {
@@ -135,7 +135,7 @@ func (r *runBuilder) overlaps(first, last super.Value) bool {
 
 func (r *runBuilder) add(o *object) {
 	r.objects = append(r.objects, o)
-	r.size += o.Size
+	r.len += o.Count
 	if r.span == nil {
 		r.span = extent.NewGeneric(o.Min, o.Max, r.cmp)
 		return
@@ -155,5 +155,5 @@ func (r *runBuilder) objectIDs() []uuid.UUID {
 func (r *runBuilder) reset() {
 	r.span = nil
 	r.objects = r.objects[:0]
-	r.size = 0
+	r.len = 0
 }
diff --git a/cmd/super/ztests/terminal-output-format.yaml b/cmd/super/ztests/terminal-output-format.yaml
index 36cda712d6..9ef1d077d6 100644
--- a/cmd/super/ztests/terminal-output-format.yaml
+++ b/cmd/super/ztests/terminal-output-format.yaml
@@ -15,4 +15,4 @@ outputs:
   - name: stdout
     data: "{a:1}\r\n\
       {b:2}\r\n\
-      p1 xxx key ts order desc\r\n"
+      p1 xxx key ts order desc objectcap 1048576 framecap 20000\r\n"
diff --git a/db/api/api.go b/db/api/api.go
index ed390641d2..3a952ff287 100644
--- a/db/api/api.go
+++ b/db/api/api.go
@@ -29,7 +29,7 @@ type Interface interface {
 	Query(ctx context.Context, query []srcfiles.Input) (vio.Scanner, error)
 	PoolID(ctx context.Context, poolName string) (uuid.UUID, error)
 	CommitObject(ctx context.Context, poolID uuid.UUID, branchName string) (uuid.UUID, error)
-	CreatePool(context.Context, string, order.SortKeys, int64) (uuid.UUID, error)
+	CreatePool(context.Context, string, order.SortKeys, uint64, uint64) (uuid.UUID, error)
 	RemovePool(context.Context, uuid.UUID) error
 	RenamePool(context.Context, uuid.UUID, string) error
 	CreateBranch(ctx context.Context, pool uuid.UUID, name string, parent uuid.UUID) error
diff --git a/db/api/local.go b/db/api/local.go
index 1c2ce6122b..ba956ef969 100644
--- a/db/api/local.go
+++ b/db/api/local.go
@@ -61,11 +61,11 @@ func (l *local) Root() *db.Root {
 	return l.db
 }
 
-func (l *local) CreatePool(ctx context.Context, name string, sortKeys order.SortKeys, thresh int64) (uuid.UUID, error) {
+func (l *local) CreatePool(ctx context.Context, name string, sortKeys order.SortKeys, objectCap, frameCap uint64) (uuid.UUID, error) {
 	if name == "" {
 		return uuid.Nil(), errors.New("no pool name provided")
 	}
-	pool, err := l.db.CreatePool(ctx, name, sortKeys, thresh)
+	pool, err := l.db.CreatePool(ctx, name, sortKeys, objectCap, frameCap)
 	if err != nil {
 		return uuid.Nil(), err
 	}
diff --git a/db/api/remote.go b/db/api/remote.go
index 9c7e79d4b8..e00559a110 100644
--- a/db/api/remote.go
+++ b/db/api/remote.go
@@ -53,14 +53,15 @@ func (r *remote) CommitObject(ctx context.Context, poolID uuid.UUID, branchName
 	return res.Commit, err
 }
 
-func (r *remote) CreatePool(ctx context.Context, name string, sortKeys order.SortKeys, thresh int64) (uuid.UUID, error) {
+func (r *remote) CreatePool(ctx context.Context, name string, sortKeys order.SortKeys, objectCap, frameCap uint64) (uuid.UUID, error) {
 	res, err := r.conn.CreatePool(ctx, api.PoolPostRequest{
 		Name: name,
 		SortKeys: api.SortKeys{
 			Order: sortKeys.Primary().Order,
 			Keys:  field.List{sortKeys.Primary().Path},
 		},
-		Thresh: thresh,
+		ObjectCap: objectCap,
+		FrameCap:  frameCap,
 	})
 	if err != nil {
 		return uuid.Nil(), err
diff --git a/db/data/object.go b/db/data/object.go
index b511d8095f..75ba56f289 100644
--- a/db/data/object.go
+++ b/db/data/object.go
@@ -15,7 +15,7 @@ import (
 )
 
 const (
-	DefaultThreshold = 500 * 1024 * 1024
+	DefaultObjectCap = 1024 * 1024
 )
 
