diff --git a/sync/src/blocks.rs b/sync/src/blocks.rs
new file mode 100644
index 000000000..683db0c7b
--- /dev/null
+++ b/sync/src/blocks.rs
@@ -0,0 +1,413 @@
+// Copyright 2015, 2016 Ethcore (UK) Ltd.
+// This file is part of Parity.
+
+// Parity is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+
+// Parity is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+
+// You should have received a copy of the GNU General Public License
+// along with Parity.  If not, see <http://www.gnu.org/licenses/>.
+
+use util::*;
+use ethcore::header::{ Header as BlockHeader};
+
+known_heap_size!(0, HeaderId, SyncBlock);
+
+/// Block data with optional body.
+struct SyncBlock {
+	header: Bytes,
+	body: Option<Bytes>,
+}
+
+/// Used to identify header by transactions and uncles hashes
+#[derive(Eq, PartialEq, Hash)]
+struct HeaderId {
+	transactions_root: H256,
+	uncles: H256
+}
+
+/// A collection of blocks and subchain pointers being downloaded. This keeps track of
+/// which headers/bodies need to be downloaded, which are being downloaded and also holds
+/// the downloaded blocks.
+pub struct BlockCollection {
+	/// Heads of subchains to download
+	heads: Vec<H256>,
+	/// Downloaded blocks.
+	blocks: HashMap<H256, SyncBlock>,
+	/// Downloaded blocks by parent.
+	parents: HashMap<H256, H256>,
+	/// Used to map body to header.
+	header_ids: HashMap<HeaderId, H256>,
+	/// First block in `blocks`.
+	head: Option<H256>,
+	/// Set of block header hashes being downloaded
+	downloading_headers: HashSet<H256>,
+	/// Set of block bodies being downloaded identified by block hash.
+	downloading_bodies: HashSet<H256>,
+}
+
+impl BlockCollection {
+	/// Create a new instance.
+	pub fn new() -> BlockCollection {
+		BlockCollection {
+			blocks: HashMap::new(),
+			header_ids: HashMap::new(),
+			heads: Vec::new(),
+			parents: HashMap::new(),
+			head: None,
+			downloading_headers: HashSet::new(),
+			downloading_bodies: HashSet::new(),
+		}
+	}
+
+	/// Clear everything.
+	pub fn clear(&mut self) {
+		self.blocks.clear();
+		self.parents.clear();
+		self.header_ids.clear();
+		self.heads.clear();
+		self.head = None;
+		self.downloading_headers.clear();
+		self.downloading_bodies.clear();
+	}
+
+	/// Reset collection for a new sync round with given subchain block hashes.
+	pub fn reset_to(&mut self, hashes: Vec<H256>) {
+		self.clear();
+		self.heads = hashes;
+	}
+
+	/// Insert a set of headers into collection and advance subchain head pointers.
+	pub fn insert_headers(&mut self, headers: Vec<Bytes>) {
+		for h in headers.into_iter() {
+			if let Err(e) =  self.insert_header(h) {
+				trace!(target: "sync", "Ignored invalid header: {:?}", e);
+			}
+		}
+		self.update_heads();
+	}
+
+	/// Insert a collection of block bodies for previously downloaded headers.
+	pub fn insert_bodies(&mut self, bodies: Vec<Bytes>) {
+		for b in bodies.into_iter() {
+			if let Err(e) =  self.insert_body(b) {
+				trace!(target: "sync", "Ignored invalid body: {:?}", e);
+			}
+		}
+	}
+
+	/// Returns a set of block hashes that require a body download. The returned set is marked as being downloaded.
+	pub fn needed_bodies(&mut self, count: usize, _ignore_downloading: bool) -> Vec<H256> {
+		if self.head.is_none() {
+			return Vec::new();
+		}
+		let mut needed_bodies: Vec<H256> = Vec::new();
+		let mut head = self.head;
+		while head.is_some() && needed_bodies.len() < count {
+			head = self.parents.get(&head.unwrap()).cloned();
+			if let Some(head) = head {
+				match self.blocks.get(&head) {
+					Some(block) if block.body.is_none() && !self.downloading_bodies.contains(&head) => {
+						needed_bodies.push(head.clone());
+					}
+					_ => (),
+				}
+			}
+		}
+		self.downloading_bodies.extend(needed_bodies.iter());
+		needed_bodies
+	}
+
+	/// Returns a set of block hashes that require a header download. The returned set is marked as being downloaded.
