A major change in our VPS/Cloud hosting infrastructure is in the making for the past couple of months. The time to show what is behind the curtains has come and it is really exciting!
If you have been a TMDHosting Customer for a while now, you probably know that we are utilizing the OpenVZ technology for our VPS and Cloud services needs. In its very core, OpenVZ is a fascinating technology that allows for streamlined management and relatively easy networking setup. This was our major choice of virtualization technology back in the day because it also allowed for great automation and hardware node based scalability.
Over time we have discovered some mishaps (beancounters, we salute you!) in the OpenVZ architecture but managed to go around them with great success. We even made it around the invasive disk I/O that OpenVZ is known to spawn at random intervals – an achievement unlocked by a very few of our fellow operations team in the industry. At a certain point we have completely tamed the OpenVZ technology and had abilities to model it for our Customers needs. Still, it did not allow us to pool idle CPU resources nor flexibly allocate RAM.
As time passed by, we began looking to utilize the idle resources we have and we felt like we exceeded what OpenVZ has to offer for our needs. From long term and at a large scale perspective, flexibility and scalability in OpenVZ came at a huge resource waste cost, making it somewhat inefficient for our infrastructure’s size, intense usage and power demands.
Today we are pleased to announce that for the sake of better resources utilization and therefore for the greater good of our Customers we have decided to switch from OpenVZ to KVM.
Technically speaking, this is a huge leap from a monolithic virtual infrastructure to a real, flexible and public cloud. We have built a new polymorphic cloud infrastructure, where we can utilize all the available resources in the pool to their greatest extent across the virtual machines or cloud instances lying beneath.
If we take the page load time as a performance metering factor, this change alone will bring at least 100% of performance increase, by simply putting idle resources to utilization. On a scale from 0 (not existent) to 10 (best performance), here are our OpenVZ vs KVM measurements:
Because 100% increase is just not enough for our firm beliefs that we must provide the most stable and lightning fast web hosting services, we have built this infrastructure on top of completely redundant and all-SSD based storage.
Simply put, it matters a lot how fast your data (website, store, forum, etc.) is being read from the storage. By shortening that time drastically, we shorten the time your page loads. A lot of technical and not-so explanations exists, but this one sums them all. Here is a sample test, measured in IOPS (input/output operations per second) on how HDD vs SSD compares:
These measures combined will bring more than 300% of performance boost on stage at a completely no price change.
We have divided the polymorphic infrastructure we have built into couple of different segments, allowing us to pool resources and divide them flexibly between instances. We have a pool of network resources, pool for computing and pool for storing your data.
By combining them altogether, we end up with an instance (be it a VPS or cloud container) that gets precise amount of allocated resources – no more urgent bursting into imaginary resource pools and unwanted hardware overcommitment. The moment you need more resources, an intelligent service scales them automagically your instance, helping you face the slashdotting caused by your recent blog post, breaking news or product promotion.
This is a huge change in our infrastructure and we have planned to migrate our existing VPS and cloud instances to it first, starting next week. We will be covering various aspects of the infrastructure upgrade, such as hardware configurations, networking setup, the process of migration and other advantages of the new platform via our blog. A precise schedule for VPS and cloud migrations is currently being prepared by our operations team and every Customer that is set to be migrated will be personally notified.
Stay tuned for more information regarding this project surfacing via our blog! Wonderful and exciting 2016 is ahead!