I am currently working on a light php framework to use with some high request ajax for my site, and have run into an interesting problem that has me completely stumped. The ajax is for a series of notifications, so the javascript sends off an ajax request for new information every 30 seconds. This ajax is active on every page of the entire site, so I realize its a lot of requests as several hundred users are browsing the site at any given moment, many with several windows open.
Syntax wise everything is perfect. The problem is, when I activated the ajax for my community, there is a build up of 30 - 40 SLEEP commands in the MySQL database. All of which seem to ignore the set timeout of 10 seconds. It effects the entire site's performance as a result.
My understanding is that left over sleep commands are the result of a connection that hasn't been closed. I am use the MySQLi object oriented class and have extended it with a few functions of my own. Everything is by the text book and double checked against the documentation, and I've made sure that every little piece of the MySQLi class is closed and released.
If anyone has any wisdom or experience with this type of effect, I would be most grateful for any advice.
Have you tried a push approach versus your current pull/polling approach? Take a look at Comet:
Comet a new approach to AJAX applications
Comet (wikipedia)
Comet programming: Using AJAX to simulate server push
Ajax Push Engine
However, there are some downsides, the main one being you have to maintain an open connection, and some browsers limit the number of connections you can maintain. The following article talks about the pros and cons of using Comet and suggests a hybrid-polling approach:
The allure of Comet
Here is a stackoverflow question that talks about comet also:
Compatibility of Comet with current technology
Related
I'm running an enterprise level PHP application. It's a browser game with thousands of users online on an infrastructure that my boss refuses to upgrade and the machinery is running on 2-3 system load (yep linux) at all times. Anyhow that's not the real issue. The real issue is that some users wait until the server gets loaded (prime time) and they bring their mouse clickers and they click the same submit button like 10 - 20 times, sending 10-20 requests at the same time while the server is still producing the initial request, thus not updated the cache and the database.
Currently I have an output variable on each request, which is valid for 2 minutes and I have "mutex" lock which is basically a flag inside memcache which if found blocks the execution of the script further, but the mouse clicker makes so many requests at the same time that they run almost simultaneously which is a big issue for me.
How are you, the majority of StackOverflow folks dealing with this issue. I was thinking of flagging the cookie/session but I think I will get in the same issue if the server gets overloaded. Optimization is impossible, the source is 7 years old and is quite optimized, with no queries on most pages (running off of cache) and only querying the database on certain user input, like the one I'm trying to prevent.
Yep it's procedural code with no real objects. Machines run PHP 5 but the code itself is more of a PHP 4. I know, I know it's old and stuff but we can't spare the resource of rewriting this whole mess since most of the original developers left that know how stuff is intertwined and yeah, I'm basically patching old holes. But as far as I know this is a general issue on loaded PHP websites.
P.S: Disabling the button with javascript on submit is not an option. The real cheaters are advanced users. One of them had written a bot clicker and packed it as a Google Chrome extension. Don't ask how I dealt with that.
I would look for a solution outside your code.
Don't know which server you use but apache has some modules like mod_evasive for example.
You can also limit connections per second from an IP in your firewall
I'm getting the feeling this is touching more on how to update a legacy code base than anything else. While implementing some type of concurrency would be nice, the old code base is your real problem.
I highly recommend this video which discusses Technical Debt.
Watch it, then if you haven't already, explain to your boss in business terms what technical debt is. He will likely understand this. Explain that because the code hasn't been managed well (debt paid down) there is a very high level of technical debt. Suggest to him/her how to address this by using small incremental iterations to improve things.
limiting the IP connections will only make your players angry.
I fixed and rewrote a lot of stuff in some famous opensource game clones with old style code:
well, i must say that cheating can be always avoid executing the right queries and logic.
for example look at here http://www.xgproyect.net/2-9-x-fixes/9407-2-9-9-cheat-buildings-page.html
Anyway, about performace, keep in mind that code inside sessions will block all others thread untill current one is closed. So be carefull to inglobe all your code inside sessions.Also, sessions should never contain heavy data.
About scripts: in my games i have a php module that automatically rewrite links adding an random id saved in database, a sort of CSRFprotection. Human user will click on the changed link, so they will not see the changes but scripts will try to ask for the old link and after some try there are banned!
others scripts use the DOM , so its easy to avoid them inserting some useless DIV around the page.
edit: you can boost your app with https://github.com/facebook/hiphop-php/wiki
I don't know if there's an implementation already out there, but I'm looking into writing a cache server which has responsibility for populating itself on cache misses. That approach could work well in this scenario.
