We are building a scaled application that uses WebSockets on AWS EC2. We were considering using the default ELB (Elastic Load Balancing) for this, but that, unnecessarily, makes the load balancer itself a bottleneck for traffic-heavy operations (see this related thread), so we are currently looking into a way to send the client the connection details of a "good instance" to connect to instead. However, the Elastic Load Balancer API does not seem to support a query of the sort "give me (public) connection details of a good instance", which is odd because that is the core functionality of any load balancer. Maybe I have just not looked at the right place?
UPDATE:
Currently, we are investigating two simple solutions using default implementations:
Use ELB in TCP mode which tunnels all traffic through the ELB.
Simply connect to the public IP of the instance that the ELB connected you to for your GET request. The second solution requires public IPs to be enabled, but does not route all traffic through the ELB.
I was concerned about that very last part because I assumed that the ELB is not in the same building as the instance it gave you. But I assume, it usually is in the same building or has some other high-speed connection to the instances? In that case, the tunneling overhead is negligible.
Both solutions seem to be equally viable, or am I overseeing something?
If your application manages to make the ELB a bottleneck, then you are a pretty big fish. Why don't you try first using their load balancer trusting that they do their job right? It is difficult to make it "better", and the most difficult part about this is to define what is "better" in the first place. You definitely did not very well define that in your question, so I am pretty sure that you are well off using just their load balancer.
In some cases it might make sense to develop your own load balancing logic, especially if your machine usage depends on very special metrics not per se accessible to the ELB system.
Yes, I'd say both solutions are viable.
The upside of the second is that it allows greater customization of the load balancing logic you may want to implement (providing an improvement over ELBs round robin), dispatching requests to a server of your convenience after an initial HTTP GET request.
The downside may be on the security front. It's not clear whether security, and SSL is part of your requirements, but in case it is, the second solution forces you to handle it at the ec2 instances level, which can be inconvenient and affect each node's performance. Otherwise websocket communications may be left unsecured.
Related
We have many internet services, what are the considerations whether to use alb per service or single alb for all using listener rule pointing to target 🎯 group.
The services has its own clusters/target group with different functionality and different url.
Can one service spike impact other services?
Is it going to be a single point of failure ?
Cost perspective ?
Observability, monitoring, logs ?
Ease of management ?
Personally I would normally use a single ALB and use different listeners for different services.
For example, I have service1.domain.com and service2.domain.com. I would have two hostname listeners in the same ALB which route to the different services.
In my experience ALB is highly available and scales very nicely without any issues. I've never had a service become unreachable due to scaling issues. ALB's scale based on "Load Balancer Capacity Units" (LBCU). As your load balancer requires more capacity, AWS automatically assigns more LBCU's which allows it to handle more traffic.
Source: Own experience working on an international system consisting of monoliths and microservices which have a large degree of scaling between timezones.
You don't have impact on service B if service A has a spike, but the identification of which service is having bad times could be a little pain.
For monitoring perspective it's is a bit hard because is not that easy to have a fast identification of which service/target is suffering.
For management, as soon as different teams need to create/management its targets it can create some conflicts.
I wouldn't encourage you using that monolith architecture.
From cost perspective you can use one load balancer with multi forward rules, but using a single central load balancer for an entire application ecosystem essentially duplicates the standard monolith architecture, but increases the number of instances to be served by one load balancer enormously. In addition to being a single point of failure for the entire system should it go down, this single load balancer can very quickly become a major bottleneck, since all traffic to every microservice has to pass through it.
Using a separate load balancer per microservice type may add additional overhead but it make single point of failure per microservice in this model, incoming traffic for each type of microservice is sent to a different load balancer.
I am reading about load balancing.
I understand the idea that load balancers transfer the load among several slave servers of any given app. However very few literature that I can find talks about what happens when the load balancers themselves start struggling with the huge amount of requests, to the point that the "simple" task of load balancing (distribute requests among slaves) becomes an impossible undertaking.
Take for example this picture where you see 3 Load Balancers (LB) and some slave servers.
Figure 1: Clients know one IP to which they connect, one load balancer is behind that IP and will have to handle all those requests, thus that first load balancer is the bottleneck (and the internet connection).
What happens when the first load balancer starts struggling? If I add a new load balancer to side with the first one, I must add even another one so that the clients only need to know one IP. So the dilema continues: I still have only one load balancer receiving all my requests...!
Figure 2: I added one load balancer, but for having clients to know just one IP I had to add another one to centralize the incoming connections, thus ending up with the same bottleneck.
Moreover, my internet connection will also reach its limit of clients it can handle so I probably will want to have my load balancers in remote places to avoid flooding my internet connection. However if I distribute my load balancers, and want to keep my clients knowing just one single IP they have to connect, I still need to have one central load balancer behind that IP carrying all the traffic once again...
How do real world companies like Google and Facebook handle these issues? Can this be done without giving the clients multiple IPs and expect them to choose one at random avoiding every client to connect to the same load balancer, thus flooding us?
