Monthly Archives: October 2011


Its been a little while so why not start by writing a review 🙂 as a shameless disclaimer this is motivated by’s offer of a free node for life over on there blog, eagle eyed readers may notice that I have previously blogged about them before:

I’ve used in various guises over the last 2 years, initially as a personal vps and then as a place to host client servers (via there own separate accounts), My general thoughts of them have varied quite wildly depending on issues they had and whether they affected my clients or not. At the moment, I dont personally actively use except for development instances which its absolutely brilliant for and mightily flexible.

Just to give a basic overview of there services, they have there core offering of vps hosting, but have since opened up “cloud hosting” which in reality sounds just like cpanel on a vps (albeit a big one) with a custom UI. They are now taking this to the next level and expanding it into “geohosting” which effectively places your site over a larger geographic region and via there DNS provided by softlayer they point users at the server closest to them and handle all the sync issues transparently. This service last I checked is still in beta and I’m unsure how well it works with sites that make use of mysql databases and am yet to be able to test it to understand it. My review focuses mostly on the server aspect of rather than the cloud hosting, I’m going to avoid rehashing some of the content in previous posts but focus on 3 specific areas: Locations, SAN’s, Support, I dont think by any means this is totally comprehensive and so I would recommend you take a look at a few other reviews and also at there forum.

Locations is available in multiple locations, Currently US East, West and Central, UK (London), EU (Amsterdam), and now just recently Japan (Tokyo), this is a good mix and combine this with there DNS failover and you can see how quite a formidable High Availability Setup could be built out. Availability does vary widely and each location is prone to its own issues for example this summer the London clouds had major issues because of the heatwave and due to the complexities of London and its Datacenters cooling is often more marginal in the summer months which means any failure of such critical systems has dramatic consequences.

The locations tend to be very similarly specced out, I believe one or 2 clouds are still on the old Radar based technology (what onapp was before it became onapp) have pledged to migrate all radar clouds to onapp and with this I think we will lose one great feature of the radar clouds – bandwidth and cpu usage charts for the most part have full, 1 hour, day, month and year graphs where as the onapp clouds only have the last approx 13-15 hours or so.

Relating to this, has a wide range of templates but as a result of the differing technologies, some templates are not available on some clouds, for example jumpbox isn’t available on radar clouds. From my experience also, any template that says it is optimised is often far from the case, either certain things that boost performance are not done, or things that bring questionable performance gains are tweaked and there isnt really any consistency – thankfully it looks like have moved away from building lots of optimised templates, i would still always recommend starting from scratch or hiring a pro to do it correct from the start having understood your needs and approximate traffic as optimisations are always relative to what you are running!

If you have read other blog posts here about updating to debian squeeze (6.0) then you will know that currently, and i believe this is still the case, the debian squeeze templates are broken and should be avoided, if you need squeeze on, then follow my guide here:


if is anything to go by then this is by far and away’s biggest issue, without getting into lots of details from the outside, it looks plain and simple like there architecture is majorly flawed and therefore is why they are having so many problems with them, they initially blamed it on software vendors that make the SAN software but its been a while since that was rectified before the latest issues have started occuring. in a nutshell things are ok until a SAN fails, when a SAN fails things are initially ok when Failover SAN takes over. The biggest issue however is resyncing data saps’ performance out of vps’ so much so that everytime it happened to some previous clients, there sites would basically be totally offline because the server could not provide enough IO capacity to serve some html and constantly went read only.

They still have the IO issues with SAN’s although its much less prominent than recent months, it is still an issue, that needs solving at an infrastructure level, which I’m told is something they have done in the new Japan cloud and all future clouds, I’m also led to believe that in time the older clouds will gradually get upgraded to the newer tech.

Backups are plagued by a lot of these issues, so much so that it led to put a disclaimer on the snapshot backups to say they are not guarenteed, yet continue to go ahead and charge customers for the pleasure of not guarenteeing to backup there servers, I think this is not acceptable and I do recommend that clients have a totally seperate backup service from and don’t even pay them for a backup service. there rsync option is still fairly reliable if slightly slow and the r1soft is ok if your happy to run your own CDP backup server and stick with version 2 without mysql (they still haven’t got around to licensing for V3 yet)

Despite the issues, performance of the SAN’s is acceptable for most uses. however if you want to do anything that is disk intensive or high traffic, (or both) I cant quite recommend it, for basic wordpress sites, with the right optimisation and caching it wont be a problem whatsoever (hint, take a look at w3 total cache!) but for more complex software like magento and a busy site, Disk IO is really going to hamper you. and cause misleading High Load issues (which are down to IO related issues). This leads me nicely onto…


Support is again a mixed bag, generally they are helpful, polite, there english is passable (they have a ukraine based support team), however I’m often left with the impression that they are programmed to first blame the issue on lack of nodes then take a look when that claim is refuted. I can recount several instances where I was told that more nodes were needed when I knew without a doubt this wasnt the case (I knew because i had identically setup servers (on the software side) running in different locations with less resources without issues. I can also remember many times when a tech has blanket asked for the root password when I have asked them to perform a task that does not require root access – so if your anal about server access then be aware of this.

The negatives of support are: sometimes tech’s don’t read whole ticket and understand the issue if being assigned to the ticket part way through, some of the support tech’s english is bad enough to sometimes be incomprehensible.

There managed suppport does leave a lot to be desired, I’ve found it via other clients to be “less than proactive.” They do an ok job at monitoring the server for downtime and resolving it, but they dont really proactively update servers (unless it is a major very public security breach to be patched) and for sure they dont proactively monitor your server for attacks (unless the attacks take the server down) so I’ve found it expensive for what you get. I have heard though that this is being looked at as part of there wider support team setup which hopefully will render a positive outcome.

I’m going to roll in communications into this point, communications for me still lacks when there are issues, it has improved so it deserves some credit but I think has spread itself a little too thinly given the percieved resources available on this front. They are absolutely brilliant when it comes to giveaways and when things are going well but I still often get the impression they hide in there shell when things go wrong. The status pages however are a massive improvement over 12 months ago, it’s actually very helpful to see things reported there because its a consistent place to track for issues arising and gives at least some info I can pass over to clients and manage there expectations. However do have some more room to improve here.

Positively though, and I’m quite excited to see how this pans out, is that are rolling out Phone support (and they dont seem to be outsourcing it to script monkeys in india either!). This could be a blessing or a curse to them, I hope its the former and I hope that it resolves many issues faster and ultimately leads to happier customers.


I mentioned at the start of this article that I don’t personally actively use on a day to day basis, I do however use it to roll up test vm’s to test out ideas and particularly test out puppet deployment scripts as this is where the service really shines, its ultimately very very flexible to meet peoples needs on a sliding resources scale.

I think the service is fine for small or medium size indivudual sites or companies that are probably not solely reliant on there website for revenue, its definitely not for people that dont know about servers (you want cloud or geohosting if you fit into that category, or to hire a pro!) and I would argue its not for clients with mission critical needs or high turn over e-commerce sites unless you are prepared to pay significantly more to build some kind of HA infrastructure on the platform.

At this stage, I’m not recommending, but I am not recomending against it either, I do hope the next few months will push me back to the former though