mLab vs Compose, on performance


#1

Has anyone done (or found) a performance shoot-out between mLab and Compose? Specifically their entry level plans ($15/mo and $31/mo, respectively). I’m wondering if Compose’s auto-scaling gives it an edge over mLab’s shared plan.


#2

Anyone? I’m wondering as well.


#3

Nothing scientific, but I have run load tests on comparable plans across mLab ($15 shared), Atlas ($65 3 replica cluster), and found mLab to be somewhat faster, surprisingly, up to the max load that I would expect on my service. I tried compose a while back, and wasn’t too impressed with the interface and experience with latency. So, I have no numbers, but mLab serves me quite well.

Using mLab for all deployments now.


#4

I had used Compose for a couple of years on a couple different projects and really liked what they had to offer from price to quality. One issue that I kept running into was the lack of options when scaling deployments up or down. Compose’s scaling features (at least how they were at the time, not sure about now) were very limited. For example, I could not really customize IOPS without hugely increasing costs because I also needed to greatly increase the rest of my hardware (memory and storage). My app at the time was very write heavy and I didn’t really need extra storage or more RAM, just better IO.

I’ve switched over to MongoDB Atlas (from the MongoDB guys) and absolutely love it. They do everything that Compose does but also allow more customizability when it comes to picking specific instance sizes, memory, IOPS, storage, replication nodes, etc.

They offer a nice real-time metrics panel to help track down issues and the deployment performance has been superb since I’ve been using it (almost a year now).

They are pretty affordable as well and their pricing is comparable to Compose but with more options.

I can’t really speak to mLab as I’ve never used it.