I need to deliver a lot of HTTP content (Lets say it simple - a Big storage with HTTTP Access - Similar to AWS S3)
The Bandwith needed for this excedds the Bandwith of one Server (We get 200MBit each Server and the question is not to change this)
For out Prog we need 1Gbit that woudl mean 5 Servers.
When I connect them togeter with mod_proxy then I have one Server in front which only has 200MBit. So thats not the right way.
But these Servers must be accassible from the Web with one Domain Name. Is there a possibillity to so that? Example: One gets the HTTP Request, but the Resonse comes from a different Server?
DNS Round Robin?
Different Idea?
Thanx
If the outbound network traffic is not CPU limited, you can use this open source Linux Network Balancer
http://lnlb.sourceforge.net/
The inbound network speed will remain at 200MBit, but with five nodes the maximum outbound limit is 5*200MBit.
A lot of people condemn round robin DNS, perhaps assuming that it will take a full TCP timeout to detect all failed node which is simply not the case. Its a simple way to solve the performance problem and improves availability a lot. This also helps to solve the potential bottleneck of your lan without having to go to 10gbit Ethernet which would be a requirement between a router and a load balancer for the rate of traffic you describe.
There may be scope for getting more throughput from your servers and hence only needing 3 or 4 servers rather than 5. But that's a very different question.
Related
Every few months when thinking through a personal project that involves sockets I find myself having the question of "How would you properly load balance sockets on a dynamic horizontally scaling WebSocket server?"
I understand the theory behind horizontally scaling the WebSockets and using pub/sub models to get data to the right server that holds the socket connection for a specific user. I think I understand ways to effectively identify the server with the fewest current socket connections that I would want to route a new socket connection too. What I don't understand is how to effectively route new socket connections to the server you've picked with low socket count.
I don't imagine this answer would be tied to a specific server implementation, but rather could be applied to most servers. I could easily see myself implementing this with vert.x, node.js, or even perfect.
First off, you need to define the bounds of the problem you're asking about. If you're truly talking about dynamic horizontal scaling where you spin up and down servers based on total load, then that's an even more involved problem than just figuring out where to route the latest incoming new socket connection.
To solve that problem, you have to have a way of "moving" a socket from one host to another so you can clear connections from a host that you want to spin down (I'm assuming here that true dynamic scaling goes both up and down). The usual way I've seen that done is by engaging a cooperating client where you tell the client to reconnect and when it reconnects it is load balanced onto a different server so you can clear off the one you wanted to spin down. If your client has auto-reconnect logic already (like socket.io does), you can just have the server close the connection and the client will automatically re-connect.
As for load balancing the incoming client connections, you have to decide what load metric you want to use. Ultimately, you need a score for each server process that tells you how "busy" you think it is so you can put new connections on the least busy server. A rudimentary score would just be number of current connections. If you have large numbers of connections per server process (tens of thousands) and there's no particular reason in your app that some might be lots more busy than others, then the law of large numbers probably averages out the load so you could get away with just how many connections each server has. If the use of connections is not that fair or even, then you may have to also factor in some sort of time moving average of the CPU load along with the total number of connections.
If you're going to load balance across multiple physical servers, then you will need a load balancer or proxy service that everyone connects to initially and that proxy can look at the metrics for all currently running servers in the pool and assign the connection to the one with the most lowest current score. That can either be done with a proxy scheme or (more scalable) via a redirect so the proxy gets out of the way after the initial assignment.
You could then also have a process that regularly examines your load score (however you decided to calculate it) on all the servers in the cluster and decides when to spin a new server up or when to spin one down or when things are too far out of balance on a given server and that server needs to be told to kick several connections off, forcing them to rebalance.
What I don't understand is how to effectively route new socket connections to the server you've picked with low socket count.
As described above, you either use a proxy scheme or a redirect scheme. At a slightly higher cost at connection time, I favor the redirect scheme because it's more scalable when running and creates fewer points of failure for an existing connection. All clients connect to your incoming connection gateway server which is responsible for knowing the current load score for each of the servers in the farm and based on that, it assigns an incoming connection to the host with the lowest score and this new connection is then redirected to reconnect to one of the specific servers in your farm.
I have also seen load balancing done purely by a custom DNS implementation. Client requests IP address for farm.somedomain.com and that custom DNS server gives them the IP address of the host it wants them assigned to. Each client that looks up the IP address for farm.somedomain.com may get a different IP address. You spin hosts up or down by adding or removing them from the custom DNS server and it is that custom DNS server that has to contain the logic for knowing the load balancing logic and the current load scores of all the running hosts.
Route the websocket requests to a load balancer that makes the decision about where to send the connections.
As an example, HAProxy has a leastconn method for long connections that picks the least recently used server with the lowest connection count.
The HAProxy backend server weightings can also be modified by external inputs, #jfriend00 detailed the technicalities of weighting in their answer.
I found this project that might be useful:
https://github.com/apundir/wsbalancer
A snippet from the description:
Websocket balancer is a stateful reverse proxy for websockets. It distributes incoming websockets across multiple available backends. In addition to load balancing, the balancer also takes care of transparently switching from one backend to another in case of mid session abnormal failure.
During this failover, the remote client connection is retained as-is thus remote client do not even see this failover. Every attempt is made to ensure none of the message is dropped during this failover.
Regarding your question : that new connection will be routed by the load balancer if configured to do so.
As #Matt mentioned, for example with HAProxy using the leastconn option.
