Suggestion Needed for Server Architecture for Single Code Base Multiple sub-domain (Multi tenant architecture) MVC .NET.
Introduction
Working on creating MVC .NET Web based application, which will be connected to different DB , based on Sub-domain name.
2k Accounts , will have approx 100 users each.
Finding the best architecture suggestions to handle this.
Description
- Account / Sub domains count will be 2K or more.
- Each Account will contain approx 100 users.
- Database will be MSSQL, each subdomain will have different db connection.
May i request you to help / expert feedback in below questions,
Questions
Type of URL structure which will suite best for above requirement.
Logic 1 Or Logic 2 OR Logic 3, which is the best ?
How server can withstand this many responses ? Do this is related to Application pool in .NET ? Please advice more on this.
2000 Subdomain X 100 Users = 2,00,000 Requests , one code base can handle this many requests ?
Logic 1 Or Logic 2 OR Logic 3, which is best to implement ? What is advantage to one over other?
Google / Facebook etc handling single domain with multiple servers.Is this is the best industry standard ?
As we know , Logic 1 , we can set multiple Application Pool for each Sub domain , will this be applicable for Logic 3 as well ?
Considering Logic 1 and Logic 3 , which is the best and why ? when there will be too much load on web server , we need to add another web server at that time which Logic will be best suitable?
Sharing the points which is known,
We can add different application pool for different sub domains.
APP server should be high end server to manage this many requests.
Need to add load balancer when load / traffic on server increases.
Need a best approach , expensive infrastructure also can be considered.
Url structure is really irrelevant here (as long as it is not "Logic 3" -- just don't expose your "database name" in query strings). Beware, though, that if you go the "subdomain for each customer" way, you will have to purchase a wildcard SSL certificates, and these are a bit more expensive. Plus, IIS still does not support wildcard subdomains, so there will be quirks in setting things up. Other than that, either option is fine.
Load balancers are pretty much a must these days. On the other hand, I would not invest in "high end" hardware. A couple of middle-of-the-road servers will be more than sufficient. What you should be worrying about is getting these 200,000 users.
Regarding database schema, see the Multi-Tenant Data Architecture article on MSDN. Basically, there's no single right answer. Each option comes with its own sets of benefits and downsides.
Single "code base" (as in, single instance of an application) can handle as much databases as you wish. The only limit here is SQL Server itself. Maximum Capacity Specifications for SQL Server says that the maximum number of "Databases per instance of SQL Server" is 32 767, but whether it's practical or not is debatable.
Related
What is the difference between the single and multiple instances sites in asp.net when using Azure cloud services?
Ok - there's a few concepts here that you need to grok to answer your question.
// Arrange.
First, I'll make some assumptions about your question mainly based on the link to documentation, so my answers have less ambiguity.
You're dealing with Azure WebSites and not a Cloud Web Role or a custom Windows Virtual Machine with IIS. **
You're trying to remember stuff with the Session object (ie: State data).
You're not sure what an instance or multiple instances are, with respect to Azure WebSites.
NOTE: My answer applies to WebSites, Web Roles and Windows VM's running IIS .. but I just wanted to be uber clear on the Q.
// Act.
When you create a website (either in an WebSite, Web-Role or a custom Windows Server with IIS) the website has some defined memory boundary/space/garden/wall/magic bubble which is called the App Pool. It means that your website is 100% isolated from other websites on that single server. You do something bad, it doesn't mess with anyone else's sites.
So that website which is installed on that single server is called an instance.
Next, we decide that we need to handle so many people hitting our websites, so we need to scale out. This means, make copies/clones of this website which has the effect of splitting the load up. If you scale out to 3 copies, then each webserver should (for simplicity) split the work load by a 3rd - so each handles about 33% of the load***.
Now, you have 1 website on 3 servers and this is called multiple instances.
So an instance is therefore a term used to describe how many servers the website is installed on.
Ok - so why is this important and what does this have to do with State (as suggested by that article you were reading/referring to) ?
Remember how I said that a instance is a single server and if you have multi instances you have more than 1 server? well .. now that the website exists on different servers - they can't share their State data between them unless you do some special stuff. That special stuff is what that document is chatting about, with lots of funky terms like Inproc, OutProc, Distributed caching, etc.