Your question doesn't sound AWS specific, so here's a generic answer (elastic LB in AWS auto-scales depending on traffic):
You're right, you can overwhelm a loadbalancer with the number of requests coming in. If you deploy a LB on a standard build machine, you're likely to first exhaust/overload the network stack including max number of open connections and handling rate of incoming connections.
As a first step, you would fine tune the network stack of your LB machine. If that still does not provide you the required throughput, there are special purpose loadbalancer appliances on the market, that are built ground-up and highly optimized to handle a large number of incoming connections and routing them to several servers. Examples of these are F5 and netscaler
You can also design your application in ways that help you split traffic to different sub domains, thereby reducing the number of requests 1 LB has to handle.
It is also possible to implement a round-robin DNS, where you would have 1 DNS entry point to several client facing LBs instead of just one as you've depicted.
Advanced load balancers like Netscaler and similar also does GSLB with DNS not simple DNS-RR (to explain further scaling)
if you are to connect to i.e service.domain.com, you let the load balancers become Authorative DNS for the zone and you add all the load balancers as valid name servers.
When a client looks up "service.domain.com" any of your loadbalancers will answer the DNS request and reply with the IP of the correct data center for your client. You can then further make the loadbalancer reply on the DNS request based of geo location of your client, latency between clients dns server and netscaler, or you can answer based on the different data centers load.
In each datacenter you typically set up one node or several nodes in cluster. You can scale quite high using such a design.
Since you tagged Amazon, they have load balancers built in to their system so you don't need to. Just use ELB and Amazon will direct the traffic to your correct system.
If you are doing it yourself, load balancers typically have a very light processing load. They typically do little more than redirect a connection from one machine to another based on a shallow inspection (or no inspection) of the data. It is possible for them to be overwhelmed, but typically that requires a load that would saturate most connections.
If you are running it yourself, and if your load balancer is doing more work or your connection is getting saturated, the next step is to use Round-Robin DNS for looking up your load balancers, generally using a combination of NS and CNAME records so different name lookups give different IP addresses.
If you plan to use amazon elastic load balancer they claim that
Elastic Load Balancing automatically scales its request handling
capacity to meet the demands of application traffic. Additionally,
Elastic Load Balancing offers integration with Auto Scaling to ensure
that you have back-end capacity to meet varying levels of traffic
levels without requiring manual intervention.
so you can go with them and do not need to handle the Load Balancer using your own instance/product
Questions about load balancers if you have time.
So I've been using AWS for some time now. Super basic instances, using them to do some tasks whenever I needed something done.
I have a task that needs to be load balanced now. It's not a public service though. It's pretty much a giant cron job that I don't want running on the same servers as my website.
I set up an AWS load balancer, but it doesn't do what I expected it to do.
It get's stuck on one server, and doesn't load balance at all. I've read why it does this, and that's all fine and well, but I need it to be a serious round-robin load balancer.
edit:
I've set up the instances on different zones, but no matter how many instances I add to the ELB, it just uses one. If I take that instance down, it switches to a different one, so I know it's working. But I really would like it to always use a different one under every circumstance.
I know there are alternatives. Here's my question(s):
Would a custom php load balancer be an ok option for now?
IE: Have a list of servers, and have php randomly select a ec2 instance. Wouldn't be scalable at all, bu atleast I could set this up in 2 mins and it can work for now.
or
Should I take the time to learn how HAProxy works, and set that up in place of the AWS ELB?
or
Am I doing it wrong, and AWS's ELB does do round-robin. I just have something configured wrong?
edit:
Structure:
1) Web server finds a task to do.
2) If it's too large it sends it off to AWS (to load balancer).
3) Do the job on EC2
4) Report back via curl to an API
5) Rinse and repeat
Everything works great. But because the connection always comes from my server (one IP) it get's sticky'd to a single EC2 machine.
ELB works well for sites whose loads increase gradually. If you are expecting an uncommon and sudden increase on the load, you can ask AWS to pre-warm it for you.
I can tell you I used ELB in different scenarios and it always worked well for me. As you didn't provide too much information about your architecture, I would bet that ELB works for you, and the case that all connections are hitting only one server, I would ask you:
1) Did you check the ELB to see how many instances are behind it?
2) The instances that you have behind the ELB, are all alive?
3) Are you accessing your application through the ELB DNS?
Anyway, I took an excerpt from the excellent article that does a very good comparison between ELB and HAProxy. http://harish11g.blogspot.com.br/2012/11/amazon-elb-vs-haproxy-ec2-analysis.html
ELB provides Round Robin and Session Sticky algorithms based on EC2
instance health status. HAProxy provides variety of algorithms like
Round Robin, Static-RR, Least connection, source, uri, url_param etc.
Hope this helps.
This point comes as a surprise to many users using Amazon ELB. Amazon
ELB behaves little strange when incoming traffic is originated from
Single or Specific IP ranges, it does not efficiently do round robin
and sticks the request. Amazon ELB starts favoring a single EC2 or
EC2’s in Single Availability zones alone in Multi-AZ deployments
during such conditions. For example: If you have application
A(customer company) and Application B, and Application B is deployed
inside AWS infrastructure with ELB front end. All the traffic
generated from Application A(single host) is sent to Application B in
AWS, in this case ELB of Application B will not efficiently Round
Robin the traffic to Web/App EC2 instances deployed under it. This is
because the entire incoming traffic from application A will be from a
Single Firewall/ NAT or Specific IP range servers and ELB will start
unevenly sticking the requests to Single EC2 or EC2’s in Single AZ.
