You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@openwhisk.apache.org by OpenWhisk Team Slack <ra...@apache.org> on 2019/12/07 09:33:01 UTC
[slack-digest] [2019-12-06] #random
2019-12-06 01:45:43 UTC - Bill Zong: Haven’t used lambda before.
But what’s the essence of AWS lambda?
Is it container or just some java-like web service?
https://openwhisk-team.slack.com/archives/C3UDXSFA6/p1575596743014100?thread_ts=1575478548.013100&cid=C3UDXSFA6
----
2019-12-06 15:08:00 UTC - chetanm: <https://aws.amazon.com/builders-library/avoiding-insurmountable-queue-backlogs/?did=ba_card&trk=ba_card|https://aws.amazon.com/builders-library/avoiding-insurmountable-queue-backlogs/?did=ba_card&trk=ba_card>
https://openwhisk-team.slack.com/archives/C3UDXSFA6/p1575644880014400
----
2019-12-06 15:08:53 UTC - chetanm: > Shuffle-sharding
> AWS Lambda is an example of a system where polling a separate queue for every Lambda customer would be too costly. However, having a single queue could result in some of the issues described in this article. So rather than using one queue, AWS Lambda provisions a fixed number of queues, and hashes each customer to a small number of queues. Before enqueueing a message, it checks to see which of those targeted queues contains the fewest messages, and enqueues into that one. When one customer's workload increases, it will drive a backlog in its mapped queues, but other workloads will automatically be routed away from those queues. It doesn't take a large number of queues to build in some magical resource isolation. This is only one of the many protections built into Lambda, but it is a technique that is also used in other services at Amazon.
https://openwhisk-team.slack.com/archives/C3UDXSFA6/p1575644933015200
----
2019-12-06 15:09:48 UTC - chetanm: Pretty similar to the the LB approach implemented in OpenWhisk
https://openwhisk-team.slack.com/archives/C3UDXSFA6/p1575644988016100?thread_ts=1575644988.016100&cid=C3UDXSFA6
----
2019-12-06 15:37:40 UTC - Carlos Santana: I agree, This part is interesting :thinking_face:
> but other workloads will automatically be routed away from those queues
In OpenWhisk this in theory could be possible via ingesting some metrics into the LB, and the LB adjusting/rotate the hash algorithm for the other not hot actions, leaving the wot one keep hitting the same invoker
https://openwhisk-team.slack.com/archives/C3UDXSFA6/p1575646660016200?thread_ts=1575644988.016100&cid=C3UDXSFA6
----