Loading…
Topic
Concurrency
2 posts about Concurrency. Every summary links to the original.
- 2 posts
- 2 companies
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
Newest first
Posts
How we raised Databricks cluster rate limits by 7.5×
Databricks’ Compute Platform team replaced fixed request-rate limits for cluster creation with adaptive concurrency limits to better handle long-running, variable-cost operations. Fixed limits turned away requests during healthy periods yet accepted too many when dependencies slowed, requiring manual tuning. Adaptive concurrency tracks active provisioning workflows with slots and adjusts the slot count based on congestion signals such as provisioning latency and thread pool saturation. The rollout analysis reported a 7.5× increase in default cluster-creation request-rate limits, more than 99.99% fewer rate-limit errors, and removal of over a hundred customer-specific overrides. The fixed API request-rate limit remains as a final safety guardrail against exceptionally high incoming traffic.
Your .NET parallel loop can become serial in a one-CPU container
In a .NET container with a one-CPU limit, Parallel.ForEachAsync defaults its MaxDegreeOfParallelism to Environment.ProcessorCount, which returns 1, causing I/O-bound calls to run serially. The author’s lab processed 200 records, each requiring three enrichment calls and one write, taking 83.6 seconds with one CPU versus 42.1 seconds with two. The solution uses a DependencyBudgets class with separate SemaphoreSlim instances per dependency to cap concurrency independently. Strategy C4, with 24 enrichment permits and 6 write permits, avoided all HTTP 503 rejections while accepting 35.08 records per second in a slow write scenario. The author recommends configuring per-dependency concurrency limits, validating them at startup, and using separate timeouts for service calls and permit waits.
Related topics
All topics →Weekly digest
The best engineering reads, once a week.