There are many similarities between hosts, so what are we doing differently?
2020 was a busy year for LEOPARD.host – we overhauled our Hosting and Server platforms from the ground up, including a change to some of our core networking vendors too. This wasn’t just cosmetic, we researched options extensively, performed harsh testing and made support teams prove themselves before making any decisions and implementing them across our PoPs.
With so many layers involved (across compute & networking), it’s important to find harmony.
It would’ve been “fine” to improve our systems to a small degree and keep them reliable as we always have, though in our industry there’s something to be said for planning more than 6 months ahead as many hosts seem to do. We went back to the drawing board, evaluated the systems we’d been depending on, and how we could be improving our platforms to better serve our customers.
New systems: Hypervisor (bye VMware), JetBackup 5.x (from 4.x) & more.
We moved all of our corporate and customer host nodes away from VMware to a more KVM-friendly option, enabling more control and flexibility for our Operations team as well as our valued customers. That removed many restrictions from VMware, such as the limiting busybox shell that they provide with their hypervisor. Virtualisation now occurs on top of a popular Linux distribution with regular packages in-use, giving us the option to expand into offering container-based packages to our customers too.
JetBackup made some bad calls with their software up to the 4.x branch, and eventually worked out that they’d made such bad decisions that they had to re-write their codebase from scratch and force their partners to navigate an unpleasant upgrade path. We’ve finalised that work and are now in the transition period where new backups are taken under 5.x before we can safely remove 4.x. It’s great that they recognised their failings, though we hope they’ve learnt their lessons and won’t require such a messy process again.
More focus on security: Data centre relocation (bye Equinix), network rebuild & even more.
It wasn’t enjoyable, however relocating our primary PoP from Equinix SY4 to NEXTDC S2 made plenty of sense. Equinix had given us several causes for concern, and through the process of trying to remediate those we were given more reasons to worry. NEXTDC have focused on security, convenience and service – a magic combination that’s been serving us well for the past half a year.
We spent a long while in Equinix SY4 after our time in their SY3 facility. In touring their newly-built SY5 facility, it’s clear that they’ve learnt several important lessons, however despite their scale we had too many reasons to steer clear of them. We’re hopeful that they’ll continue to listen to feedback and improve, though our customers are too important to wait it out with them. NEXTDC have been tremendously good since our migration, and we have no plans to leave their ecosystem.
Our network has had some major changes too – we changed our switching vendor and have begun preparations for further network expansion later on this year (or in early 2022, yet to be finalised). Throughput potential is exponentially higher than it was beforehand, and we’ve enjoyed more capable firmware on the new hardware too. Final tweaks on this front are being completed this month (May 2021) so that we can apply some final security improvements, to then move on to providing more convenience and automation for you.
Eager to learn more about these major works? Chat with Tech Support!