Basically you need a mechanism to mark a cache slot as pending on a miss; a read of a pending value should cause the client to sleep a small but random amount of time and retry; population of pending data in a traditional model would be done by the client encountering a miss instead of pending.
In this context, the script is the client, not the browser.
I am a programmer at an internet marketing company that primaraly makes tools. These tools have certian requirements:
They run in a browser and must work in all of them.
The user either uploads something (.csv) to process or they provide a URL and API calls are made to retrieve information about it.
They are moving around THOUSANDS of lines of data (think large databases). These tools literally run for hours, usually over night.
The user must be able to watch live as their information is processed and is presented to them.
Currently we are writing in PHP, MySQL and Ajax.
My question is how do I process LARGE quantities of data and provide a user experience as the tool is running. Currently I use a custom queue system that sends ajax calls and inserts rows into tables or data into divs.
This method is a huge pain in the ass and couldnt possibly be the correct method. Should I be using a templating system or is there a better way to refresh chunks of the page with A LOT of data. And I really mean a lot of data because we come close to maxing out PHP memory and is something we are always on the look for.
Also I would love to make it so these tools could run on the server by themselves. I mean upload a .csv and close the browser window and then have an email sent to the user when the tool is done.
Does anyone have any methods (programming standards) for me that are better than using .ajax calls? Thank you.
I wanted to update with some notes incase anyone has the same question. I am looking into the following to see which is the best solution:
SlickGrid / DataTables
GearMan
Web Socket
Ratchet
Node.js
These are in no particular order and the one I choose will be based on what works for my issue and what can be used by the rest of my department. I will update when I pick the golden framework.
First of all, you cannot handle big data via Ajax. To make users able to watch the processes live you can do this using web sockets. As you are experienced in PHP, I can suggest you Ratchet which is quite new.
On the other hand, to make calculations and store big data I would use NoSQL instead of MySQL
Since you're kind of pinched for time already, migrating to Node.js may not be time sensitive. It'll also help with the question of notifying users of when the results are ready as it can do browser notification push without polling. As it makes use of Javascript you might find some of your client-side code is reusable.
I think you can run what you need in the background with some kind of Queue manager. I use something similar with CakePHP and it lets me run time intensive processes in the background asynchronously, so the browser does not need to be open.
Another plus side for this is that it's scalable, as it's easy to increase the number of queue workers running.
Basically with PHP, you just need a cron job that runs every once in a while that starts a worker that checks a Queue database for pending tasks. If none are found it keeps running in a loop until one shows up.
I recently inherited a PHP application using the CodeIgniter framework - which handles authentication, user sessions, CRUD operations, routing, templating, and pretty much all aspects of the application just beautifully. However, one feature requires the use of long-polling. Certain admins need near real-time updates on what other users are doing. Everything is working fine now for a few hundred users, but we're scaling up to support a few thousand, and I'm worried the long-polling will cause some performance issues.
Here's the basics of the long-poll process:
The browser makes a GET request which kicks off the long-poll process.
The long-poll process checks the timestamp of a txt file every 1/4 seconds.
a) If the txt file is updated, the long-poll process returns the changes to the browser and the view is updated.
b) If no changes are found in 25 seconds, the long-poll ends and returns null.
Repeat step 1.
This process takes place outside of the codeigniter framework. I'm thinking that it's a good idea to replace this process with a socket.io implementation.
Will socket.io be a better solution than the long-polling? If so, what convincing evidence do I have without actually building a demo and doing load testing? It seems like a good idea in my head, but I need to justify the time and effort to make the switch.
Also, sorry if this doesn't make sense, or 'is not a real question' by SO standards. I'm pretty new to back-end scalability and most of this stuff is brand new to me. Please offer helpful guidance for rewording (if needed) before downvoting. Thanks.
EDIT: The preference here is to keep things as-is, because it's expensive to rip out code and replace it with new things, especially when new things are untested. So I guess my question is, when/if the long-polling solution hits a brick wall, will socket.io be a viable replacement?
