I‘m rather new to distributed databases, though I have already studied related literature (e.g. CAP theorem, CRDT) and implemented some POC to allow scaling my application horizontally.
Now I however face a challenging problem. In ordere to scale the app horizontally, communication between services is done via a distributed queue. As a background here, I do require a custom CRDT method to keep the data eventually consistent, and I do require my application to work like a cache (remotely related to REDIS).
The challenge is now that I also need to persist the data. That requires me to keep the data within the application cache and database eventually consistent. I‘ve checked Cassandra, I saw a ticket [1] where somebody tried to add functionality for custom CRDT merge functionality (which as I mentioned do require for a reason). That never made it into Cassandra, and seems to have a few issues to resolve.
What are my options, either in form of a concrete distributed database engine allowing custom merging, or an algorithm that could help solve the problem (e.g. in form of a db trigger or something like this).
[1] https://issues.apache.org/jira/browse/CASSANDRA-6412
As far as I know, there are very few databases that allow you to specify your own custom conflict resolution algorithms. Tbh. the only one I really found - disclaimer: I'm not a Microsoft Advocate - is Azure CosmosDB. It has MongoDB-compatible API and can be configured to use master-master replication strategy, where you need to specify your own conflict resolution algorithm (using JavaScript). You can use it to define your own merge operation.
If you'll take a look outside of database-native solutions into application-level ones, there are several tools, like ie. Akka (available in both JVM or .NET version) which enables you to write custom CRDTs inside of distributed-data module. JVM version additionally supports multi-datacenter persistence, which is conceptually closer to how commutative CRDTs work and can be integrated with Cassandra backend.
I've implemented a MerkleClock CRDT at my merkle-crdt repository.
You could use an approach that when you update the database record column, you fetch the column's value and then you merge it with your CRDT of your current state and then when you save, you serialise the CRDT as JSON and store it in the database.
I'm currently working on a multiplayer game which will be using two databases(MONGODB). One for authentication(login) and one to contain all game-specific data.
What I've done is to separate the user and game specific data. This way i'll be able to build micro services around the user in the future.
I'm a bit uncertain on how to handle/validate the game-specific database operations tho.
When i log into my game, i perform a POST request to my rest api, which validates the user and returns some data.
The game itself however, is using a TCP socket connection to handle real-time gameplay and will be saving game-specific data to the database on the authoritative server(all game logic is done on the server) . How would you go about to link the data on the game-specific database to a specific user found in the authentication database?
Theoretically, the best option is to not share data, only unique identifiers (primary keys).
In practice, you should not divide the data this way. Part of the user belongs to the game, other parts might be shared across different games. I'm assuming that's why you separated the two.
Have a look at the DDD principle of Bounded Contexts and how you should/could create separate services this way. That being said, defining bounded contexts the proper way is the hardest thing to do in SOA and/or microservices.
Context: Eclipse 4 platform RCP presentation layer (on top of a SOA backend)
1) ¿Where is the best place to store authenticated user information (HttpSession-like store, including security sensitive data)?
2) ¿Is there any advice or best practice on how to implement SWT widget/control enablement/visibility based on user permissions/profiles? (swt presentation logic being in its own bundle, agnostic of any authorization logic).
I read that in version 3.x this could be done with org.eclipse.ui.activities, there is something like that in E4?
Thanks
In Eclipse 4 we store most things in the model. For 'global' info like sign-ins / session data would be in the MApplication's 'persistentData' field, likely with one entry per userId.
* but * calling this secure is a flat lie; the model is an open book so anybody could find the info. Encryption is up to you (as is how truly secure the result is, Eclipse makes no effort to be secure internally...
As far as modifying the UI is concerned then if it's just a matter of tweaking what a person can see it's not too bad. You can find all the elements in the model and individually set their 'toBeRendered' state to false if they shouldn't be available; true if they should. This will likely work for many things but you may find that you'll have to tweak up your command's 'visibleWhen' clauses to ensure that the system doesn't make them show up again.
I'm planning my first iPhone app and I'd like to get some inputs as to how to build it, right from the start. The iPhone app is being built to be paired with a public facing web application that is already built in PHP.
I'd like the web platform to be central (data is housed in a mySQL database), and have the
iPhone clients talk to it and use REST'ful methods to perform the functions of the site
(fetching latest content, posting content, voting, account management as examples).
I'd like the clients to get a local copy of the data in a SQLite database, but refresh to get the latest version of the feed (similar to the Twitter app).
Couple of thoughts I have right now:
Use something like ASIHTTPRequest to send/recieve data to PHP files on the server listening for requests
JSON - would I be better off to send the GET/POSTS to a PHP that returns JSON objects, and work with some sort of wrapper that manages the data and communicates changes to the local SQLite database?
Am I totally off in how I should be building this thing to communicate with the web? Is
there a best practice for this?
I'd really appreciate any input on how you would architect this sort of a setup.
Thank you,
EDIT: After reading my own post again, I know it sounds like a Twitter client, but it is NOT, although it has similar features/structure of a Twitter type setup. Thanks!