// Assert.
So the TL;DR; is that you now know that when you scale out and have multiple copies of your website on separate hardware, that's called multi instances and when you do that (have more than one copy) then you need to consider some special code to handle sharing of the State across these multiple servers -- if you need to share state.
Now - have a pic of a beautiful Mola Mola for reading all of this :)
*** Yes yes yes .. there's a number of algorithms to handle load balancing of scaled out sites, like round robin, etc. Lets just keep it really simple for the purpose of this question. K? thxgoodbai.
The "multiple instances" means what it says - more than a single instance of your website hosted on two IIS instances. If you utilise in-process state tracking you will not be able to balance traffic between the two web servers (typically sticky sessions are used to get around this but this is not an option with the load balancing capabilities of Azure).
I am building an ASP.NET web application that will use SQL Server for data storage. I am inheriting an existing structure and I am not able to modify it very much. The people who use this application are individual companies who have paid to use the application. Each company has about 5 or 10 people who will use the application. There are about 1000 companies. The way that the system is currently structured, every company has their own unique database in the SQL Server instance. The structure of each database is the same. I don't think that this is a good database design but there is nothing I can do about it. There are other applications that hit this database and it would be quite an undertaking to rewrite the DB interfaces for all of those apps.
So my question is how to design the architecture for the new web app. There are times of the month where the site will get a lot of traffic. My feeling is that the site will not perform well at these times because I am guessing that when we have 500 people from different companies accessing the site simultaneously that they will each have their own unique database connection because they are accessing different SQL Server databases with different connection strings. SQL Server will not use any connection pooling. My impression is that this is bad.
What happens if they were to double their number of customers? How many unique database connections can SQL Server handle? Is this a situation where I should tell the client that they must redesign this if they want to remain scalable?
Thanks,
Corey
You don't have to create separate connections for every DB
I have an app that uses multiple DBs on the same server. I prefix each query with a "USE dbName; "
I've even run queries on two separate DB's in the same call.
As for calling stored procs, it's a slightly different process. Since you can't do
Use myDB; spBlahBLah
Instead you have to explicity change the DB in the connection object. In .Net it looks something like this:
myConnection.ChangeDatabase("otherDBName");
then call your stored procedure.
Hopefully, you have a single database for common items. Here, I hope you have a Clients table with IsEnabled, Logo, PersonToCallWhenTheyDontPayBills, etc. Add a column for Database (i.e. catalog) and while you're at it, Server. You web application will point to the common database when starting up and build the list of database connetions per client. Programmatically build your database connection strings with the Server and Database columns in the table.
UPDATE:
After my discussion with #Neil, I want to point out that my method assumes a singleton database connection. If you don't do this then it would be silly to follow my advice.
Scaling is a complex issue. However why are you not scaling the web aspect as well? Then the connection pooling is limited to the web application.
edit:
I'm talking about the general case here. I know tha pooling occurs at many levels, not just the IDbConnection (http://stackoverflow.com/questions/3526617/are-ado-net-2-0-connection-pools-pre-application-domain-or-per-process). I was wondering whether the questioner had considered scaling at the we application level.
Hi, We are developing a multi-tenant application in Asp.Net with separate Database for each tenant, in which one of the requirement is to monitor the bandwidth usage for each tenant,
i have tried to search but not found much help on the topic,we want to monitor exactly how much bandwidth is being used for each tenant while each tenant can have its own top level domain or a sub domain or a combination of both.
so what are the available options, the ones which i can think of can be
IIS Log Monitoring means a separate application which will calculate the bandwidth for each tenant.
Log Each Request and Response for a tenant from within the application and then calculate the total bandwidth usage based on that.
Use some third part components if available
So what do you think will be the best approach, also if there is any other way to do this.
Ok, here is an idea (that I have not test, leave that to you)
On global.asax
use one of this function (find the one that have a valid final size)
Application_PostRequestHandlerExecute
Application_ReleaseRequestState
and get the size that you have send with
Response.Filter.Length
No need to metion, that you get the filename of the call using the
HttpContext.Current.Request.Path
This functions called with every single request, so you can get your size and you do the rest.
