Bit late with this month’s post, been a crazy month for me!
Since last month, DigitalOcean has released Projects which doesn’t really affect these benchmarks, but is notable because the concept of grouping servers has been something that both Linode and Vultr have had in some capacity for a while now.
Also quite notable, DigitalOcean’s UI has seemed to have received a bit of an upgrade. It’s been the better looking of the bunch for a while now and they just keep pushing the visual appeal way ahead of the rest.
For these benchmarks I spun up 3 boxes on each provider at the $5 price point running Ubuntu 18.04.1 LTS in the New York / New Jersey data centers. The results are then averaged and presented to you.
Worth noting, because people keep asking. I originally dumped Lightsail because it seemed like nobody was all that into it but since people have been asking about it more, I was going to include it in this month’s results.
Sadly, Amazon’s currently only offering older Linux distros, specifically Ubuntu 16.04 LTS. They tend to be slow to roll out support for things like that and since I like to run my benchmarks on the latest version of Ubuntu (perhaps moving to Debian soon ;) I am most likely going to continue to omit Lightsail from these reviews.
Another request I get from time to time is that I should include some sort of metric about each company’s support. I get it, support is important but I do feel that bogging down each company’s support team with erroneous requests just for the sake of these posts isn’t very cool.
That said, I’ve had to interact with each company’s support team in a legitimate capacity in the past and I found Linode and DigitalOcean’s support to be top notch (quick turn around, very friendly).
Couple more notes, I’m still omitting the database metrics as I haven’t had the
time to research the issues I’ve been having with
sysbench and I also adjusted
ab tests to run 500 concurrent instead of 1000. The reason for that change
is that for whatever reason running
ab today was throwing connection reset
errors (regardless of the site / server).
Suspect perhaps Spectrum (or my router) has decided to throttle connections
ab was looking like some sort of brute force attack.
Enough about my day, how about those benchmarks??!?
|Location||New York 1||Newark, NJ||New York (NJ)|
|RAM||1 GB||1 GB||1 GB|
|CPU||1 Core||1 Core||1 Core|
|SSD||25 GB||25 GB||25 GB|
|Transfer||1 TB||1 TB||1 TB|
|Cache Size (KB)||28928.00||16384.00||16384.00|
|Events per Second||883.78||770.61||771.06|
|Ops per Second||3581264.20||3295047.66||3233764.98|
|Ops per Second||3578008.75||3307136.49||3212908.55|
|Reads per Second||2023.05||1936.46||1219.23|
|Writes per Second||1348.69||1290.98||812.82|
|Fsyncs per Second||4308.02||4124.13||2591.94|
Apache Benchmark (against
nginx on the servers)
|Requests per Second||213.85||215.15||188.86|
|Time per Request (ms) (mean)||2349.07||2333.17||2649.64|
|Transfer Rate (Kbytes/sec)||179.39||180.48||158.43|
Aside from having the slowest CPU speeds, DigitalOcean crushed the competition in CPU cycles and memory reads and writes. File I/O was also favorable for DigitalOcean but their maximum event time was 6 times as long as that of Linode and Vultr.
What I found to be quite interesting is that Vultr, who touts having the fastest network, was only trivially faster than DigitalOcean. Linode, being the closest to the test speed test server reported very poor download results.
As always, YMMV but looking at the benchmark results and factoring in UI and platform improvements seemingly every month or so, DigitalOcean seems to be the best bet these days.
If you found this comparison helpful in making your host decision, I’d be quite thankful if you used one of my referral links below.
Until next month, happy hosting!