CDN edge computing solutions

[pdgetrf]

Research the solutions from commercial CDNs that provide edge computing solution.

Note the scope of this is in edge computing related features.

A few examples of such CDNs:

• Cloudflare
• Fastly
• Akamai
• anything else?

We want to understand

• what kind of user cases?
• what these offer?
• what their designs are?
• what kind of feature and performance they claim to provide (e.g. kv store)?
• any relevant/prominent we want to look into adding in our Fornax project?
• any trend for future edge computing direction?

Depth: High-level

Time estimate: ~a week, by 12/20, possibly talk at next week communication meeting (Monday 5pm)

[pdgetrf]

some reference on KV store

1. Akamai's EdgeKV
2. Cloudflare's Workers KV
3. Fastly's Edge Dictionaries
4. RIAK DB's RIAK KV, and usage in League of Legends

How do these differ? what user cases do these handle respectively? In specific, how do these support cross-cluster storage, if at all?

[kxu1985]

Links for high-level overview of some CDN edge solutions.

Cloudflare:
https://blog.cloudflare.com/cloudflare-workers-serverless-week/
https://blog.cloudflare.com/cloud-computing-without-containers/

Fastly:
https://explore.fastly.com/compute-edge-promotion
https://www.fastly.com/blog/announcing-lucet-fastly-native-webassembly-compiler-runtime
https://www.fastly.com/blog/announcing-edge-dictionaries-make-faster-decisions-edge

Cloudflare vs. Fastly:
https://www.bizety.com/2020/08/29/cloudflare-vs-fastly-the-future-of-cdn-edge-computing/

Akamai:
https://www.akamai.com/products/serverless-computing-edgeworkers
https://www.akamai.com/content/dam/site/en/documents/product-brief/akamai-edgeworkers-product-brief.pdf

Amazon CloudFront:
https://aws.amazon.com/cloudfront/
https://docs.aws.amazon.com/AmazonCloudFront/latest/DeveloperGuide/IntroductionUseCases.html

Verizon Edgecast:
https://www.edgecast.com/blog/edge-computing-in-a-sea-of-multiple-edges/
https://www.edgecast.com/delivery/control/

[kxu1985]

Akamai EdgeKV Key Takeaways:

• Use eventual consistency model
• Build for fast, frequent reads, infrequent writes
• Concern and warning on compliances
• Key use cases: fast content delivery, localization and personalization, microservices

CloudFlare Workers KV Key Takeaways:

• Use eventual consistency model, up to 60 seconds for global propagation
• Support frequent and fast reads, infrequent writes
• Use cases: lightweight web applications
• All values are encrypted at rest with 256-bit AES-GCM

Fastly Edge Dictionaries Key Takeaways:

• A global key-value store called Edge Dictionaries
• Use 'Varnish Cache' underlyingly
• Dictionary container is tied to versions, but data is 'versionless'
• Use cases: lightweight web applications
• Private and sensitive data cannot be handled

Common questions:

• KV store is in memory or on disks?

Thoughts:

• The common focus is fast reads. Is this feature sufficient for near-future use cases?
• Security compliance is a common concern. None can handle private and sensitive data. Is this a big drawback?
• Consistency model??? Strong consistency vs. eventual consistency vs. other?
• Global definition? Multi-cluster sync and propagation?