 // A FileKind is the first part of a file name, used to differentiate files
diff --git a/db/pools/config.go b/db/pools/config.go
index 4a7180c803..fd492aa30d 100644
--- a/db/pools/config.go
+++ b/db/pools/config.go
@@ -3,6 +3,7 @@ package pools
 import (
 	"uuid"
 
+	"github.com/superdb/super/bsup"
 	"github.com/superdb/super/db/data"
 	"github.com/superdb/super/db/journal"
 	"github.com/superdb/super/order"
@@ -16,24 +17,29 @@ type Config struct {
 	Name      string         `super:"name"`
 	ID        uuid.UUID      `super:"id"`
 	SortKeys  order.SortKeys `super:"layout"`
-	Threshold int64          `super:"threshold"`
+	ObjectCap uint64         `super:"objectcap"`
+	FrameCap  uint64         `super:"framecap"`
 }
 
 var _ journal.Entry = (*Config)(nil)
 
-func NewConfig(name string, sortKeys order.SortKeys, thresh int64) *Config {
+func NewConfig(name string, sortKeys order.SortKeys, objectCap, frameCap uint64) *Config {
 	if sortKeys.IsNil() {
 		sortKeys = order.SortKeys{order.NewSortKey(order.Desc, field.Dotted("ts"))}
 	}
-	if thresh == 0 {
-		thresh = data.DefaultThreshold
+	if objectCap == 0 {
+		objectCap = data.DefaultObjectCap
+	}
+	if frameCap == 0 {
+		frameCap = bsup.DefaultFrameCap
 	}
 	return &Config{
 		Ts:        nano.Now(),
 		Name:      name,
 		ID:        uuid.NewV7(),
 		SortKeys:  sortKeys,
-		Threshold: thresh,
+		ObjectCap: objectCap,
+		FrameCap:  frameCap,
 	}
 }
 
diff --git a/db/root.go b/db/root.go
index fa3f0ac0c3..f357869e08 100644
--- a/db/root.go
+++ b/db/root.go
@@ -10,6 +10,7 @@ import (
 
 	arc "github.com/hashicorp/golang-lru/arc/v2"
 	"github.com/superdb/super"
+	"github.com/superdb/super/bsup"
 	"github.com/superdb/super/bsupbytes"
 	"github.com/superdb/super/compiler/dag"
 	"github.com/superdb/super/db/branches"
@@ -315,20 +316,23 @@ func (r *Root) RenamePool(ctx context.Context, id uuid.UUID, newName string) err
 	return r.pools.Rename(ctx, id, newName)
 }
 
-func (r *Root) CreatePool(ctx context.Context, name string, sortKeys order.SortKeys, thresh int64) (*Pool, error) {
+func (r *Root) CreatePool(ctx context.Context, name string, sortKeys order.SortKeys, objectCap, frameCap uint64) (*Pool, error) {
 	if name == "HEAD" {
 		return nil, fmt.Errorf("pool cannot be named %q", name)
 	}
 	if r.pools.LookupByName(ctx, name) != nil {
 		return nil, fmt.Errorf("%s: %w", name, pools.ErrExists)
 	}
-	if thresh == 0 {
-		thresh = data.DefaultThreshold
-	}
 	if len(sortKeys) > 1 {
 		return nil, errors.New("multiple pool keys not supported")
 	}
-	config := pools.NewConfig(name, sortKeys, thresh)
+	if objectCap == 0 {
+		objectCap = data.DefaultObjectCap
+	}
+	if frameCap == 0 {
+		frameCap = bsup.DefaultFrameCap
+	}
+	config := pools.NewConfig(name, sortKeys, objectCap, frameCap)
 	if err := CreatePool(ctx, r.engine, r.logger, r.path, config); err != nil {
 		return nil, err
 	}
diff --git a/db/writer.go b/db/writer.go
index 8500443dd8..e716832cd5 100644
--- a/db/writer.go
+++ b/db/writer.go
@@ -50,14 +50,17 @@ func (w *Writer) newObject() *data.Object {
 }
 
