By: Dr. Maria Seraphina Astriani<\/p>\n
<\/p>\n
Do you want to save money when your server is idle?<\/p>\n
Try SERVERLESS!<\/p>\n
\u00a0<\/strong><\/p><\/blockquote>\n
What is serverless computing?<\/strong><\/p>\n
Serverless computing enables developers to build applications faster by eliminating the need for them to manage infrastructure. With serverless applications, the cloud service provider automatically provisions, scales, and manages the infrastructure required to run the code [1].<\/p>\n
Serverless computing is an execution model for the cloud in which a cloud provider dynamically allocates\u2014and then charges the user for\u2014only the compute resources and storage needed to execute a particular piece of code<\/strong> or \u201cpay as you go\u201d [2]. If you do not use serverless, you still must pay for the unused resources.<\/p>\n
<\/p>\n
Resources and Demand<\/strong><\/p>\n
Source: Salma REBAI, Allocation et federation des ressources informatiques dans le Cloud<\/em> (Resource allocation in Cloud federation)<\/p>\n
<\/p>\n
What is Function-as-a-Service?<\/strong><\/p>\n
Function-as-a-Service (FaaS) is a serverless way to execute modular pieces of code on the edge. FaaS lets developers write and update a piece of code on the fly, which can then be executed in response to an event, such as a user clicking on an element in a web application. This makes it easy to scale code and is a cost-efficient way to implement microservices [3].<\/p>\n
<\/p>\n
Using serverless code like FaaS, web developers can focus on writing application code, while the serverless provider takes care of server allocation and backend services.<\/p>\n
<\/p>\n
Should I use serverless?<\/strong><\/p>\n
These are the four main reasons people switch to serverless [4]:<\/p>\n
\n
- it scales with demand automatically<\/li>\n
- it significantly reduces server cost (70-90%), because you don\u2019t pay for idle<\/li>\n
- it eliminates server maintenance<\/li>\n
- it frees up developer resources to take on projects that directly drive business value (versus spending that time on maintenance)<\/li>\n<\/ul>\n
<\/p>\n
Reasons not to use serverless [5]:<\/p>\n
\n
- your workloads are constant<\/li>\n
- you fear vendor lock-In<\/li>\n
- you need advanced monitoring<\/li>\n
- you have long-running functions<\/li>\n
- you use an unsupported language<\/li>\n<\/ul>\n
<\/p>\n
Providers<\/strong><\/p>\n
There are many providers that already support serverless and these are some of them:<\/p>\n
\n
- Lambda (Amazon Web Services)<\/li>\n
- Azure Functions (Microsoft)<\/li>\n
- Google Cloud Functions (Google)<\/li>\n
- IBM Cloud Functions (IBM)<\/li>\n<\/ul>\n
<\/p>\n
Please have a look on the following table to know which one is the most suitable for you [6]:<\/p>\n
\n\n
\n \n Features<\/span><\/b><\/p>\n<\/td>\n
\n Amazon Web Services (AWS)<\/span><\/b><\/p>\n<\/td>\n
\n Microsoft<\/span><\/b><\/p>\n<\/td>\n
\n Google<\/span><\/b><\/p>\n<\/td>\n
\n IBM<\/span><\/b><\/p>\n<\/td>\n<\/tr>\n
\n \n Serverless Compute Offerings<\/span><\/i><\/p>\n<\/td>\n
\n Lambda<\/span>
\nLambda@Edge<\/span><\/span><\/p>\n<\/td>\n\n Azure Functions<\/span>
\nAzure Event Grid<\/span><\/span><\/p>\n<\/td>\n\n Google Cloud Functions<\/span>
\nGoogle App Engine<\/span><\/span><\/p>\n<\/td>\n\n IBM Cloud Functions<\/span><\/p>\n<\/td>\n<\/tr>\n
\n \n Maximum Functions<\/span><\/i><\/p>\n<\/td>\n
\n Lambda<\/span>\u00a0\u2013 Unlimited