Here must note, that you need first to test this idea to see if its work, and maybe improve it, and have in mine that if you have compress the pages on server the length is not the correct and maybe you need to compress it on Global.asax to have the actually lenght.
Hope this help.
Well, since the IIS logs already contain the request size and response size, it doesn't seem like too much trouble to develop a small tool to parse them and calculate the total per day/week/month/whatever.
Trying to segment traffic based on host is difficult in my experience. Instead, if you give each tenant their own IP(s) for the applications you should be able to find programs that will monitor bandwidth based on IP.
ADDITION Is the structure of IIS that you have one website to rule them all for all tenants and on login the system forks to the proper database? If so, this may create problems with respect to versioning in that all tenant's sites will all have to have exactly the same schema and would all need to be updated simultaneously when you update the application such that a schema change is required.
Another structure, which sounds like what you may have, is that each tenant has their own website like so:
tenant1_site/appvirtualdir
tenant2_site/appvirtualdir
...
Where the appvirtualdir points to the same physical path for all tenant's sites. When all clients have the same application version, they are all using literally the same code. If you have this scenario and some sort of authentication, then you will need one IP per tenant anyway because of SSL. SSL will only bind to IP and port unlike non-SSL which will bind to IP, port and host. If that were the case, then monitoring traffic based on IP will still be simpler and more accurate as it could be done at the router or via a network monitor.
I was wondering whether it would be possible to change the sqlConnectionString used for SessionState in ASP.net based upon the domain an application is running on?
A scenario; We have 20 sites running from one application all talking to different databases depending which domain (site) they are browsing from.
When browsing www.domain1.com the application talks to the database 'db1'. The site www.domain2.com on the other hand talks to the database 'db2' etc, thus selecting the relevant content and also spreading the load to each database rather than using one master database to handle all connections for the sites.
An issue that has arisen though - for this setup we use SqlServer mode for the SessionState so all users to all sites sessions are stored in 1 aspstate database, now as the sites get busier / number of sites increase this database comes under increasing strain to handle all the session requests for all the sites and we are starting to get some timeout errors where the connections to this database are bottlenecking.
We can seperate out the sites to from their own application and set up different applications with the same code but within each application set a different Session database in each Web.Config and thus lightening the load. This task would be quite time consuming though and would result in more management in the long term. SO.. I would love to know if it's possible to modify within the code the sqlConnectionString used for SessionState, based upon a domain, before the session object is created? Can we inherit from System.Web.HttpApplication and use the Application_AcquireRequestState event to create the required setup of the HttpSessionState object?
Hopefully this makes sense and that someone can provide some pointers and prove to me that this isn't a pipe dream!
Cheers,
Steve
I think you are missing a big point--putting things in separate databases on the same server isn't going to help things at all if the bottleneck is sql server--it is either SQL running out of headroom or the network running out of bandwidth. I'd try and figure out which one it was before doing anything.
Your issue isn't so much that the connections to the database are bottlenecking, its that you are overwhelming the network connection to the database with data from all of the sessions.
By default, the Sql Server state provider simply serializes your data and ships it to the database. This is VERY inefficient and takes a LONG time to transfer on a fast network.
We solved this problem by going to a custom provider, like DOTSS that compresses session content before shipping it to the database. The compression rates we see are 80%-90% and the compression time is less than 10ms.
You can implement a custom session state provider. See MSDN for details. I've never done it, but with a little luck you can wrap the SqlServer session state module and redirect it based on the domain
First of all, I don't see there is advantage of "I would love to know if it's possible to modify within the code the sqlConnectionString used for SessionState, based upon a domain, before the session object is created" compared to set this in web.config.
Secondly, I think you need change that connection string setting in App_Start, so all the request will use that changed settings.Application_AcquireRequestState probably too late for this.
Why not split up the sites into sperate web applications and use hostheader to differentiate between the web sites. That way you could easily configure which session database you want your web application to use since each web application would have a seperate web.config file.
You could partition your session across different databases by implementing IPartitionResolver, and using a different partition for each domain.
Here's an example showing how to implement a custom partition resolver. (The example partitions by session ID, but it would be trivial to change it to partition by domain instead.)
We have several dozen development sites whose database connections are handled via the project's main Web.Config.
