Digital Oceans has officially opened new AMS2 datacenter in Amsterdam. This event forced me to do some small tests on quality of services, from my own (Poland's perspective) point of view.
I'm not talking about Quality of Service, because this is beyond any discussion. With standard (simple to answer) tickets being answered with 3-5 minutes (!!!) and with more complex cases solved with 3-5 hours, Digital Ocean's set itself a new world-standard. I have never found or met support of this scale.
So, if we're not talking about QoS, lets talk about performance.
I've gathered together six pings to six different destinations (five in Digital Ocean's cluster):
Each test (ping) included standard ping (default size and timeout) sent 1000 times (enough, in my opinion, for a medium-value analysis) to six different locations.
This includes all four Digital Ocean datacenters currently available (AMS1 was not available during tests):
- two in New York (three droplets -- standard in NY1, powerful in NY1 and standard in NY2),
- one in San Francisco,
- one in Amsterdam,
- one datacenter in Poland (home.pl) for doing some comparisons.
Since all four DO's droplets setup for this test were smallest available one (512 MB of RAM, 1 core, 20 GB of SSD disk space, priced $5 per month or $0.007 per hour) I've decided to add one extra droplet (second in NY1 datacenter), a little bit more powerful (32 GB of RAM, 12 cores, 320 GB of SSD disk space, priced $320 / month or $0.476 / hour). Again, to compare.
Here are results caputred by me:
In first (left) column we have (starting from top):
22.214.171.124-- home.pl Polish datacenter added for compare,
126.96.36.199-- AMS2 datacenter standard DO droplet,
188.8.131.52-- SFO datacenter standard DO droplet.
Second (right) column shows:
184.108.40.206-- NY1 datacenter standard DO droplet,
220.127.116.11-- NY2 datacenter standard DO droplet,
18.104.22.168-- NY1 datacenter powerful DO droplet.
All tests were performed from Poland (that's why first test gave best results, as it was ping to local, Polish datacenter) and were conducted in a really top-level network environment (main core up to 200 Mbps).
Final tests (each after running 1000 cycles of ping) showed up, that we have:
22.214.171.124-- home.pl: 9 ms min, 43 ms max, 9 ms med,
126.96.36.199-- AMS2: 35 ms min, 42 ms max, 35 ms med,
188.8.131.52-- NY1 standard: 103 ms min, 109 ms max, 103 ms med,
184.108.40.206-- NY1 power: 104 ms min, 363 ms max, 104 ms med.
220.127.116.11-- NY2: 115 ms min, 891 ms max, 118 ms med,
18.104.22.168-- SFO: 194 ms min, 206 ms max, 195 ms med.
All Digital Ocean's droplets returned 100% of pings (1000 / 1000). Home.pl server lost only one ping, which is highly acceptable. Most Digital Ocean's droplets returned medium response time the same or only 1-3 ms above minimum time (see below remarks).
This proves, that entire DO's infrastructure is not only one of most powerful in the entire world, but it is very stable in the same time. We need to remember, that stability and reproducibility of results is often much more anticipated than overal performance.
Conclusions, that we can rise, after analysing above are quite obvious:
- Geographic location of datacenter has direct influence to performance and times. Local datacenter gave ping response times below 10 ms, while San Francisco datacenter, located farest from test end-point, have slowest results.
- Hardware configuration has no visual influence on ping response times. A powerful server with 12 times more CPU power (12 cores to 1) and 64 times more memory (32 GB to 512 MB) gave only 1 ms slower minimum and medium result than corresponding standard or slow server.
- NY2 datacenter gave unusually high maksimum response time (891 ms), which influenced medium response time for this server (118 ms). For this reason, this test was restarted once again and produced results of 115 ms min, 137 ms max and 115 ms med. So, it is back to normal and that nearly 900 ms maximum response time was just a one accident.
General conclusion is even more obvious.
For services hosted in Europe, Digital Ocean's AMS2 datacenter is the best possible option. It offers fastest response times for Europe, great stability and (as of writing this) is Digital Ocean's the only location, which offers most powerful droplet -- 96 GB of RAM, 24 cores, 960 GB SSD disk space and 10 TB of monthly transfer -- priced $960 per month or $1.411 per hour.
Nearly thousand dollars per month or nearly 1.5 USD per hour is surely high price. But, on the other hand, Digital Ocean's AMS2 datacenter is currently one of the few in Europe (and probably in the entire world) to offer single customer (with just a credit card, without all this unnecessary business crap, long-term agreements and costly obligations) a nearly 1 TB of SSD disk space.
This is an offer, that will surely suit most demanding among us.