|
SF allows you to automatically spin up and shut down VMs running your microservice instances, based on a schedule or just reacting to periods of heavy use. Coupled with auto-scaling of storage (e.g. CosmosDB) you can meet the demands of peak usage whilst keeping the hosting costs to a minimum.
Adam
|
|
|
|
|
thanks for the use case 
M.D.V. 
If something has a solution... Why do we have to worry about?. If it has no solution... For what reason do we have to worry about?
Help me to understand what I'm saying, and I'll explain it better to you
Rating helpful answers is nice, but saying thanks can be even nicer.
|
|
|
|
|
I've just realised that it reads like a sales pitch from MS... not intentional!
|
|
|
|
|
adambl wrote: not intentional!
Supposed so
M.D.V.
If something has a solution... Why do we have to worry about?. If it has no solution... For what reason do we have to worry about?
Help me to understand what I'm saying, and I'll explain it better to you
Rating helpful answers is nice, but saying thanks can be even nicer.
|
|
|
|
|
Prepare to chew some foot 
One reason to work with microservices is that you can deploy the services independently.
If you have a sales service and a stock service you can implement new features in sales and deploy them as a single unit, no need to build and test stock so deployments should be fast and cheap (and different teams can work on and deploy different services independently).
That also means, unlike traditional monoliths, that some parts of the system can be way busier than others.
Since the stock service is used by the sales, production, and purchasing services it's probably a lot busier than any of those other services.
What Service Fabric (or Kubernetes) can do for you is scale up a couple of extra instances (on different servers if need be) of the stock service if the load is heavy and scale them down when the load decreases.
Service Fabric also takes on the role of a load balancer in that case.
Another benefit is that if you have multiple instances on different servers you can update a server and still have at least one instance of a service so a server update has no downtime for your applications.
And, of course, the same goes for deployments, no downtime whatsoever 
|
|
|
|
|
Forgot to look on CP, there's an interesting article: Azure Service Fabric demo[^]
"Why don't you go where fashion sits, putting on the Ritz" 
|
|
|
|
|
We have been using Service Fabric to host microservices for about a year while now. We created a few Stateless services with minimal use of the SF SDK so most of the time when developing you can ignore the fact they are hosted in SF, and moving away from SF should not be too expensive. You can run the SF host on your dev machine, but because we have kept the SF references to just the 'shell' we can also run the services locally in a console host if you want, which makes dev a little easier.
Our deployments are via Octopus, which does have some built-in support for SF deployments, not great but you can get it running smoothly with a little fiddling. Later versions of Octopus might be better at this now.
What I really like about SF is that zero-downtime deployments are easy and it manages the drain-stopping of load balancer for you. You can package up multiple services into single application package that is deployed as a unit (and rolled back as a unit in the case of failure), which is great, but of course, if you want that you should be thinking of containers anyway.
So, we are looking at moving to containers this year (it is the MS recommended direction of travel too). You can currently host containers in SF but it is a bit awkward and there is some new stuff in the pipeline (Seabreeze) that supports server-less hosting of containers which is where I think we will end up.
Adam
|
|
|
|
|
adambl wrote: and moving away from SF should not be too expensive
Sounds good, especially taking Afzaal's comment into consideration.
I deploy using VSTS, I'm guessing Microsoft has made deployment using their own products seamless (it's literally a button click for App Services).
adambl wrote: we are looking at moving to containers this year (it is the MS recommended direction of travel too)
Are you saying SF will be discontinued in favor of Docker and Kubernetes or that it's better to take that route right away?
|
|
|
|
|
I haven't any knowledge of the MS roadmap for SF, but I guess they would keep support for both SF Apps and containers for the foreseeable future.
For some (i.e. those .Net devs without container experience) the barrier to entry is lower for SF apps, as you can Project->New in VS, then right click and publish your app direct to SF. But as soon as you have multiple services to deploy as a unit, containers are your friend, and worth getting to grips with the new tech.
Obviously if you are using containers outside the MS ecosystem, the fact MS now support them means that the entry barrier for you to start using MS hosting has lowered. Keeping both means that MS keeps both groups of people happy (= paying customers).
|
|
|
|
|
|
Nice ones, thanks
|
|
|
|
|
I started working with Service Fabric before it was generally available. I built (but not yet released) a consumer, cloud-based, mobile-first app on top of service fabric.
I was first drawn to SF by the coolness and speed of retaining application data in memory (stateful reliable services). Once I became comfortable with it as a whole I decided to build a new consumer SaaS app on top of it. It has great scalability and its zero downtime abilities made it a great fit for me.
This has been a one-man project. I'm trying to finish the product and build a company.
Learning curve: There's quite a bit to know and keep in your head. And it crosses several disciplines that are not strictly coding related: networking, security, capacity planning, disaster recovery, etc. Learning SF was drawn out over a long period of time for me because it started out as a side project.
Gotchas: Early on there were some edge cases that weren't handled well (or at all), but that's much better now. SF is still relatively new so sometimes google isn't much help if you have a problem in an unusual use case/situation. It is MUCH better than it used to be though. And SF team members monitor Stack Overflow and github so there is assistance available. The biggest gotcha I have experienced though is the number of external dependencies that exist in my project. This is NOT specific to SF, but SF has contributed to it.
Overall experience: Good. I recommend it.
SF is a big topic. Feel free to contact me privately for more specific discussion or guidance.
Have fun!
Mike
|
|
|
|
|
Cool, sounds good!
What struck me is that SF is not included in MS exams 70-532, 533 and 535 (formerly 534).
I'm probably going to start using it in the next few weeks or so, if I have any questions I'll be sure to contact you, thanks
|
|
|
|
|
sorry for the late reply
I love SF and highly recommend it. I used it for ~ 1 year and half. I will talk about the first version, a new one will soon complement it (without discontinuing the 1st)
You can build highly scalable (planet scale) and reliable apps with just C# , visual studio (integrates perfectly) and an azure subscription. Containers are optional but you can do without.
you need approximately 1 week to get started (the doc is excellent) 2 weeks start tinkering and 1 month to start being productive
SF enforces some interesting paradigms to write better microservices apps . I love their reliable actors paradigm.
The new version will target a "Serverless" approach so it will be quite different and containers here are required. I have not yet studied it but I'm sure the team will make a great work as usual.
There is a great team with nice people, always ready to support the community, they do a monthly meeting, you could join to have a look (but community makes deep tech questions)
Go for it
Antonio
|
|
|
|
|
AReady wrote: you need approximately 1 week to get started (the doc is excellent) 2 weeks start tinkering and 1 month to start being productive
I need it implemented yesterday 
AReady wrote: The new version will target a "Serverless" approach so it will be quite different and containers here are required
Interesting! Do you know what this version is called?
|
|
|
|
|
|
AReady wrote: it's called Service fabric mesh
Cool
|
|
|
|
|
|
|
I'd forgotten about that one ...
Sent from my Amstrad PC 1640
Bad command or file name. Bad, bad command! Sit! Stay! Staaaay...
AntiTwitter: @DalekDave is now a follower!
|
|
|
|
|
|
|
That's a brilliant ironic spam.
[Edit]Removed the phone number from your title as this was doing their advertising for them. PO'H
I'd rather be phishing!
modified 4-Jul-18 8:52am.
|
|
|
|
|
Please remove the phone number from the subject
Bryian Tan
|
|
|
|
|
done (by someone else).
I'd rather be phishing!
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