+	pub fn needed_headers(&mut self, count: usize, ignore_downloading: bool) -> Option<(H256, usize)> {
+		// find subchain to download
+		let mut download = None;
+		{
+			for h in &self.heads {
+				if ignore_downloading || !self.downloading_headers.contains(&h) {
+					self.downloading_headers.insert(h.clone());
+					download = Some(h.clone());
+					break;
+				}
+			}
+		}
+		download.map(|h| (h, count))
+	}
+
+	/// Unmark a header as being downloaded.
+	pub fn clear_header_download(&mut self, hash: &H256) {
+		self.downloading_headers.remove(hash);
+	}
+
+	/// Unmark a block body as being downloaded.
+	pub fn clear_body_download(&mut self, hash: &H256) {
+		self.downloading_bodies.remove(hash);
+	}
+
+	/// Get a valid chain of blocks ordered in descending order and ready for importing into blockchain.
+	pub fn drain(&mut self) -> Vec<Bytes> {
+		if self.blocks.is_empty() || self.head.is_none() {
+			return Vec::new();
+		}
+
+		let mut drained = Vec::new();
+		let mut hashes = Vec::new();
+		{
+			let mut blocks = Vec::new();
+			let mut head = self.head;
+			while head.is_some() {
+				head = self.parents.get(&head.unwrap()).cloned();
+				if let Some(head) = head {
+					match self.blocks.get(&head) {
+						Some(block) if block.body.is_some() => {
+							blocks.push(block);
+							hashes.push(head);
+							self.head = Some(head);
+						}
+						_ => break,
+					}
+				}
+			}
+
+			for block in blocks.drain(..) {
+				let mut block_rlp = RlpStream::new_list(3);
+				block_rlp.append_raw(&block.header, 1);
+				let body = Rlp::new(&block.body.as_ref().unwrap()); // incomplete blocks are filtered out in the loop above
+				block_rlp.append_raw(body.at(0).as_raw(), 1);
+				block_rlp.append_raw(body.at(1).as_raw(), 1);
+				drained.push(block_rlp.out());
+			}
+		}
+		for h in hashes {
+			self.blocks.remove(&h);
+		}
+		trace!("Drained {} blocks, new head :{:?}", drained.len(), self.head);
+		drained
+	}
+
+	/// Check if the collection is empty. We consider the syncing round complete once
+	/// there is no block data left and only a single or none head pointer remains.
+	pub fn is_empty(&self) -> bool {
+		return self.heads.len() == 0 ||
+			(self.heads.len() == 1 && self.head.map_or(false, |h| h == self.heads[0]))
+	}
+
+	/// Chech is collection contains a block header.
+	pub fn contains(&self, hash: &H256) -> bool {
+		self.blocks.contains_key(hash)
+	}
+
+	/// Return heap size.
+	pub fn heap_size(&self) -> usize {
+		//TODO: other collections
+		self.blocks.heap_size_of_children()
+	}
+
+	/// Check if given block hash is marked as being downloaded.