 func (w *Writer) Push(vec vector.Any) error {
-	// XXX We need a threshold here (or in BSUP column writer)
-	// so we can write lots of smaller row-group-sized column frames
-	// per super frame.  This will come in a subsequent PR.
 	w.builder.Write(vec)
+	if uint64(w.builder.Len()) >= w.pool.ObjectCap {
+		return w.flush()
+	}
 	return nil
 }
 
 func (w *Writer) flush() error {
+	// Each object is one big vector limited in length by ObjectCap.
+	// Inside of the superframe are Frames which are in turn limited
+	// in length by FrameCap.
 	vec := w.builder.Build()
 	if vec.Len() == 0 {
 		return nil
@@ -78,7 +81,7 @@ func (w *Writer) flush() error {
 		object.Min = minVal
 		object.Max = maxVal
 	}
-	writer := bsup.NewColumnWriter(out)
+	writer := bsup.NewColumnWriterWithCap(out, w.pool.FrameCap)
 	size, err := writer.WriteSuperFrame(vec)
 	if err != nil {
 		out.Close()
diff --git a/db/ztests/aggregate-ts.yaml b/db/ztests/aggregate-ts.yaml
index 527cf27e33..4d9f03dea8 100644
--- a/db/ztests/aggregate-ts.yaml
+++ b/db/ztests/aggregate-ts.yaml
@@ -3,9 +3,9 @@ script: |
   super db init -q
   # aggregate-ts.sup contains records over two days, mapping to two partitions, and
   # the small target size here causes each partition to have several segments.
-  super db create -q -S 256B logs
+  super db create -q -objectcap=50 -framecap=10 logs
   # causes parition to have several chunk files.
-  super db load -q -use logs aggregate-ts.sup
+  super db load -q -use logs -inputcap=10 aggregate-ts.sup
   echo ===
   super db -s -c 'from logs | head 1'
   echo ===
diff --git a/db/ztests/delete-where.yaml b/db/ztests/delete-where.yaml
index c770e1c74f..4db99f8244 100644
--- a/db/ztests/delete-where.yaml
+++ b/db/ztests/delete-where.yaml
@@ -3,8 +3,8 @@ script: |
   super db init -q
   for order in desc asc; do
     echo === $order ===
-    super db create -use -q -S 1KB -orderby ts:$order test
-    seq 1000 | super -c '{ts:this-1,s:f"val{this-1}"}' - | super db load -q -
+    super db create -use -q -framecap 50 -objectcap 500 -orderby ts:$order test
+    seq 1000 | super -c '{ts:this-1,s:f"val{this-1}"}' - | super db load -q -inputcap=10 -
     super db delete -q -where 'ts > 400 and ts <= 500'
     super db -s -c 'from test | count:=count()'
     super db delete -q -where 's == "val1" or s == "val999"'
diff --git a/db/ztests/deterministic-merge.yaml b/db/ztests/deterministic-merge.yaml
index 29a70954e9..9b3c3453f0 100644
--- a/db/ztests/deterministic-merge.yaml
+++ b/db/ztests/deterministic-merge.yaml
@@ -1,8 +1,8 @@
 script: |
   export SUPER_DB=test
   super db init -q
-  super db create -q -S 32B -orderby ts:asc logs
-  super db load -q -use logs in.sup
+  super db create -q -orderby ts:asc -objectcap 2 -framecap 2 logs
+  super db load -q -use logs -inputcap 2 in.sup
   super db -s -c "from logs" > 1.sup
   super db -s -c "from logs" > 2.sup
 
