Amazon Elastic Load Balancing vs. Fastly Edge Cloud Platform

We use Amazon Elastic Load Balancers to serve mobile applications and websites. It works really well. We have not had any problems until now. Last year we integrated the AWS ELB with the EC2 Auto Scaling and now we have a fully working elastic solution. We increase/decrease EC2s instances based on traffic over our load balancers.

Incentivized

The service is really well-suited for pretty much any site that is primarily display-driven (that is, mostly GET requests). The network is able to handle massive volumes of traffic and their POPs have spread out pretty much anywhere that it's easy to get them (so basically everywhere but China and Russia). My team witnessed several large-scale attack attempts on some high-profile websites (attacks in the 10s of millions of requests per second) that were mitigated before ever coming back to the actual application; in one case we didn't realize the attack had happened until we looked at the logs the next day. Because it's a cache store option, the default configuration does not cache POST responses, and it can be difficult to set up things like authenticated paywalls as a result.

Incentivized

• Most obviously it works great for routing traffic between components hosted on Amazon web services
• The ability to dynamically spin up connections is fantastic.
• In general the ease of use and configuration is a selling point.

Incentivized

• Very performant and fast -- major companies have drastically reduced page load times with Fastly's edge technology.
• Support for many types of media, including video.
• Image add-on makes serving and manipulating images very simple.

Incentivized

• Occasionally we have a huge number of users using our network at once, and Amazon ELB isn't quite fast enough to scale effectively when that occurs. But this doesn't happen very often as our usage is usually quite stable
• If we want to add another application to our learning suite, we would have to add another load balancer, which would incur additional cost
• The setup was not easy and could really only be handled by one person on our team with the technical background to do so

Incentivized

• WAF is hard to configure.
• No comprehensive rate limiting without a pricey upgrade.
• Account access controls are extremely limited.

Incentivized

AWS Elastic Load Balancing has this trick. First, you need to know how it works. ELB is not the only piece here. ELB has a very close relation with AWS Target Groups. You create or select a target group every time you create a Load balancer. Target groups allow you to connect the load balancer to EC2 autoscaling groups, Lambda functions, or even a single EC2 instance. While this sounds complex, it becomes easy, once you know his tricks. Thanks to the user interface, managing a ELB is an easy task. The rules editor is really useful, although it will need a bit of improvement to some interface items

Incentivized

AWS gives you several support plans. On the free plan, you basicaly need to google for help, but the good news is that AWS Elastic Load Balancing works. We has more than 15 load balancers and we never run into a problem that require support. But you mght consider a support plan if you are going to do something more complex or critical

Incentivized

We have not used any other solution out there in the market but our dev-ops team did deep research and AWS provided us the solution we needed to be cost-effective. Also, the decision to keep working with Amazon was strategic. We were already using other AWS features and [Amazon Elastic Load Balancing] integrates great with those.

Incentivized

It’s the fastest and most configurable.

Incentivized

• Currently it is too soon to say for sure what kind of impact this will have.
• The ideal goal is that this will be cheaper than having to host our own routing on site.

Incentivized

• Fastly dramatically reduces load times for clients all over the world thanks to their global system of POPs.
• Fastly is a cost-effective choice when a highly-performant CDN is required and makes it simple to support data that is quickly invalidated.

Incentivized