The business side

As we approach coming out of beta, we've had a review of the industry and in order to ensure we are competitive both in terms of specifications and prices we're adjusting our offering. From the 1st of March we'll start charging the same prices as we do now, but in US dollars! Due to the way invoicing works, you may not see the changes happen immediately on the 1st, but all instances will be charged at the new prices from that date when the site switchover happens in early March.

However, to be more inline with other cloud hosting providers, the specification of instances will change too.

X Small : 1 Core / 1GB RAM / 25GB Storage - $5/month

Small : 1 Core / 2GB RAM / 25GB Storage - $10/month

Medium : 2 Core / 4GB RAM / 50GB Storage - $20/month

Large : 4 Core / 8GB RAM / 100GB Storage - $40/month

X Large : 6 Core / 16GB RAM / 150GB Storage - $80/month

And we're also launching a brand new size:

2X Large : 8 Core / 32GB RAM / 200GB Storage - $160/month

Previously our instances were balanced with the number of cores equal to the number of RAM, but from speaking with some of our customers it's more important that the servers have more RAM than cores. So this new balance formula has enabled us to buy new equipment and charge less for the instance sizes.

This has never really been a problem for our customers (the cost/competitiveness) but as we look to exit our beta phase and launch we wanted to make sure we were launching with the right offering.

The technical side

Previously we've used a variety of CPUs with the oldest generation being Broadwell-based. Therefore up to now, we've set all instances' virtual CPU to be "Intel Core Processor (Broadwell)". Customers would still get the speed of the newer generation chips that the instances are generally running on, but it didn't look good (as well as potentially removing some of the more advanced CPU features).

So, as all instances launched from the 1st of March will be launched on our new compute platform - they'll all be seen as a minimum of a Skylake architecture (as we run Intel Xeon Gold 6154 CPUs running at 3.0GHz).

This fixes one aspect that customers mention a lot, but the other aspect of how we host revolves around where the disk images are stored. The default setup of OpenStack for most companies revolves around using Nova as the compute platform with Ceph as the disk/volume storage cluster. Ceph is REALLY good, we're not taking anything away from it - upgrades have generally been smooth, it's pretty performant and with it storing everything in triplicate it's generally really safe.

However... when Ceph loses a disk, it can go in to a bit of a meltdown as it tries to rebalance everything. Sometimes, one individual disk can get full and again it goes in to a meltdown, potentially corrupting disks and this involves recovery as well as trying to tweak the balancing settings.

So, we've experimented with using NetApp's SolidFire all-flash storage for clouds and have found it to be really good. Highly reliable, great performance, easy management, great support, etc. It still keeps three copies of every bit of data and closely monitors the cluster to ensure the data is always consistent. Pretty much everything we'd want from a storage platform. Our sister company Serverchoice has been using it for a while in production and they've been raving about it!

From posts we've made on the OpenStack mailing list we seem to be the only hosting company using SolidFire to store instances' root disks, rather than just attached volumes. This took some figuring out and lots of playing, but we have it working perfectly now - instances launch in the same speed they did before, feel much more performant in usage and snapshots are lightning fast now!

So now our platform is ready, we've got some tweaks in the code of our site to make to be ready for non-beta launch, but it feels like we're really close now.

Wish us luck and please send us any feedback about the new offering, we'd love to hear from you!