Cloud Update

This post is unlike all my previous posts in that it is not a tutorial. However, I have reviewed many cloud providers in the past, and there have been significant changes that I think others may be interested in, so here is my first "Cloud Update".

Digital Ocean

It appears that Digital Ocean is coming to London! I found this out because "Lon" appeared in the images for the API, which caused a bug in the CLI application I had built. I raised a ticket with D.O. and they have pretty much confirmed it. Very recently, they even released the following on Google+ so it seems to be imminent.

Amazon Web Services

Amazon have released t2 instances which revolve around "bursting" the CPU.

This is basically an implementation of the credit scheduler from Xen if you want to do this yourself. I have switched almost all of my AWS instances over to this since it has proved very effective. My biggest issue with AWS was the fact that the t1 micro cost considerably more than a Digital Ocean instance, and performed significantly worse, with a huge amount of steal and network wait time. Now I only have to worry about network wait time which isn't a deal-breaker. These instances are great for running the little services that you need "sitting around" such as your SVN or Satis package server. (I actually run them all on one instance with a reverse proxy because I'm super cheap). I am thinking of moving some of my Digital Ocean instances over to AWS simply because of the elastic and object storage. Storage really does not scale well on Digital Ocean.


Wable has just released a new location in New York! This brings their location count to three, but sadly they are all in the USA, the land of the NSA.

Also, they have drastically reduced the point of entry to the services since I last reviewed them to just $0.75 which makes them the cheapest offering on the net I believe.

It appears that the "bundles" are allocated to specific servers rather than mixing all the customers together. This means that if you have bundle 1 or 2, you are stuck on the same servers as all the other bundle 1 & 2 customers. This does not sound like an issue, but it means all the "cheaper" customers are grouped together on the same servers, separated from the "enterprise" users, which will have a better experience. This is because the company takes care to ensure that the servers are balanced such that customers are always able to live-resize upwards whenever they desire (a feature that does not exist in D.O. or AWS). This means that if you are a higher bundle customer, the chances are far greater that the server you are allocated to is under-utilized, so you will benefit from a better network and less/non-existent CPU steal time.

Upgrading to a higher bundle does not mean your existing deployed instances will immediately migrate to a better server, instead you need to shut them down and just re-deploy. All new instances will be on the different physical server grouping.

My impression of Wable is as follows. If AWS released a chocolate covered stick, I would immediately know about it and there would be five articles/blogs talking about how amazing it is, Wable is the opposite. It doesn't have the brand-name/reputation yet, and it is generally pretty difficult to find anything out except if you manage to get in touch with the guys running the show. I haven't had a single issue so far, and it has provided an excellent service at a brilliant price-point, but I would feel uncomfortable recommending it for deploying a company's infrastructure, at the moment, simply because it hasn't got the brand recognition.

This is an underdog to keep an eye on.

No comments:

Post a Comment