diff --git a/db/ztests/ls.yaml b/db/ztests/ls.yaml
index 88adb8fb07..3595880ce9 100644
--- a/db/ztests/ls.yaml
+++ b/db/ztests/ls.yaml
@@ -34,7 +34,8 @@ outputs:
             ]
           }
         ]::order.SortKeys,
-        threshold: 524288000
+        objectcap: 1048576::uint64,
+        framecap: 20000::uint64
       }
       ===
       {
diff --git a/db/ztests/merge-by-addr.yaml b/db/ztests/merge-by-addr.yaml
index 3d2d64d336..8a5f5e6805 100644
--- a/db/ztests/merge-by-addr.yaml
+++ b/db/ztests/merge-by-addr.yaml
@@ -1,8 +1,8 @@
 script: |
   export SUPER_DB=test
   super db init -q
-  super db create -q -S 75B -orderby addr:asc logs
-  super db load -q -use logs in.sup
+  super db create -q -framecap 5 -objectcap 5 -orderby addr:asc logs
+  super db load -q -use logs -inputcap 5 in.sup
   super db -s -c 'from logs | cut addr'
 
 inputs:
diff --git a/db/ztests/meta.yaml b/db/ztests/meta.yaml
index 139c1946e4..5401acd290 100644
--- a/db/ztests/meta.yaml
+++ b/db/ztests/meta.yaml
@@ -41,7 +41,8 @@ outputs:
             ]
           }
         ]::order.SortKeys,
-        threshold: 524288000
+        objectcap: 1048576::uint64,
+        framecap: 20000::uint64
       }
       {
         name: "poolB",
@@ -53,7 +54,8 @@ outputs:
             ]
           }
         ]::order.SortKeys,
-        threshold: 524288000
+        objectcap: 1048576::uint64,
+        framecap: 20000::uint64
       }
       ===
       {
diff --git a/db/ztests/objectcap.yaml b/db/ztests/objectcap.yaml
new file mode 100644
index 0000000000..c6afc53c7e
--- /dev/null
+++ b/db/ztests/objectcap.yaml
@@ -0,0 +1,11 @@
+script: |
+  export SUPER_DB=test
+  super db init -q
+  super db create -use -q -objectcap 50 -framecap 5 test
+  seq 100 | super db load -q -inputcap 5 -
+  super db -s -c "from test:objects | count() by size:=count"
+
+outputs:
+  - name: stdout
+    data: |
+      {size:50::uint64,count:2}
diff --git a/db/ztests/rename.yaml b/db/ztests/rename.yaml
index 85e1623155..4a93f929d5 100644
--- a/db/ztests/rename.yaml
+++ b/db/ztests/rename.yaml
@@ -17,8 +17,8 @@ outputs:
     regexp: |
       pool [0-9a-f-]{36} renamed from p2 to p3
       ===
-      p1 [0-9a-f-]{36} key ts order desc
-      p3 [0-9a-f-]{36} key ts order desc
+      p1 [0-9a-f-]{36} key ts order desc objectcap 1048576 framecap 20000
+      p3 [0-9a-f-]{36} key ts order desc objectcap 1048576 framecap 20000
       ===
       pool deleted: p1
       ===
diff --git a/service/handlers.go b/service/handlers.go
index 117ca7bca9..689f1caeea 100644
--- a/service/handlers.go
+++ b/service/handlers.go
@@ -259,7 +259,7 @@ func handlePoolPost(c *Core, w *ResponseWriter, r *Request) {
 	if len(req.SortKeys.Keys) > 0 {
 		sortKeys = append(sortKeys, order.NewSortKey(req.SortKeys.Order, req.SortKeys.Keys[0]))
 	}
-	pool, err := c.root.CreatePool(r.Context(), req.Name, sortKeys, req.Thresh)
+	pool, err := c.root.CreatePool(r.Context(), req.Name, sortKeys, req.ObjectCap, req.FrameCap)
 	if err != nil {
 		w.Error(err)
 		return
diff --git a/service/ztests/curl-create.yaml b/service/ztests/curl-create.yaml
index 44db00dd84..f2a8ed9709 100644
--- a/service/ztests/curl-create.yaml
+++ b/service/ztests/curl-create.yaml
@@ -25,7 +25,8 @@ outputs:
               ]
             }
           ],
-          threshold: 524288000
+          objectcap: 1048576,
+          framecap: 20000
         },
         branch: {
           ts: 0,
diff --git a/service/ztests/curl-pool-rename.yaml b/service/ztests/curl-pool-rename.yaml
index 7d11c09df6..16c62ee9d2 100644
--- a/service/ztests/curl-pool-rename.yaml
+++ b/service/ztests/curl-pool-rename.yaml
@@ -31,5 +31,6 @@ outputs:
             ]
           }
         ],
-        threshold: 524288000
+        objectcap: 1048576,
+        framecap: 20000
       }
diff --git a/service/ztests/delete-where.yaml b/service/ztests/delete-where.yaml
index a2bda0720a..e70fe19082 100644
--- a/service/ztests/delete-where.yaml
+++ b/service/ztests/delete-where.yaml
@@ -2,8 +2,8 @@ script: |
   source service.sh
   for order in desc asc; do
     echo === $order === | tee >(cat >&2)
-    super db create -use -q -S 1KB -orderby ts:$order test
-    seq 1000 | super -c '{ts:this-1,s:f"val{this-1}"}' - | super db load -q -
+    super db create -use -q -objectcap 50 -framecap 10 -orderby ts:$order test
+    seq 1000 | super -c '{ts:this-1,s:f"val{this-1}"}' - | super db load -q -inputcap 10 -
     super db delete -q -where 'ts > 400 and ts <= 500'
     super db -s -c 'from test | count()'
     super db delete -q -where 's == "val1" or s == "val999"'
diff --git a/service/ztests/ignore-trailing-slash.yaml b/service/ztests/ignore-trailing-slash.yaml
index baffef54a4..98258a923b 100644
--- a/service/ztests/ignore-trailing-slash.yaml
+++ b/service/ztests/ignore-trailing-slash.yaml
@@ -10,4 +10,4 @@ inputs:
 outputs:
   - name: stdout
     regexp: |
-      test [0-9a-f-]{36} key ts order desc
+      test [0-9a-f-]{36} key ts order desc objectcap 1048576 framecap 20000
diff --git a/service/ztests/ls-pools.yaml b/service/ztests/ls-pools.yaml
index 47feb96517..383c6c3f22 100644
--- a/service/ztests/ls-pools.yaml
+++ b/service/ztests/ls-pools.yaml
@@ -24,8 +24,8 @@ outputs:
       pool created: p1 [0-9a-f-]{36}
       pool created: p2 [0-9a-f-]{36}
       ===
-      p1 [0-9a-f-]{36} key ts order desc
-      p2 [0-9a-f-]{36} key ts order desc
+      p1 [0-9a-f-]{36} key ts order desc objectcap 1048576 framecap 20000
+      p2 [0-9a-f-]{36} key ts order desc objectcap 1048576 framecap 20000
       ===
       {
         name: "p1"
diff --git a/service/ztests/offline-config-updates.yaml b/service/ztests/offline-config-updates.yaml
index 79ede236c0..9cb39e56b9 100644
--- a/service/ztests/offline-config-updates.yaml
+++ b/service/ztests/offline-config-updates.yaml
@@ -13,7 +13,7 @@ inputs:
 outputs:
   - name: stdout
     regexp: |
-      test [0-9a-f-]{36} key ts order desc
+      test [0-9a-f-]{36} key ts order desc objectcap 1048576 framecap 20000
       ===
   - name: stderr
     data: ""
diff --git a/sio/anyio/lookup.go b/sio/anyio/lookup.go
index 2fb6734dd6..a8368ff93f 100644
--- a/sio/anyio/lookup.go
+++ b/sio/anyio/lookup.go
@@ -34,7 +34,7 @@ func lookupReader(ctx context.Context, sctx *super.Context, r io.Reader, opts Re
 	case "line":
 		return newVioPuller(sctx, lineio.NewReader(r)), nil
 	case "json":
-		return jsonio.NewReader(context.Background(), sctx, r, opts.Pushdown, opts.ConcurrentReaders), nil
+		return jsonio.NewReader(context.Background(), sctx, r, opts.Pushdown, opts.ConcurrentReaders, opts.InputCap), nil
 	case "parquet":
 		return parquetio.NewReader(ctx, sctx, r, opts.Pushdown, opts.ConcurrentReaders)
 	case "sup":
diff --git a/sio/anyio/reader.go b/sio/anyio/reader.go
index 40dd5b5ce5..486202d534 100644
--- a/sio/anyio/reader.go
+++ b/sio/anyio/reader.go
@@ -29,6 +29,7 @@ type ReaderOpts struct {
 	Pushdown          sbuf.Pushdown
 	ConcurrentReaders int
 	CSV               csvio.ReaderOpts
+	InputCap          int
 }
 