+	pub fn is_downloading(&self, hash: &H256) -> bool {
+		self.downloading_headers.contains(hash) || self.downloading_bodies.contains(hash)
+	}
+
+	fn insert_body(&mut self, b: Bytes) -> Result<(), UtilError> {
+		let body = UntrustedRlp::new(&b);
+		let tx = try!(body.at(0));
+		let tx_root = ordered_trie_root(tx.iter().map(|r| r.as_raw().to_vec()).collect()); //TODO: get rid of vectors here
+		let uncles = try!(body.at(1)).as_raw().sha3();
+		let header_id = HeaderId {
+			transactions_root: tx_root,
+			uncles: uncles
+		};
+		match self.header_ids.get(&header_id).cloned() {
+			Some(h) => {
+				self.header_ids.remove(&header_id);
+				self.downloading_bodies.remove(&h);
+				match self.blocks.get_mut(&h) {
+					Some(ref mut block) => {
+						trace!(target: "sync", "Got body {}", h);
+						block.body = Some(body.as_raw().to_vec());
+					},
+					None => warn!("Got body with no header {}", h)
+				}
+			}
+			None => trace!(target: "sync", "Ignored unknown/stale block body")
+		};
+		Ok(())
+	}
+
+	fn insert_header(&mut self, header: Bytes) -> Result<H256, UtilError> {
+		let info: BlockHeader = try!(UntrustedRlp::new(&header).as_val());
+		let hash = info.hash();
+		if self.blocks.contains_key(&hash) {
+			return Ok(hash);
+		}
+		match self.head {
+			None if hash == self.heads[0] => {
+				trace!("New head {}", hash);
+				self.head = Some(info.parent_hash);
+			},
+			_ => ()
+		}
+
+		let mut block = SyncBlock {
+			header: header,
+			body: None,
+		};
+		let header_id = HeaderId {
+			transactions_root: info.transactions_root,
+			uncles: info.uncles_hash
+		};
+		if header_id.transactions_root == rlp::SHA3_NULL_RLP && header_id.uncles == rlp::SHA3_EMPTY_LIST_RLP {
+			// empty body, just mark as downloaded
+			let mut body_stream = RlpStream::new_list(2);
+			body_stream.append_raw(&rlp::NULL_RLP, 1);
+			body_stream.append_raw(&rlp::EMPTY_LIST_RLP, 1);
+			block.body = Some(body_stream.out());
+		}
+		else {
+			self.header_ids.insert(header_id, hash.clone());
+		}
+
+		self.parents.insert(info.parent_hash.clone(), hash.clone());
+		self.blocks.insert(hash.clone(), block);
+		Ok(hash)
+	}
+
+	// update subchain headers
+	fn update_heads(&mut self) {
+		let mut new_heads = Vec::new();
+		let old_subchains: HashSet<_> = { self.heads.iter().map(Clone::clone).collect() };
+		for s in self.heads.drain(..) {
+			let mut h = s.clone();
+			loop {
+				match self.parents.get(&h) {
+					Some(next) => {
+						h = next.clone();
+						if old_subchains.contains(&h) {
+							trace!("Completed subchain {:?}", s);
+							break; // reached head of the other subchain, merge by not adding
+						}
+					},
+					_ => {
+						new_heads.push(h);
+						break;
+					}
+				}
+			}
+		}
+		self.heads = new_heads;
+	}
+}
+
+#[cfg(test)]
+mod test {
+	use super::BlockCollection;
+	use ethcore::client::{TestBlockChainClient, EachBlockWith, BlockId, BlockChainClient};
+	use ethcore::views::HeaderView;
+	use ethcore::header::BlockNumber;
+	use util::*;
+
+	fn is_empty(bc: &BlockCollection) -> bool {
+		bc.heads.is_empty() &&
+		bc.blocks.is_empty() &&
+		bc.parents.is_empty() &&
+		bc.header_ids.is_empty() &&
+		bc.head.is_none() &&
+		bc.downloading_headers.is_empty() &&
+		bc.downloading_bodies.is_empty()
+	}
+
+	#[test]
+	fn create_clear() {
+		let mut bc = BlockCollection::new();
