1. 27
  1.  

  2. 10

    2bluesc on HN made a pricing/“spec” provider comparison gist.

    1. 6

      … and someone else expanded it to include DigitalOcean, Vultr, Linode, OVH, and Scaleway:

      https://gist.github.com/justjanne/205cc548148829078d4bf2fd394f50ae

    2. 9

      One caveat here that I haven’t seen much attention brought to is this piece of their FAQ:

      How do Lightsail instances perform?

      Lightsail instances are specifically engineered by AWS for web servers, developer environments, and small database use cases. Such workloads don’t use the full CPU often or consistently, but occasionally need a performance burst. Lightsail uses burstable performance instances that provide a baseline level of CPU performance with the additional ability to burst above the baseline. This design enables you to get the performance you need, when you need it, while protecting you from the variable performance or other common side effects that you might typically experience from over-subscription in other environments.

      If you need highly configurable environments and instances with consistently high CPU performance for applications such as video encoding or HPC applications, we recommend you use Amazon EC2.

      This infers that these instances perform similarly to EC2’s T2 instances, and so CPU will be throttled for anything more demanding than a web application.

      Their documentation doesn’t mention anything about CPU credits however, so some testing needs to be done to see how burstable/fixed performance is. I’ll probably write a blog post investigating this further.

      1. 7

        They are t2 instances from my testing. If you launch one and query the EC2 metadata service, it can return back the type of instance it is.

        1. 1

          Interesting. I’m assuming you queried it through the instance itself and not an API?

          1. 2

            I SSH’d to the box and curl’d the metadata URL. For my test $5 Ubuntu box it returned t2.nano.

            1. 1

              I don’t know if they have, but I’ve seen some people query it on the instance and see a T2 instance as a result

        2. 4

          This is a very similar to what Digital Ocean offers. I wonder if Amazon was feeling any kind of threat from companies like DO, or just looking for another lucrative hosting business? I am glad they are taking steps to simplify getting started on AWS. Having taken a class in using AWS, there is a lot of functionality there that can be confusing to navigate.

          1. 2

            Yes, this is very similar, except that Digital Ocean doesn’t provide a built-in firewall (you have to mess with iptables), doesn’t offer something similar to S3 (you have to use S3 which means network latency and increased cost for egress) and can’t manage my database for me (with AWS RDS for MySQL or PostgreSQL). DO will be perfect for my use case when they add this, and I hope they’ll do :-)

            1. 1

              Another area that I’ve found DO falls short, is that they allow only 1 external IP per droplet. Almost every other host sells extra IPs.

              1. 3

                That’s true, but with TLS Server Name Indication being supported by almost all clients in use nowadays, I think this is becoming less important.

                1. 2

                  You’re right, SNI fixed one use-case for multiple IPs, but many others remain. There’s the privacy argument and SEO penalties for starters. My biggest issue right now, though, is that I want to run email for multiple domains on one droplet. However, one IP only gets one PTR record, and some email hosts count a mismatched PTR record as an indication of spam.

                  1. 3

                    I’ve been hosting email for two domains on one VPS with one IP for a while with no problems. It’s true that you should have your mailserver’s forward and reverse DNS match, so for mail purposes it can only really have one hostname if it has one IP. I.e. the same IP can’t be addressed as both mail.example.com and mail.example.org, because one of these will then fail to have a matching PTR. But there’s no requirement that the MX record for a domain point to a server in the same domain. So you can have the MX records for both example.com and example.org point to mail.example.com, which has one IP and a matching PTR record of mail.example.com. Works for outgoing mail as well, as SPF and DKIM also don’t have any same-domain requirement.

                2. 2

                  Can you not use DO’s floating IP service to provide multiple external IPs (I haven’t tried to do so, hence asking)?

                  I’m still working through the Lightsail documentation, but I don’t see any mention of IPv6 support? I do wonder why Amazon hasn’t yet been able to deliver VPC IPv6 support.

                  1. 1

                    Not really. DO gives only one “floating IP” per droplet, so they’re very limited too. Also, the floating IP will only route incoming traffic to the “anchor IP” on the droplet. So you can use it to respond to incoming connections, but you can’t initiate an outgoing connection from it (I think, not 100% sure there). Also, I’m not sure if you can even set the PTR record for the floating IP.

                    I really don’t get why DO doesn’t just allow extra IPs. It’s not a problem at any other host, and lots of their customers have been asking for years [0].

                    I’ve been seriously considering migrating to another provider over this. It’s a shame, because I like everything else about DO.

                    [0] https://www.digitalocean.com/community/questions/can-i-get-additional-ip-addresses-for-a-droplet

                    1. 1

                      Ah, thanks for the info. I assumed (never a good thing!) that it would be possible to add more than one to a droplet.

                      Yes, the single IP per droplet is odd. I can’t help but think it’s intentional, forcing you to use multiple droplets instead of a single droplet hosting services on unique IPs.