There is a separate configuration section corresponding to each URL on our intranet (e.g. http://development11, http://development12). We have SQL instances with a similar naming convention (DEVDB1\SQL1, DEVDB1\SQL2).
Based on the URL configured on the intranet IIS server, the app grabs the appropriate config. For testing we can easily modify the user, the database server or individual databases utilized for a particular site.
I am looking for information on how to create an ASP.NET web farm - that is, how to make an ASP.NET application (initially designed to work on a single web server) work on 2, 3, 10, etc. servers?
We created a web application which works fine when, say, there are 500 users at the same time. But now we need to make it work for 10 000 users (working with the web app at the same time).
So we need to set up 20 web servers and make something so that 10 000 users could work with the web app by typing "www.MyWebApp.ru" in their web browsers, though their requests would be handled by 20 web-servers, without their knowing that.
1) Is there special standard software to create an ASP.NET web farm?
2) Or should we create a web farm ourselves, by transferring requests between different web servers manually (using ASP.NET / C#)?
I found very little information on ASP.NET web farms and scalability on the web: in most cases, articles on scalability tell how to optimize and ASP.NET app and make it run faster. But I found no example of a "Hello world"-like ASP.NET web app running on 2 web servers.
Would be great if someone could post a link to an article or, better, tell about one's own experience in ASP.NET "web farming" and addressing scalability issues.
Thank you,
Mikhail.
1) Is there special standard software
to create an ASP.NET web farm?
No.
2) Or should we create a web farm
ourselves, by transferring requests
between different web servers manually
(using ASP.NET / C#)?
No.
To build a web farm, you will need some form of load balancing. For up to 8 servers or so, you can use Network Load Balancing (NLB), which is built in to Windows. For more than 8 servers, you should use a hardware load balancer.
However, load balancing is really just the tip of the iceberg. There are many other issues that you should address, including things like:
State management (cookies, ViewState, session state, etc)
Caching and cache invalidation
Database loading (managing round-trips, partitioning, disk subsystem, etc)
Application pool management (WSRM, pool resets, partitioning)
Deployment
Monitoring
In case it might be helpful, I cover many of these issues in my book: Ultra-Fast ASP.NET: Build Ultra-Fast and Ultra-Scalable web sites using ASP.NET and SQL Server.
I'd say you should configure an NLB cluster (Network Load Balancing), which basically splits all requests between cluster nodes (And as an added benefit detects if things are down and stops sending them requests). There's features built into windows for this, but they don't compare to a hardware device for performance or scalability. If you're using Windows 2008 it really is simple to set one up. If you do this make sure you have a shared machine key or you'll start getting exceptions for viewstate being invalid (When 1 server submits the form and it posts to the other and they're using different keys to encode the data).
You can also use DNS round-robin but at 20 servers presumably in 1 datacenter I wouldn't see a point to going to such crazy lengths. If you've got multiple data centers though this is definitely worth considering (As NLB won't really work well between data centers).
You'll also want to be sure if a user swaps servers they don't loose their session. The simplest way would be to use a Session State database (Configurable in the web.config, or you can do it server-wide in IIS's configs). If you don't use sessions though just turn them off in the Pages directive of the web.config and call it a day. You could also use a session state server, but I don't have any experience with this.
It may also be worth considering spending some time optimizing the code or adding caching directives to static content - it can be very cost-effective even if you only trim the need for a few of those servers.
Hope that helps.
If you keep your server stateless, it is easy with a good router that implements some round-Robbin protocol (that send each call to the single published server ip to a different web server).
if it is not stateless (like - if a login is required, or ssl) than you need to keep each session to the same server.
Here is some info about MS Application Request Routing - you will get everything there:
IIS Load balancing
I would not recommend #2. You will do much better off with a load balancer.
Pay attention to session state management. Unless you configure the load balancer to keep each user on the same web server, you will have to use the session state server or database.
Also, check your code's usage of Application and Cache variables. These will be different on every web server. If those values are static, you may not have a problem. But if they can change, you can end up with different values on each web server.
There used to be a problem with ViewState in 1.x, as explained here. I'm not sure if this problem still exists.
Then, there are some changes that you need to make to the Machine Key in web.config, as explained here.