\nLambda@Edge<\/span>\u00a0\u2013
\n25 per AWS account<\/span><\/p>\n<\/td>\n\n Unlimited<\/span><\/p>\n<\/td>\n
\n 1000 per
\nproject<\/span><\/p>\n<\/td>\n\n Unlimited<\/span><\/p>\n<\/td>\n<\/tr>\n
\n \n Scalability & Availability<\/span><\/i><\/p>\n<\/td>\n
\n Transparent \u2013
\nAutomatic Scaling<\/span><\/p>\n<\/td>\n\n Automatic
\nscaling<\/span><\/p>\n<\/td>\n\n Automatic<\/span><\/p>\n<\/td>\n
\n Automatic<\/span><\/p>\n<\/td>\n<\/tr>\n
\n \n Languages supported<\/span><\/i><\/p>\n<\/td>\n
\n \u00a7\u00a0 <\/span><\/span>Lambda<\/span>\u00a0\u2013 Node.js, Python, Java, C# (.NET)<\/span><\/p>\n
\u00a7\u00a0 <\/span><\/span>Lambda@Edge<\/span>\u00a0\u2013 Node.js 6.10<\/span><\/p>\n<\/td>\n
\n Node.js, C#,
\nPython, F#, PHP,
\nBash, Batch, executable for Azure Functions<\/span><\/p>\n<\/td>\n\n Node.js<\/span><\/p>\n<\/td>\n
\n Binaries in
\nDocker, Python, Swift, Node.js, Java<\/span><\/p>\n<\/td>\n<\/tr>\n\n \n Max. execution time<\/span><\/i><\/p>\n<\/td>\n
\n 300 seconds
\n(5 minutes) in Lambda<\/span><\/p>\n<\/td>\n\n 300 seconds<\/span><\/p>\n<\/td>\n
\n 540 seconds<\/span><\/p>\n<\/td>\n
\n 600 seconds<\/span><\/p>\n<\/td>\n<\/tr>\n
\n \n Max Code Size<\/span><\/i><\/p>\n<\/td>\n
\n 50 MB compressed
\n250 MB uncompressed<\/span><\/p>\n<\/td>\n\n None
\nUser pays the storage cost<\/span><\/p>\n<\/td>\n\n 100 MB
\ncompressed
\n500 MB uncompressed<\/span><\/p>\n<\/td>\n\n 48 MB<\/span><\/p>\n<\/td>\n<\/tr>\n
\n \n Monitoring<\/span><\/i><\/p>\n<\/td>\n
\n Dashbird,
\nCloudWatch<\/span><\/p>\n<\/td>\n\n Azure
\nApplication Insights<\/span><\/p>\n<\/td>\n\n Stackdriver<\/span><\/p>\n<\/td>\n
\n IBM Cloud
\nShell<\/span><\/p>\n<\/td>\n<\/tr>\n\n \n Concurrent Executions<\/span><\/i><\/p>\n<\/td>\n
\n 1000
\nconcurrent executions per account per region<\/span><\/p>\n<\/td>\n\n 10 concurrent
\nexecutions per function<\/span><\/p>\n<\/td>\n\n 400 per
\nfunction<\/span><\/p>\n<\/td>\n\n 1000 per
\nproject<\/span><\/p>\n<\/td>\n<\/tr>\n\n \n Triggers<\/span><\/i><\/p>\n<\/td>\n
\n \u00a7\u00a0 <\/span><\/span>Lambda<\/span>\u00a0\u2013 S3, Kinesis, DynamoDB, Simple Notification,
\nCognito, Simple Email Service, CloudWatch Logs, CloudFormation,
\nCodeSummit,CloudWatch Events, Echo, Scheduled, Echo, API Gateway, Lex<\/span><\/p>\n\u00a7\u00a0 <\/span><\/span>Lambda@Edge<\/span>\u00a0\u2013 Functions will automatically trigger against
\nfollowing Amazon CloudFront events \u2013 viewer requests, viewer response, origin
\nrequest, origin response<\/span><\/p>\n<\/td>\n\n Queues,
\nSchedule, HTTP, Events, Blob Storage<\/span><\/p>\n<\/td>\n\n Cloud
\nStorage, HTTP, Cloud Pub\/Sub<\/span><\/p>\n<\/td>\n\n Periodic
\nTriggers, webhook triggers for GitHub, Cloudant noSQL DB, Push Notification
\nService, Message Hub Service, Watson, Slack APIs, websocket, weather<\/span><\/p>\n<\/td>\n<\/tr>\n\n \n Web Editing<\/span><\/i><\/p>\n<\/td>\n
\n Web Editing<\/span><\/p>\n<\/td>\n
\n Provided<\/span><\/p>\n<\/td>\n
\n Via Google
\nCloud Source<\/span><\/p>\n<\/td>\n\n Provided<\/span><\/p>\n<\/td>\n<\/tr>\n
\n \n Logs<\/span><\/i><\/p>\n<\/td>\n
\n Via
\nCloudWatch Logs<\/span><\/p>\n<\/td>\n\n Available<\/span><\/p>\n<\/td>\n
\n Via CLI and
\nStackdriver<\/span><\/p>\n<\/td>\n\n Yes<\/span><\/p>\n<\/td>\n<\/tr>\n
\n \n HTTP<\/span><\/i><\/p>\n<\/td>\n
\n Must be
\ntriggered through an API gateway<\/span><\/p>\n<\/td>\n\n Directly
\ntriggered via HTTP<\/span><\/p>\n<\/td>\n\n Can be