As you already outlined in your plan, XML and REST are a great way to communicate with a web application. I want to suggest few details about how to actually design and build it, or what you should keep in mind.
First of all, I believe it's important to stick with MVC. I've seen people creating HTTP connections in view-controllers, controllers being NSXMLParser's delegate, controllers containing data in member variables. I've even seen UITableCells establishing HTTP connections. Don't do it!
Your model and its basic manipulation code should be as much extracted from user interface as possible. As you already have created the model in your web-application, try to recreate the entities in your iPhone project. Don't be afraid of having some simple methods in entity classes, but do not make them use external resources, especially tcp connections. As an example of methods in entity class you might have methods that formats data in specific ways (dates as an example, or returning fullname as concatenation of firstname and surname), or you can even have a method like - (void)update that would act as a wrapper to call class responsible to update the model.
Create another class for updating the model - fetching the XMLs from web-app. Do not even consider using synchronous connections, not even from a dedicated thread. Asynchronous connections with delegate is the way to go. Sometimes multiple requests need to be made to get all required data. You might want to create some kind of state-machine to keep the information about in which stage of downloading you are, and progress from stage to stage, skipping to the end if error occurs, re-executing from failed stage after some moments.
Download data somewhere temporarily, and first when you have it all, make a switch and update user interface. This helps responsiveness during launching the app - user gets to work immediately with data stored locally, while the update mechanism is downloading the new data.
If you need to download lots of files, try to download them simultaneously, if dependencies between files allow for that. This involves creating a connection per request, probably delegate instance for each of them. You can of course have only one delegate instance for all of those connections, but it gets a bit more complex to track the data. Downloading simultaneously might decrease latency considerably, making the mechanism much faster for the user.
To save the time and bandwidth, consider using HTTP's If-Modified-Since and/or ETag headers. Remember the time or tag when you requested the data the last time, and next time send it in HTTP's header. Your web-application should return HTTP code 304 if content has not been changed. iPhone app should react on this code accordingly in connection:didReceiveResponse:.
Create a dedicated class to parse the XML and update the model. You can use NSXMLParser, but if your files are not huge I strongly recommend TouchXML, it's such a pleasure to work with XML as document (it also supports XPath), instead of an event based API. You can use this parser also when files are downloaded to check their validity - re-download if parsing fails. That's when dedicated class for parsing comes handy.
If your dataset is not huge, if you do not need to persist downloaded data on iPhone forever, you probably don't need to store them in SQLite database, you can simply store them in XML format - just a simple caching. That at least might be the way for a twitter app. It gets easier that way, but for bigger data sets XML consumes lots of memory and processing power - in that case SQLite is better.
I'd suggest using Core Data, but you mention this is your first iPhone app, so I suggest you don't use it. Yet.
Do not forget about multitasking - your app can go to sleep in the middle of download, you need to cancel connections, and cleanup your update mechanisms. On app's wake-up you might want to resume the update.
Regarding the view part of the application - use Interface Builder. It might be painful in the beginning, but it pays off in the long run.
View controllers are the glue between model and views. Do not store data in there. Think twice about what to implement where, and who should call it.
This is not related to architecture of the app, but I want to remind that Objective-C is very expressive language. Code should read much like a sentence. Extend classes with protocols. As an example the other day I needed first line of a string. Sure, you can write a one-liner where you find first occurrence of a new-line, and get a substring from beginning till there. But it doesn't look right. I've added - (NSString*)firstLine into my NSString's protocol. Code looks so much better this way, it doesn't need any comments.
There are lots of things to consider in both architecture and design of any project, they both should go hand in hand. If one is causing trouble to the other, you need to adapt. Nothing is written in stone.
I'm currently working on an app that sounds similar to yours. I'd also suggest ASIHTTPRequest, and probably something like TouchJSON for JSON parsing, or extending/making a delegate of NSXMLParser if you want to parse XML.
As suggested by JosephH, depending on how your app works you may want to consider alternate authentication methods: I'd take a look at something token-based like OAuth, which has ready-made libraries for people to dig in to.
SQLite is totally viable for feed caching, although I prefer NSCoding so that you can freeze-dry your custom data structures.
As a general suggestion, make sure to spend a lot of time thinking about every use case and corner case for connections: it's easy to assume a user will only contact the server in certain ways and at certain times, and then after you throw in multitasking/incoming calls/lock screen/memory warnings, things can get hairy without any planning.
All in all, you seem to be on the right track, just make sure you plan out everything beforehand :)
Apple have a brand new in depth piece of sample code - MVCNetworking that shows in depth how to use subclasses of NSHTTPRequests and NSOperationQueues.
As others mentioned, I think you are asking the right questions and are heading in the right direction. All of the replies above are valuable advice. Here is my advice, and I hope you'll find it useful.
No matter which method/library you choose to talk to your web services, I think it's important to make a clean separation in the way you design your data model on the phone VS. the data model in your web application. You have 3 major distinctions to keep in mind for your design:
Data model on the web application (reflected by your existing mySQL database)
Since this is already there, there is not much to say about it, except that it will influence a lot your design for the following 2 parts. I suggest to make this model the 'master reference' for how your data is represented across platforms.
