Rootconf Hyderabad edition
Rootconf
Rootconf

Rootconf Hyderabad edition

On SRE, systems engineering and distributed systems

Tickets

Loading…

All submissions
This submission has been added to the schedule

Powered by VideoKen

Powered by VideoKen

Session video
Submission video

Ankur Jain

@anvinjain

Fast object distribution using P2P

Submitted May 11, 2019

Section: Full talk Technical level: Beginner Session type: Discussion Category: Distributed systems Section: Full talk (40 mins)

In Flipkart, objects like VM/container images, media, ML models, etc are stored in a homegrown object storage service. To improve download latencies and avoid getting rate limited, teams normally introduce nginx or some other cache to access frequently used objects. This seemingly works for small objects, less dynamic content or when only few clients are downloading concurrently. But for large objects which are required to be downloaded concurrently by hundreds of clients, network bandwidth of cache cluster gets choked, thereby increasing download latency. Cache misses further worsen the total download duration.

Shatabdi solves the above problem by transparently converting any given http resource to a torrent and downloading it on all participating clients in peer-to-peer(p2p) fashion. It can also be extended to support non-http sources of truth as well. Shatabdi daemon abstracts out all the noise and provides a REST API hosted locally to manage downloads/uploads.

Presently, Shatabdi has gone live in Flipkart datacenters and used for distributing VM and container images to motherships. We foresee this platform to be used for a myriad of other scenarios, such as:

  • Warming up nodes in caching cluster on startup

  • Distributing pre-computed datasets to multiple instances of a datastore

  • Distributing ML models that are required by multiple nodes

Some numbers for the interested, these involve no cache apart from a single download from a http source of truth(object storage system in Flipkart):

  • Disk-based download
    512 clients downloading 3.2GB concurrently in ~(2-4 mins) with upload speed capped to 24 mbps

  • TmpFS-based download
    50 clients downloading 3.2GB concurrently in ~15-25seconds with warmed up seeder

This talk will focus on the motivation behind building Shatabdi and a technical deep dive in its implementation. Broadly, we will cover performance tuning of open-source torrent clients, functional flows to ensure maximal bandwidth consumption and always healthy magnets and benchmarking large deployments container based clusters in Flipkart.

For any usecase involving object distribution at scale for large object size and a lot of concurrent clients, this talk should give valuable insights around architecting solutions using p2p.

Outline

  • Motivation for fast object distribution
  • Goals for a torrent-based distribution layer
  • Exploring open-source alternatives
  • Design, architecture and functional flows of Shatabdi
    • Indexing store and leader election using etcd
    • Achieving HA for torrent tracker
  • Performance tuning of libtorrent
  • Benchmarking container deployments powered by Shatabdi
  • Adoption @ Flipkart

Speaker bio

Ankur works in Platform-as-a-service team in Flipkart Cloud Platform where he builds dev-centric platforms that work at Flipkart scale.

Slides

https://docs.google.com/presentation/d/144puJStK8ZVvd43XRRP5Jxi9TBGxzdg0xx1CTmmmGyw/edit?usp=sharing

All submissions

Comments

Login to leave a comment

  • AS

    Anwesha Sarkar

    @anweshaalt

    Hello Ankur,

    It was a good rehearsal taday. Here are the feedback from the rehearsal:

    1. Why are the target audience of the talk? And why it is so?
    2. Explain the goal of the talk in the start.
    3. Include some references and resources.
    4. Include diagrams.
    5. Include the contact details in the last slides
    6. Include a demo.
    7. Avoid having text heavy slides.
    8. Include comparison of Shatabdi with other open source alternatives.
    9. Include the failures.
    10. Change the title of the talk. Construct it around the problem and not around the tool.
    11. Explain the terminologies in a seperate slide.

    Submit your revised slides by 10th September 2019.

    Regards
    Anwesha

    Posted 5 years ago

  • saurabh hirani

    @saurabh_hirani

    Hi Ankur and Yashpal,

    First of all - clever name :)

    Thank you for submitting the proposal. This seems like an interesting approach. I had the following queries:

    1. The distribution overview image has D42 - is that https://www.device42.com/ or something else?
    2. Can you add slides + preview video explaining the flow. An approach that really works well in such a proposal would be -
      2.1 State of the world before Shatabdi (not the train), the problems, some stats
      2.2 What were the design goals for Shatabdi
      2.3 What problems did it solve + a before/after comparison
    3. Is Shatabdi open source? Do you plan to open source it? If it isn't and there are no plans to do so - the audience would benefit more from the architectural aspects of how it solved a specific problem it because they won't have the tool to play with after the talk. If it is - then please link the relevant github repos and call it out as it will help your talk gain more traction.
    Posted 5 years ago
Hybrid access (members only)

Hosted by

Rootconf
Rootconf
We care about site reliability, cloud costs, security and data privacy