 func NewReader(ctx context.Context, sctx *super.Context, r io.Reader, opts ReaderOpts) (vio.Puller, error) {
@@ -38,7 +39,11 @@ func NewReader(ctx context.Context, sctx *super.Context, r io.Reader, opts Reade
 	if opts.Format != "" && opts.Format != "auto" {
 		return lookupReader(ctx, sctx, r, opts)
 	}
-
+	if opts.InputCap != 0 {
+		// XXX There ought to be a better way to do this.  Mostly this
+		// is for testing so not a big deal.
+		sbuf.PullerBatchValues = opts.InputCap
+	}
 	track := NewTrack(r)
 
 	bsupErr := bsup.Probe(track)
@@ -74,7 +79,7 @@ func NewReader(ctx context.Context, sctx *super.Context, r io.Reader, opts Reade
 	// sake of tests.
 	jsonErr := isJSONStream(track, 10)
 	if jsonErr == nil {
-		return jsonio.NewReader(ctx, sctx, track.Reader(), opts.Pushdown, opts.ConcurrentReaders), nil
+		return jsonio.NewReader(ctx, sctx, track.Reader(), opts.Pushdown, opts.ConcurrentReaders, opts.InputCap), nil
 	}
 	jsonErr = fmt.Errorf("json: %w", jsonErr)
 	track.Reset()
@@ -152,7 +157,7 @@ func isCSVStream(track *Track, delim rune, name string) error {
 }
 
 func isJSONStream(track *Track, want int) error {
-	r := jsonio.NewValReader(track)
+	r := jsonio.NewValReader(track, 0)
 	for range want {
 		if _, err := r.Next(); err != nil {
 			if errors.Is(err, io.EOF) {
diff --git a/sio/dbio/writer.go b/sio/dbio/writer.go
index 220162b658..a6fea5cdad 100644
--- a/sio/dbio/writer.go
+++ b/sio/dbio/writer.go
@@ -4,6 +4,7 @@ import (
 	"bytes"
 	"fmt"
 	"io"
+	"strconv"
 	"strings"
 	"uuid"
 
@@ -117,6 +118,10 @@ func formatPoolConfig(b *bytes.Buffer, p *pools.Config) {
 	b.WriteString(p.SortKeys.Primary().Path.String())
 	b.WriteString(" order ")
 	b.WriteString(p.SortKeys.Primary().Order.String())
+	b.WriteString(" objectcap ")
+	b.WriteString(strconv.FormatUint(p.ObjectCap, 10))
+	b.WriteString(" framecap ")
+	b.WriteString(strconv.FormatUint(p.FrameCap, 10))
 	b.WriteByte('\n')
 }
 
diff --git a/sio/jsonio/reader.go b/sio/jsonio/reader.go
index 8732246fe1..964afc2a2f 100644
--- a/sio/jsonio/reader.go
+++ b/sio/jsonio/reader.go
@@ -17,7 +17,7 @@ import (
 	"github.com/superdb/super/vector"
 )
 
-var VecBatchSize uint32 = 10 * 1024
+const defaultInputCap = 10 * 1024
 
 type Reader struct {
 	sctx       *super.Context
@@ -28,7 +28,10 @@ type Reader struct {
 	hasClosed atomic.Bool
 }
 