Data model on the iPhone app (reflected by the information you need to display in the iPhone app)
This is where the fun begins. First, you need a good understanding of what data you need to display in the phone app. So have a good, high level design of your app first (use pen and paper, draw mock-ups of each view and the interactions between them, model the navigation between your view controllers etc.). It really helps to understand the interactions between your view controllers and the various bits and pieces of data you want to show in the app. This will help you create the requirements for the data model on the phone. Based on these requirements, map the existing (web) data model to a new model, suited to your iPhone app. This new model may or may not include all tables and fields found in your web app. But the general representation of the 2 models should be very similar (e.g. relationships, data types, etc.)
Data model used to communicate between the 2 above (this is your 'data exchange protocol')
Once you have the 2 representations of your data above, you need to 'translate' from one to the other, both ways. Design your data exchange protocol to be as simple and compact as possible. You don't want to waste bytes on useless information, as transmissions over the network are costly. (As a side note, you might think of compressing the transmitted data later on, but it's just as important to have a good design from the beginning). It's probably best to begin with a protocol in which the metadata is the same as the one in your web application model (e.g. same relationships, names of tables, attributes, etc.). But remember, you'll only have to serialize/de-serialize those entities and relationships that you listed in point 2) above. So design accordingly. Your exchange protocol may also include session tokens, authentication info, a version number, or other metadata, if you need it.
Remember: your data exchange protocol is what will de-couple your web application and iPhone application models. I found that it's best to de-couple them because they may both evolve over time. The data model on the iPhone for example, may evolve a lot especially when you will find that you need to re-model some relationships or add/remove attributes from your entities in order to improve application responsiveness, or the user experience, the navigation, or whatever.
Since this is a whole concern in and by itself, well, you need to design a generic serialization/de-serialization mechanism on top of your (JSON/XML/whatever parser you choose) that is flexible enough to sustain the potential differences between your 2 data models. These differences might be: entity/attribute/relationship names, primary key identifier names, data types, attributes to ignore, and the list goes on. I would definitely implement a serializer/de-serializer utility class in the iPhone app, backed by a .plist configuration file containing all supported entities, concerns, aliases you might have. Of course, each model object should 'know' how to serialize, de-serialize itself and its relationships (i.e. the required object graph depth).
One last note, since you will end up with 2 representations of your data, you will need a way to uniquely identify an object on both sides. So for example, think of adding a uuid attribute to all data that needs to be exchanged, or use any other approach that suits your needs.
I am building an app that has similar requirements to yours, and these are the approaches I found to be best so far. Also, you might find this video useful (it inspired me a lot on how to implement some of the issues I mentioned above and is especially interesting if you're using CoreData) :
http://itunes.apple.com/ca/podcast/linkedin-important-life-lessons/id384233225?i=85092597
(see the lecture entitled "LinkedIn: Important Life Lessons on CoreData & GameKit (March 12, 2010)" )
Good luck!
It's quite a broad question, and I think you're going in the right way anyway, however I'll do my best to give some advice:
JSON, ASIHTTPRequest and POSTs to PHP scripts sound like a great way to go.
If the data is not really sensitive, I'd use http most of the time, and use https only for a login page that either sets a cookie or returns a "token" that you use in subsequent requests. (HTTPS can be quite slow over a 3G connection as the overhead in terms of number of packets to setup an SSL connection is higher than a plain TCP connection.)
You should make sure you correctly pass any data from the input to the PHP scripts to the database, to avoid any SQL injection attacks - ie. used parameterised SQL, don't create sql queries by doing "SELECT * from users where username="+$_GET['username']"
I would do this like I have done with a lot of AJAX web-page stuff. i.e.:
Have a URL on your server side package the information to be transmitted into XML format. (This can be through a CGI/PHP script or whatever). Your transmitting XML in the message body - so it's easy to human read and debug with a standard web browser.
Use the standard iPhone NSXMLParser methods to parse out the individual data fields from the XML doc, and write it back to your database. This method is equiped to both fetch the data from a URL and parse it in one call - like:
NSURL *xmlURL = [NSURL URLWithString:#"http://www.example.com/livefeed.cgi"];
NSXMLParser *myParser = [[NSXMLParser alloc] initWithContentsOfURL:xmlURL];
Walk through the data hierarchy with the NSXMLParser methods and populate your database accordingly.
I'd know if it exists a kind of CoreData server ?
The point is to get an "automatic API server" providing data to clients.
It can be useful to implement very quickly standalone Forum app, Games or anything on the cloud...
There isn't a Core Data server.
Strictly speaking Core Data isn't a data base system but rather an object-graph management system with persistence tools added on. It's primary purpose is to handle the complexity of object relationships internal to a single app. You don't even have to save the graph to disk if you don't want to.
NextStep/Apple used to have a nifty technology called Enterprise Objects which was a kind of object oriented wrapper for non-OOP databases. It did allow you to setup or interface with any database no matter how heavy weight. Unfortunately, they have stopped selling it and just use it internally now. It was rather nifty.