I think websockets are better than long-polling, with polling you have unnecessary network throughput overhead primarily as your application scale. Check this two links:
http://www.websocket.org/quantum.html
http://dsheiko.com/weblog/websockets-vs-sse-vs-long-polling
I am in the middle of making a social network, and i want it to be as smooth as facebook.
Like if you look in a console and look at logging, it doesn't update all the time with ajax calls.
on my site i have to load: notifications(the number of new notifs and the notifs themselves), friend requests(same as notifications), online friends(if there are any online it will load the pictures of the online users.) thats 6 ajax calls that is loaded every 10 second. and this causes a huge bandwidth waste and server requests.
Therefore i thought, what if the SERVER told the CLIENT when there was a new update instead of the CLIENT asking the SERVER every 10 seconds.
i have googled this problem and read about ajax push, and a framework called comet.
i just can't seem to find any info on how to implement this on jQuery.
I looked briefly into Comet. It appears to be ambitious, experimental and won't run on just any old server.
As I understand it, Comet doesn't really push as such but does something called "long polling", which I won't try to describe here. The web already has several good texts on the subject.
Personally, I would stick with the current plan (conventional AJAX) but make one general purpose call with all the necessary data bundled into an object and JSON encoded. This will reduce 6 requests down to one (every 10 seconds).
You can box-clever by returning nulls within the returned object for information that hasn't changed thereby minimising the length of each response.
As far as I know, you must make significant modifications on your webserver to get this thing to work. Also, server side php is not really a good option.
Somebody had already asked something similar here: Using comet with PHP?
You can try socket.io on node.js too. It works great for real time communication
http://socket.io/
What way will be best to write online chat with js? If i would use AJAX and update information about users and messages every 5sec - HTTP requests and answers will make big traffic and requests will make high server load.
But how another? Sockets? But how..
You seem to have a problem with the server load, so I'll compare the relevant technologies.
Ajax polling:
This is the most straightforward. You do setTimeout loop every 5 seconds or so often to check for new chat messages or you set an iframe to reload. When you post a message, you also return new messages, and things shouldn't get out of order. The biggest drawback with this method is that you're unlikely to poll with a frequency corresponding to how often messages are posted. Either you'll poll too quickly, and you'll make a lot of extra requests, or you'll poll too slowly and you'll get chunks of messages at a time instead of getting them in a real-time-ish way. This is by far the easiest method though.
HTTP Push
This is the idea that the server should tell the client when there are new messages, rather than the client continually bothering the server asking if there are any new ones yet. Imagine the parent driving and kid asking "are we there yet?", you can just have the parent tell the kid when they get there.
There are a couple ways to both fake this and do it for real. WebSockets, which you mentioned, are actually creating a stream between the client and the server and sending data in real time. This is awesome, and for the 4 out of 10 users that have a browser that can do it, they'll be pretty psyched. Everyone else will have a broken page. Sorry. Maybe in a couple years.
You can also fake push tech with things like long-polling. The idea is that you ask the server if there are any new messages, and the server doesn't answer until a new message has appeared or some preset limit (30 seconds or so) has been reached. This keeps the number of requests to a minimum, while using known web technologies, so most browsers will work with it. You'll have a high connection concurrency, but they really aren't doing anything, so it should have too high a server cost.
I've used all of these before, but ended up going with long polling myself. You can find out more about how to actually do it here: How do I implement basic "Long Polling"?
You should chose sockets rather than AJAX Polling, However there isn't much around about how you can integrate socket based chats with MySQL.
I have done a few tests and have a basic example working here: https://github.com/andrefigueira/PHP-MySQL-Sockets-Chat
It makes use of Ratchet (http://socketo.me/) for the creation of the chat server in PHP.
And you can send chat messages to the DB by sending the server JSON with the information of who is chatting, (if of course you have user sessions)
There are a few ways that give the messages to the client immediately:
HTML5 Websockets
good because you can use them like real sockets
bad because only a few browsers support it
Endlessly-loading frame
good because every browser supports it
not so cool because you have to do AJAX requests for sending stuff
you can send commands to the client by embedding <script> tags in the content
the script gets executed immediately, too!
...
So, in conclusion, I would choose the second way.
The typical approach is to use long polling. Though better not do this in PHP (PHP will need one process for each connection thus drastically limiting the number of possible visitors to your site). Instead use node.js. It's perfect for chats.