\ndirectly triggered via HTTP<\/span><\/p>\n<\/td>\n\n Yes<\/span><\/p>\n<\/td>\n<\/tr>\n
\n \n Orchestration<\/span><\/i><\/p>\n<\/td>\n
\n Via the step
\nfunctions<\/span><\/p>\n<\/td>\n\n Via Logic
\nApps<\/span><\/p>\n<\/td>\n\n No<\/span><\/p>\n<\/td>\n
\n Via Rules<\/span><\/p>\n<\/td>\n<\/tr>\n
\n \n Version Control<\/span><\/i><\/p>\n<\/td>\n
\n Versioning
\nand aliases<\/span><\/p>\n<\/td>\n\n Via GitHub
\nand other similar platforms<\/span><\/p>\n<\/td>\n\n Via Google
\nCloud Source<\/span><\/p>\n<\/td>\n\n No<\/span><\/p>\n<\/td>\n<\/tr>\n
\n \n Deployments<\/span><\/i><\/p>\n<\/td>\n
\n .ZIP to S3 or
\nLambda<\/span><\/p>\n<\/td>\n\n Via GitHub,
\nVisual Studio, Local git, Dropbox, Bitbucket<\/span><\/p>\n<\/td>\n\n Google Cloud
\nSource and .ZIP to cloud storage<\/span><\/p>\n<\/td>\n\n GitHub,
\nBluemix DevOps<\/span><\/p>\n<\/td>\n<\/tr>\n\n \n Access Management<\/span><\/i><\/p>\n<\/td>\n
\n IAM roles
\n(identity and access management)<\/span><\/p>\n<\/td>\n\n IAM roles<\/span><\/p>\n<\/td>\n
\n IAM roles<\/span><\/p>\n<\/td>\n
\n IAM roles<\/span><\/p>\n<\/td>\n<\/tr>\n
\n \n Dependencies<\/span><\/i><\/p>\n<\/td>\n
\n Deployment
\npackages<\/span><\/p>\n<\/td>\n\n NuGet, Npm,
\nVisual Studio Team Services<\/span><\/p>\n<\/td>\n\n Npm
\npackage.json<\/span><\/p>\n<\/td>\n\n NA<\/span><\/p>\n<\/td>\n<\/tr>\n
\n \n Limits<\/span><\/i><\/p>\n<\/td>\n
\n \u00a7\u00a0 <\/span><\/span>Lambda<\/span>\u00a0\u2013 Memory Allocation \u2013 128 MB to 1536 MB<\/span><\/p>\n
Ephemeral disk capacity \u2013
\n512 MB<\/span><\/p>\n\u00a7\u00a0 <\/span><\/span>Lambda@Edge<\/span>\u00a0\u2013 Viewer request and response event limits = 128
\nMB, request timeout = 30 seconds<\/span><\/p>\n<\/td>\n\n Azure
\nFunctions under consumption plan, limits the execution time to 5 minutes<\/span><\/p>\n<\/td>\n\n Function
\ncalls per second: 1,000,000 per 100 seconds<\/span><\/p>\n<\/td>\n\n Memory \u2013 min
\n256 MB and max 512 MB<\/span><\/p>\nTimeout \u2013 60000 ms to 600000 ms<\/span><\/p>\n<\/td>\n<\/tr>\n
\n \n Pricing<\/span><\/i><\/p>\n<\/td>\n
\n \u00a7\u00a0 <\/span><\/span>Lambda<\/span>\u00a0\u2013 1M requests for free, after that, $0.20\/1M
\ninvocations along with $0.00001667\/GB-sec<\/span><\/p>\n\u00a7\u00a0 <\/span><\/span>Lambda@Edge<\/span>\u00a0\u2013 Request Pricing is $0.60 per 1 million requests
\n($0.0000006 per request), Lambda@Edge functions are metered at a granularity
\nof 50ms<\/span><\/p>\n<\/td>\n\n Azure Event
\nGrid \u2013 Price per million operations $0.60<\/span><\/p>\n<\/td>\n\n 2 million
\ninvocations free. $0.40\/million invocations<\/span><\/p>\n$0.0000025\/GB-sec
\nwith 400,00 GB-sec\/month for free<\/span><\/p>\n$0.0000100\/GHz-sec with 200,000 GHz-sec\/month for free<\/span><\/p>\n<\/td>\n
\n Basic Cloud
\nFunction Rate: $0.000017 per second of execution, per GB of memory allocated<\/span><\/p>\nAPI Gateway: Free\/No Limits<\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n
REFERENCES:<\/p>\n
[1] https:\/\/azure.microsoft.com\/en-us\/overview\/serverless-computing\/<\/a><\/p>\n
[2] https:\/\/www.infoworld.com\/article\/3406501\/what-is-serverless-serverless-computing-explained.html<\/a><\/p>\n
[3] https:\/\/www.cloudflare.com\/learning\/serverless\/glossary\/function-as-a-service-faas\/<\/a><\/p>\n
[4] https:\/\/www.serverless.com\/blog\/when-why-not-use-serverless<\/a><\/p>\n
[5] https:\/\/blog.runscope.com\/posts\/why-you-might-not-need-serverless<\/a><\/p>\n