+		assert!(is_empty(&bc));
+		let client = TestBlockChainClient::new();
+		client.add_blocks(100, EachBlockWith::Nothing);
+		let hashes = (0 .. 100).map(|i| (&client as &BlockChainClient).block_hash(BlockId::Number(i)).unwrap()).collect();
+		bc.reset_to(hashes);
+		assert!(!is_empty(&bc));
+		bc.clear();
+		assert!(is_empty(&bc));
+	}
+
+	#[test]
+	fn insert_headers() {
+		let mut bc = BlockCollection::new();
+		assert!(is_empty(&bc));
+		let client = TestBlockChainClient::new();
+		let nblocks = 200;
+		client.add_blocks(nblocks, EachBlockWith::Nothing);
+		let blocks: Vec<_> = (0 .. nblocks).map(|i| (&client as &BlockChainClient).block(BlockId::Number(i as BlockNumber)).unwrap()).collect();
+		let headers: Vec<_> = blocks.iter().map(|b| Rlp::new(b).at(0).as_raw().to_vec()).collect();
+		let hashes: Vec<_> = headers.iter().map(|h| HeaderView::new(h).sha3()).collect();
+		let heads: Vec<_> = hashes.iter().enumerate().filter_map(|(i, h)| if i % 20 == 0 { Some(h.clone()) } else { None }).collect();
+		bc.reset_to(heads);
+		assert!(!bc.is_empty());
+		assert_eq!(hashes[0], bc.heads[0]);
+		assert!(bc.needed_bodies(1, false).is_empty());
+		assert!(!bc.contains(&hashes[0]));
+		assert!(!bc.is_downloading(&hashes[0]));
+
+		let (h, n) = bc.needed_headers(6, false).unwrap();
+		assert!(bc.is_downloading(&hashes[0]));
+		assert_eq!(hashes[0], h);
+		assert_eq!(n, 6);
+		assert_eq!(bc.downloading_headers.len(), 1);
+		assert!(bc.drain().is_empty());
+
+		bc.insert_headers(headers[0..6].to_vec());
+		assert_eq!(hashes[5], bc.heads[0]);
+		for h in &hashes[0..6] {
+			bc.clear_header_download(h)
+		}
+		assert_eq!(bc.downloading_headers.len(), 0);
+		assert!(!bc.is_downloading(&hashes[0]));
+		assert!(bc.contains(&hashes[0]));
+
+		assert_eq!(&bc.drain()[..], &blocks[0..6]);
+		assert!(!bc.contains(&hashes[0]));
+		assert_eq!(hashes[5], bc.head.unwrap());
+
+		let (h, _) = bc.needed_headers(6, false).unwrap();
+		assert_eq!(hashes[5], h);
+		let (h, _) = bc.needed_headers(6, false).unwrap();
+		assert_eq!(hashes[20], h);
+		bc.insert_headers(headers[10..16].to_vec());
+		assert!(bc.drain().is_empty());
+		bc.insert_headers(headers[5..10].to_vec());
+		assert_eq!(&bc.drain()[..], &blocks[6..16]);
+		assert_eq!(hashes[15], bc.heads[0]);
+
+		bc.insert_headers(headers[16..].to_vec());
+		bc.drain();
+		assert!(bc.is_empty());
+	}
+
+	#[test]
+	fn insert_headers_with_gap() {
+		let mut bc = BlockCollection::new();
+		assert!(is_empty(&bc));
+		let client = TestBlockChainClient::new();
+		let nblocks = 200;
+		client.add_blocks(nblocks, EachBlockWith::Nothing);
+		let blocks: Vec<_> = (0 .. nblocks).map(|i| (&client as &BlockChainClient).block(BlockId::Number(i as BlockNumber)).unwrap()).collect();
+		let headers: Vec<_> = blocks.iter().map(|b| Rlp::new(b).at(0).as_raw().to_vec()).collect();
+		let hashes: Vec<_> = headers.iter().map(|h| HeaderView::new(h).sha3()).collect();
+		let heads: Vec<_> = hashes.iter().enumerate().filter_map(|(i, h)| if i % 20 == 0 { Some(h.clone()) } else { None }).collect();
+		bc.reset_to(heads);
+
+		bc.insert_headers(headers[2..22].to_vec());
+		assert_eq!(hashes[0], bc.heads[0]);
+		assert_eq!(hashes[21], bc.heads[1]);
+		assert!(bc.head.is_none());
+		bc.insert_headers(headers[0..2].to_vec());
+		assert!(bc.head.is_some());
+		assert_eq!(hashes[21], bc.heads[0]);
+	}
+}
+