Note: Users encounter this usually during load test, so it is ideal to
load test AWS Infra from multiple distributed agents.
More info at the Point 9 in the following article http://harish11g.blogspot.in/2012/07/aws-elastic-load-balancing-elb-amazon.html
HAProxy is not hard to learn and is tremendously lightweight yet flexible. I actually use HAProxy behind ELB for the best of both worlds -- the hardened, managed, hands-off reliability of ELB facing the Internet and unwrapping SSL, and the flexible configuration of HAProxy to allow me to fine tune how things hit my servers. I've never lost an HAProxy instance yet, but it I do, ELB will just take that one out of rotation... as I have seen happen when the back-end servers have all become inaccessible, which (because of the way it's configured) makes ELB think the HAProxy is unhealthy, but that's by design in my setup.
I have created two Amazon EC2 instances. After that I created an Elastic Load Balancer and registered the two instances in it.
Now what I would like to know is, when we use the DNS name of the load balancer, which instance will the load balancer use?
The idea of Load balancing is to distribute workload across multiple computers or a computer cluster, network links, central processing units, disk drives, or other resources [...].
While there are many algorithms conceivable, the general goal is to achieve optimal resource utilization, maximize throughput, minimize response time, and avoid overload, which usually implies transparent distribution of the load between the load balanced resources. Therefore you usually won't know (and shouldn't need to know), which load balanced resource serves a particular request.
Accordingly, Elastic Load Balancing (ELB) automatically distributes incoming application traffic across multiple Amazon EC2 instances.
How this is done specifically is a fairly complicated topic, mostly due to the ELB routing documentation falling short of being non existent, so one needs to assemble some pieces to draw a conclusion - see my answer to the related question Can Elastic Load Balancers correctly distribute traffic to different size instances for a detailed analysis including all the references I'm aware of.
For the question at hand I think it boils down to the somewhat vague AWS team response from 2009 to ELB Strategy:
ELB loosely keeps track of how many requests (or connections in the
case of TCP) are outstanding at each instance. It does not monitor
resource usage (such as CPU or memory) at each instance. ELB
currently will round-robin amongst those instances that it believes
has the fewest outstanding requests. [emphasis mine]
stf ,
you cannot come to know, for which server load is distributing through EBS , EBS internally take care of request distribution .
Of course you can figure out which server your request goes to!
On each server you are going to need something akin to a health_check.html file (can be named anything, someone suggested index.htm but that is a bad idea and is another discussion entirely) so the load balancer can call it and determine how long it took to get a response.
On server #1 put the following in the health_check.html file: <HTML><BODY>1</BODY></HTML>
On server #2 put this in the health_check.html file: <HTML><BODY>2</BODY></HTML>
Now when you navigate to www.YourDomain.com/health_check.html you will know exactly which server you are on.
Clear your cookies and re-navigate to the same URL to see which server you get next. Good luck cloud developer!
Hey guys I was wondering if this seems like a viable solution to the age old problem of Amazon Elastic Load Balancer's lacking a dedicated IP, and thus A record support.
What if I created a micro/small instance and hooked it to an elastic IP. I can then use that IP as my A record address for my website. That instance will forward 100% of its traffic to an ELB load balancer address (Haproxy?), which will then operate normally and forward that traffic to my server pool.
With this architecture I can use my A-record and an ELB.
Are there any downsides to this aside from the cost of the initial instance that forwards its traffic to the ELB?
Will this double forwarding create too much lag or is it really negligible since they're all in AWS?
Thanks for feedback.
If you are currently using Route53 for you DNS, it does have support for handling zone apex.
https://forums.aws.amazon.com/message.jspa?messageID=260459
Not sure if this answers your question since you didn't mention why you need a dedicated ip.
Are there any downsides to this aside from the cost of the initial instance that forwards its traffic to the ELB?
Er, yes. You're loosing about 99.9% of the benefits of ELB.
Will this double forwarding create too much lag or is it really negligible since they're all in AWS?
No, the lag should be small (sub-milisecond). The two main problems are:
1) Your instance will become a bottleneck when your traffic increases. You won't be able to survive a sudden rush, such as being linked from a high-traffic website like Slashdot or Oprah.
The whole point of ELB is that they can manage scaling (the frontend and the backend) for you. If you insert a single box in the flow, it kinda prevents ELB from doing anything useful.
Also, the micro instance can take very little traffic. You have to go to at least a m1.large if you won't want your network packets throttled.
2) Your instance will become a Single Point Of Failure. When your box dies, your website will be down. ELB can prevent problems on both the front and backend with redundancy.
Perhaps if you explained why you needed an A record?
(It is also possible to run your own front-end(s): Just create a box with an EIP, and put nginx and/or HAProxy on it. But as with everything, there are trade-offs.)