Explainer · 5 min read

Checkpoint restart and resume: why a 90% transfer should never start from zero

What checkpoint restart means in file transfer, how block-level resume differs from file-level resume, why browser uploads usually restart, and what RIPTON CLOUD does on an interruption.

Author
RIPTON CLOUD team, Product and engineering
Technical review
RIPTON CLOUD engineering
Updated

Key takeaways

  • File-level resume repeats the whole file that was interrupted; block-level resume repeats only the missing blocks.
  • Most browser uploads and email-style tools restart from zero after a dropped connection.
  • A verified checkpoint is only useful if the receiver can prove what it already has.
  • RIPTON CLOUD resumes browser batches from the last completed batch and Desktop transfers from the last verified block, and verifies a package before marking it delivered.

Three kinds of resume

  • None: the connection drops and the transfer begins again. Typical of email attachments, many web upload forms and consumer sharing links.
  • File-level: the tool knows which files completed and skips them, but a file interrupted at 90% is sent again in full. rsync and most sync clients behave this way for a single large file.
  • Block-level: the receiver tracks which blocks of each file arrived and asks only for the rest. An interruption costs the blocks in flight, not the file.

Why verification is part of resume

Resuming from a checkpoint is only safe if both sides agree on what the receiver already holds. A protocol that resumes from a byte offset without checking can splice a corrupted tail onto a good head. Block-level protocols keep a map of received blocks and verify the assembled file before declaring it complete; RIPTON CLOUD marks a package delivered only after that verification.

What RIPTON CLOUD does on an interruption

PathOn interruptionOn completion
Browser (Web Edge)Resumes from the last completed batch of files; large files resume by chunk.Package assembled on the node and verified.
RIPTON Desktop and node-to-node (RIPT)Resumes from the last verified block; missing blocks are retransmitted, not the file.Every block accounted for before the package is marked delivered.
Recipient downloadBrowser download managers handle resume per their own rules; the archive is served as one stream.Download events recorded per recipient.

How to test resume in an evaluation

  1. 1Start a transfer of a multi-gigabyte file or a large directory.
  2. 2At around 60 to 90 percent, pull the network cable or stop the process.
  3. 3Restore and restart. Record how much was repeated and the total elapsed time.
  4. 4Compare the delivered files with the source. A resume that saves time but corrupts data is worse than a restart.

Related

Interrupting a run on purpose is part of every RIPTON CLOUD proof of concept, so you see resume behaviour on your own route.

Plan a proof of concept