Historically, nine built their own VM management solution around open source tools. While this was sufficient for many years, it was time to revisit this decision.
While looking at the pros and cons of different solutions we came to the conclusion that we want to spend more time on engineering services and supporting our customers in running their workloads than spending time on the maintenance of our infrastructure.
At this point, we made a decision to use Nutanix as our future partner and platform of choice to deliver our clients the next generation of virtual machines that is based on a hyper-converged infrastructure.
What does this mean for our current and future customers? What can you expect regarding performance and scalability?
How do customers benefit from this new infrastructure?
Besides significantly higher processing speed, each looking at CPU, memory and disk performance, customers do also benefit from a more resilient platform where single node failures do not result in major outages.
With the new platform we provide way more flexible and customizable virtual machines. CPU, memory and disk requirements can be adjusted in small steps to reflect your technical and financial needs. We’re happy to support you to find the ideal solution for your application and budget. As the whole Nutanix project simultaneously acts as lifecycle management for our current infrastructure, existing customers will benefit from this solution as well.
Our plans to migrate your workload
We spent some time to engineer and properly test a migration path for existing virtual machines that requires no form of customer interaction. All customer data and databases as well as all settings will be migrated, there will be no change in software versions and therefore no need for extensive testing on your side.
Once this migration is started, your virtual machine will still be available and accessible as always. After the background synchronization has finished we stop all services on the old platform and start your virtual machine on our new Nutanix platform. This is where a short downtime of approximately 3 to 5 minutes will happen.
As we are writing this blog post, a dozen early adopters already benefit from the new solution.
Did you login to our cockpit at https://cockpit.nine.ch lately? In fact, this was one of the first customer facing applications of our own we migrated. Customers who are eligible for migrations will be contacted in the next days and weeks.
What does eligible mean?
Our first Nutanix cluster is covering one of our two data centers, therefore all managed virtual server customers that currently run their workload in this datacenter are eligible and will be contacted soon. Currently, only (new or existing) managed v-server customers in this datacenter are eligible for a migration.
Of course we will also implement this new solution for our other datacenter site, though this will need some more time and is planned for a later stage.
How do I read the performance benchmarks?
When talking about superior performance, we made sure to have some numbers to reflect this. But please keep in mind that benchmarks are not real world examples, still they show the (huge) advantage for the new platform and it’s capabilities. All CPU, memory and OLTP benchmark results below were generated by using “sysbench”, a Linux system benchmarking utility. All tests have been performed on a system with 4 CPU cores and 3 GB of memory.
Better Memory Performance
Memory throughput is especially important when handling huge datasets and working with databases. We see a 25% increase here with the new platform.
Faster CPU Processing
Not only the memory receives a performance boost: the multi-core CPU benchmark below shows huge multi-core processing improvements. Raw Power!
The latency of the single-core benchmark is revealing how fast the CPU can process the requests that we generated. Lower is better!
Besides multi core processing, there are still a lot of use cases where the single core performance is also relevant, for example single huge database queries or your nightly cronjobs. 31% more events could have been processed during this single-core cpu test.
Hybrid high-performance SSD/HDD storage
Huge improvements can be seen in the latency and the throughput of the disk i/o subsystem. A OLTP test shows us a combined view of all areas in a more real world benchmark: the workload of a typical mysql server setup combines CPU, Memory and Disk I/O performance.
If you have any questions about the new environment for your products, do not hesitate to contact us. We will inform you upfront about the migration of your managed vServer and keep you informed about future improvements of our infrastructure.
You can as well expect more content on our blog regarding the new platform in the next weeks.