-func NewReader(ctx context.Context, sctx *super.Context, r io.Reader, p sbuf.Pushdown, concurrentReaders int) *Reader {
+func NewReader(ctx context.Context, sctx *super.Context, r io.Reader, p sbuf.Pushdown, concurrentReaders, inputCap int) *Reader {
+	if inputCap == 0 {
+		inputCap = defaultInputCap
+	}
 	var fields field.List
 	if p != nil {
 		fields = p.Projection().Paths()
@@ -36,7 +39,7 @@ func NewReader(ctx context.Context, sctx *super.Context, r io.Reader, p sbuf.Pus
 	return &Reader{
 		sctx:       sctx,
 		ctx:        ctx,
-		stream:     newStream(ctx, r, concurrentReaders),
+		stream:     newStream(ctx, r, concurrentReaders, inputCap),
 		projection: fields,
 	}
 }
diff --git a/sio/jsonio/stream.go b/sio/jsonio/stream.go
index 4bb9bf0e9f..df156c75c6 100644
--- a/sio/jsonio/stream.go
+++ b/sio/jsonio/stream.go
@@ -10,19 +10,21 @@ import (
 )
 
 type stream struct {
-	r    io.Reader
-	ch   chan result
-	done chan struct{}
-	once sync.Once
-	ctx  context.Context
+	r        io.Reader
+	ch       chan result
+	done     chan struct{}
+	once     sync.Once
+	ctx      context.Context
+	inputCap int
 }
 
-func newStream(ctx context.Context, r io.Reader, n int) *stream {
+func newStream(ctx context.Context, r io.Reader, n int, inputCap int) *stream {
 	return &stream{
-		r:    r,
-		ch:   make(chan result, n),
-		ctx:  ctx,
-		done: make(chan struct{}),
+		r:        r,
+		ch:       make(chan result, n),
+		ctx:      ctx,
+		done:     make(chan struct{}),
+		inputCap: inputCap,
 	}
 }
 
@@ -44,7 +46,7 @@ func (s *stream) next() (*vector.BytesTable, int, error) {
 
 func (s *stream) run() {
 	defer close(s.ch)
-	r := NewValReader(s.r)
+	r := NewValReader(s.r, s.inputCap)
 	for {
 		batch, startLineNum, err := readBatch(r)
 		select {
@@ -66,20 +68,20 @@ func (s *stream) close() {
 
 var bytesTablePool sync.Pool
 
-func newBytesTable() *vector.BytesTable {
+func newBytesTable(inputCap int) *vector.BytesTable {
 	b, ok := bytesTablePool.Get().(*vector.BytesTable)
 	if !ok {
-		b = new(vector.NewBytesTableEmpty(VecBatchSize))
+		b = new(vector.NewBytesTableEmpty(uint32(inputCap)))
 	}
 	b.Reset()
 	return b
 }
 
 func readBatch(r *valReader) (*vector.BytesTable, int, error) {
-	t := newBytesTable()
+	t := newBytesTable(r.inputCap)
 	start := r.lineNumber()
 	var err error
-	for range VecBatchSize {
+	for range r.inputCap {
 		var b []byte
 		if b, err = r.Next(); err != nil {
 			break
diff --git a/sio/jsonio/valreader.go b/sio/jsonio/valreader.go
index e094797ab1..848f3a0e2d 100644
--- a/sio/jsonio/valreader.go
+++ b/sio/jsonio/valreader.go
@@ -13,15 +13,16 @@ import (
 const maxSize = 512 * 1024 * 1024
 
 type valReader struct {
-	r      io.Reader
-	buf    []byte
-	cursor []byte
-	EOF    bool
-	line   int
+	r        io.Reader
+	buf      []byte
+	cursor   []byte
+	EOF      bool
+	line     int
+	inputCap int
 }
 
-func NewValReader(r io.Reader) *valReader {
-	return &valReader{r: r, buf: make([]byte, 512*1024)}
+func NewValReader(r io.Reader, inputCap int) *valReader {
+	return &valReader{r: r, buf: make([]byte, 512*1024), inputCap: inputCap}
 }
 
 func (r *valReader) Next() ([]byte, error) {