I'm setting up a relay and my understanding is that the ORPort is required to be open to relay traffic from the outside world but do I need to enable the DirPort too?
No, you don't need to enable DirPort. If you set the DirPort, your relay will also serve as a directory mirror.
As a directory mirror other relays can query you for information about other relays on the network (for example to get a list of relays for building circuits).
Note that enabling this can significantly increase your bandwidth usage (full directory listings can be fairly large) and you'll have a lot more incoming connections. For example, one of my relays serving as a mirror has over 1200 incoming connections for dir requests and pushes an extra 10-20 Mbps as a result.
I believe the BandwidthRate option includes limiting directory traffic (which is separate from RelayBandwidthRate).
If you want to just run a relay, it's fine to leave DirPort as 0 so you can dedicate as much bandwidth to relaying. There are a lot of relays running as mirrors so I think the capacity for them is pretty good, but running one when possible is encouraged.
I'm having the following dilemma, I have a website on IIS with two internal IPs, each one of those IPs are NATed to different external IPs (each IP is from a different ISP). I also configured a RoundRobin DNS Service (two A hosts with the same name but with a different IP). Basically what this does is that the traffic is balanced between the two ISPs, and that's what we want. The thing is that apparently this configuration (DNS Roundrobin) is meant for when you have a cluster of server so each server has its own ISP on its own NIC, so the traffic from the webserver to the client is made over that ISP.
Right now we are being told that no matter where our inbound traffic comes from, the outbound traffic is always through our main WAN, which is also OK, because we have tested that when the primary WAN link is down, the website keeps working on the secondary link.
OK, the question is, do you think there may be problem with this configuration? Is the DNS Rounrobin also useful on this configuration?.
Thanks a lot for your feedback.
normally when you host a web service the responses are much bigger compared to the inbound traffic (normally you receive an HTTP GET/ and deliver the whole content back) - so it would make much more sense to balance the outbound traffic over your ISPs to get value out of your additional bandwidth.
does it make sense - yes - you can loose one ISP and your site is still available (assuming you do Healthchecks on your DNS server to determine if the sites are available before you send the IP address back - if you always deliver both IPs even when one ISP is down it won't help you at all)
it would be better to add an additional server - OR do policy based routing on your single server - so sending the response out of the interface where it was received.
hope that helps!
I want to add a load balancer infront of my nodejs websockets server. The plan is to add another node on another physical machine and have a load balancer in front. The load balancer will also be on its own physical machine.
The requirement is that several 1000s of simultaneous connections could be handled and I'm a bit worried about bouncys upper limitations.
I like the consistency of using bouncy since it is a node module, but at the same time it seems like nginx could handle more socket connections or be a bit more stable.
Anyone who has experience with bouncy or nginx as load balancer and could give me some advices?
Thanks!
nginx is pretty good for mass connections, check these answer.
https://stackoverflow.com/a/16289251/2325522
there you can see how to use Nginx as load balacer.
The only problem that you can have is the mass band-width needed to serve 1000's of simultaneous connections.
Example:
5000 clients * 0.25Mb/request (a little one)
=
1250mb (1.25Gb outgoing band-width)
Hope these solve your doubts.
It has recently come to my attention that setting up multiple A records for a hostname can be used not only for round-robin load-balancing but also for automatic failover.
So I tried testing it:
I loaded a page from our domain
Noted which of our servers had served the page
Turned off the web server on that host
Reloaded the page
And indeed the browser automatically tried a different server to load the page. This worked in Opera, Safari, IE, and Firefox. Only Chrome failed to try a different server.
But after leaving that server offline for a few minutes and looking at the access logs, I found that the number of requests to the other servers had not significantly increased. With 1 out of 3 servers offline, I had expected accesses to each of the remaining 2 servers to roughly increase by 50%, but instead I only saw 7-10%. That can only mean DNS-based failover does not work for the majority of browsers/visitors, which directly contradicts what I had just tested.
Does anyone have an idea what is up with DNS-based web browser failover? What possible reason could there be why automatic failover works for me but not the majority of our visitors?
What's happening is that the browsers are not doing automatic DNS failover.
If you have multiple A records on a domain then when your nameserver requests the IP for the domain you typed into your browser, it'll request one from the SOA. It could be any of those A records. Then it passes it along.
Some nameservers are 'smart' enough to request a new A record if the one it gets doesn't work and some aren't. So if you set multiple A records then you will have set up a pseudo redundancy failover, but only for those people with 'smart' nameservers. The rest get a toss of the dice on which IP they get and if it works then good, and if not then it will fail to load as it did for you in Chrome.
If you want to specifically test this then you can use your hosts file C:\Windows\system32\drivers\etc\hosts in Windows and /etc/hosts in Linux to specify what IP you want to go with what domain to see if you get a true failover - as what you'll run into in practicality is that DNS servers across the net will cache your domain name resolution based on its TTL. So if/when you get a real failure, that IP will still need to be resolve and be otherwise farmed out to another nameserver.
Another possible explanation is that, for most public websites, the bulk of traffic comes from bots not from browsers. Depending on the bot it is possible that they aren't quite as smart as the browsers when it comes to handling multiple A records for a domain.
Also, some bots use keep-alives to keep the TCP connections open & make multiple HTTP requests over the same connection. Given that the DNS lookup is only done when a connection is made, they will continue to make requests to the old IP address at least as long as the connection is kept open.
If the above explanation has any weight you should be able to see it in your logs by examining the user agent strings.