Resource Limit Increases for EdgeWorkers Yield Big Results
Exciting news! Akamai recently launched a batch of resource limit increases for the EdgeWorkers platform that dramatically reduced errors resulting from resource limit constraints. The limit increases came as two releases in the fall of 2022 — the first in September and the second in late November.
Additional architectural changes to enable dynamic bursting of CPU/memory and wall time through their limits (when possible) were also released in the November rollout. These combined changes unlock even more opportunities for businesses to harness the power of EdgeWorkers to boost performance by running their business logic workloads on the edge, closer to their end consumers.
Immediate results
EdgeWorkers engineers have measured an overall 10-fold reduction of resource-based error rates for EdgeWorkers compared with historical averages. Specifically, the error rates moved from an historical average percentage of requests of 0.01% (which isn’t too bad to begin with) to an average of 0.0007% in December.
This is worth repeating: The average resource-based error rates were reduced by more than a factor of 10.
The impact of the resource limit increases
Here are a few graphs to illustrate the impact of the resource limit increases a bit more. On the day the change was rolled into production, EdgeWorkers teams noticed an immediate reduction in errors (Figure 1).
Fig. 1: The decrease in resource-based errors on the day of the last update illustrates the immediate impact of the engineering changes
Figure 2, which gives a wider view of the errors starting at the beginning of November, shows the dramatic improvements.
Fig. 2: Resource-based errors for the month of November
Figure 3 illustrates the CPU timeout errors alone beginning in late September.
Fig. 3: CPU timeout errors from September through November
Finally, Figure 4 highlights the overall average error rates for a two-month window.
Fig. 4: The overall average resource-based error rates from October through November
What changed exactly?
The limit changes are automatically applied — there is nothing EdgeWorkers developers need to do to take advantage of these increases. The EdgeWorkers Basic Compute tier also received an upgrade inheriting the previous settings of the Dynamic Compute tier. Full details of the changes are outlined in our EdgeWorkers Release Notes.
Details of the rate increases
The table outlines the details of the rate increases in CPU, wall time, and memory for the Dynamic Compute EdgeWorkers tier.
Event handlers |
Limit type |
Original value |
New values |
|---|---|---|---|
on*Event handlers |
CPU time |
10 ms |
20 ms |
Wall time |
4 s |
5.5 s |
|
Memory |
1.5 MB |
2.5 MB |
|
Subrequests |
1 |
4 |
|
Response provider |
CPU time |
100 ms |
200 ms |
Wall time |
4 s |
8 s |
|
Memory |
1.5 MB |
4 MB |
|
Subrequests |
50 |
50 |
Table: The CPU, wall time, and memory upgrades
What are customers saying?
Throughout 2022, EdgeWorkers engineering teams have been focused on the reduction of error rates; - this increase in resource limits is a culmination of their efforts and analysis. Teams evaluated options, ran multiple models and tests, and were able to achieve this monumental shift in reducing resource-limit errors.
Some hawkeyed EdgeWorkers customers noticed the improvements right away and sent us messages via Slack asking if everything was okay since the shift was so dramatic. One customer remarked that Christmas had come early!
Summary
A significant increase in resource limits combined with architectural changes have resulted in a noteworthy 10x reduction in EdgeWorkers error rates for customers. While the Akamai EdgeWorkers team continues to monitor results, they will always look for ways to improve performance, the feature set, and the developer experience of the EdgeWorkers platform.
Find out more
Visit techdocs.akamai.com/edgeworkers/docs for more EdgeWorkers resources, and sign up for a free trial today. Join Akamai’s Serverless Slack Channel to talk directly with EdgeWorkers experts, architects, developers, and product managers; to get the latest news on EdgeWorkers; and to participate in the conversation about upcoming product features.
Resources
For more information, please refer to these resources:
