What is our primary use case?
We implemented the solution for web caching in 2022. We auto-update the tool so that we don't face any security issues. We are a broadcasting company. We have many OTT clients that request a lot of images. We use a customized server that resizes images on the fly or transforms them in a different format. It’s a legacy product. It's no longer fully supported. We don't have much choice there, but getting any fixes is difficult. However, with our growth and volume over the years, these things just don't cope anymore.
Ultimately, the traffic just kills the servers. So, we put Varnish in front of the cache and reduced the traffic to the service by about 98%. Instead of being bombarded with 40 million requests every day, it's now down to one million. We used more than 15 servers before. Now, with a bit of redundancy, we have reduced it to four. It is quite a lot. I haven't seen any issues on the server for one and a half years since we deployed Varnish, so it's pretty good.
What is most valuable?
The programmability is fun to use. I get direct access to the pipeline. I use it to correct a few things that our clients and servers don't like. So I do some on-the-fly corrections and a little hacking, and it works better. The tool is really easy to install, very light, and extremely fast. It requires low maintenance. I'm quite pleased with the product. It is easy to learn the product. It's all documented properly.
The person who wants to learn the tool must know how to deal with such products. However, it is relatively simple. It's running well in our organization, and nobody needs to touch it. We haven't really touched any configuration in over a year as it's still working great. It has been running my same VCL code ever since initial deployment. It has been running stable since, apart from some minor optimizations here and there.
What needs improvement?
The monitoring features could be improved for an enterprise solution. We can see quite a few things, but getting better visibility on what is going wrong would be nice. If I really need to pinpoint an issue or find something in the ongoing traffic, it's a bit clunky and laborious but not impossible. It can be done, but it's all command-line. These monitoring features could use a bit more modernized interfaces for easier access. Things like Elasticsearch and Kibana integration would be nice. It would be nice to have something graphical/queryable like that to dig deeper into what's happening.
For how long have I used the solution?
I have been using the solution for more than one and a half years. I am using the latest version of the solution.
What do I think about the stability of the solution?
The tool is very stable. I haven't seen any issues.
What do I think about the scalability of the solution?
Our company was taken over some time ago and other teams are using Varnish as a CDN product, but they are in a different country. My team has two maintainers/operators for varnish.
How was the initial setup?
Initially, I deployed the product using the open-source version, and I'm still running that VCL code. Technically, I can switch back to the open-source version at any point because I'm not running any enterprise code except for the service itself,makes negotiating for a good deal very effective too.
What about the implementation team?
I did all of the grunt work (test setup, load test, VCL code, etc) myself before presenting the solution and getting it approved for production use.
What was our ROI?
We have basic access to supportm but so far we're good with that. We got quite a good deal for three years, so I'm not unhappy with our investment. Our ROI seems to be quite decent for the deal we got in the end.
What's my experience with pricing, setup cost, and licensing?
We got a decent deal on the product. Since we had very limited feature usage, we pushed for a good deal. Since we are not using high-end features and open source would suffice, we can get the tool for a cheaper price.
Which other solutions did I evaluate?
We were using some vendor products before. They were relatively slow, and we had to add more servers as the vendor just likes to scale horizontally. The product was not performing well with increasing load and issues kept appearing leading to end-user impact and a lot of manual maintenance. We ended up with 20 servers for a very minimalistic setup. It did nothing and still had issues. I just googled a bit and chose Varnish (open source) to try it, impressed with the results we went for the Enterprise version to ensure support.
What other advice do I have?
My recommendation: it depends on the use cases. If you have a lot of files, if your servers get hammered quite badly, or if you are doing images or video delivery, first look at your traffic patterns and analyze what's going on. Generally, Varnish’s out-of-the-box product is quite simple but the power comes from the VCL code. Just running it should bring an improvement but make sure to look at the VCL code, the programmatic layer where we can play with the pipeline and make all kinds of adjustments.
We can filter, rectify or block things via VCL. The caching mechanisms ensure that all our clients receive proper delivery of their images. Traffic doesn't kill our legacy image servers anymore.
We connected our load balancer and configured Varnish to connect to the image servers to pick them up. The integration was very simple.
Our use case itself is quite simple. Overall, I rate the solution a nine out of ten, given that the monitoring and statistical analysis could be better. If we consider the product as a whole and look at the added value of Enterprise, I would rate it a six out of ten, but we do not use the features much. Enterprise has a nice UI where I can do basic monitoring. Though the UI is nice to have, I can also run it through the command line.
Overall, I rate the solution an eight and a half out of ten.
Which deployment model are you using for this solution?
On-premises
