New Boron cloud release from Pan-Net up and running in Salzburg and Vienna!
Mar 5, 2021 · 610 words · 3 minute read
We are happy to announce that we have successfully upgraded our cloud in Bijelo Polje, Podgorica, Budapest and Salzburg to Boron and deployed a new Boron datacenter in Vienna, so we have production applications already successfully running on the new infrastructure version!
Boron is a new Common Infrastructure Team (CIT) cloud release from Pan-Net, designed for both IT and NT workloads – dedicated compute with carrier grade acceleration for NT and cost-effective scalable shared compute for IT. Boron is based on OpenStack version Stein.
It offers many benefits, among others focus on carrier-grade stability, improved customer experience and efficient cost-effective powerful hardware 2.0 based on AMD processors that enable high performance and flexibility. There is also a possibility to choose your level of control and involvement with our flexible operational models, from fully managed model to franchise, where customers take over cloud operations.
Naturally, we put stability as one of the main goals in Boron target design. We reworked the architecture to detect and remove known issues from previous releases, changing the components which are proven to be the main sources of operational issues in the past.
We have gained a lot of experience based on lessons learned in daily operations and troubleshooting during the last years and managed to significantly improve our expertise not only in technology itself, but also in operational processes and collaboration with our suppliers.
In the end, we identified the main pain points of the previous architecture and implemented changes in the main functional blocks to address the application requirements, KPIs and operational issues. Telco specific KPIs were developed together with our main application vendors to provide carrier grade operations.
Flexibility for wide variety of Telekom workloads is provided by the new hardware that is optimized for specific targets, such as NFV compute, IT compute, performance storage nodes, or networking nodes:
Everyday computing, e. g. web hosting, publishing, file storage, etc.
High-performance computing and networking, e. g. data planes of virtual network functions, transcoding for streaming, etc.
High memory workloads, such as large in-memory sessions caching for virtual network functions, large in-memory data Bases, in-memory analytics, etc.
As far as Enhanced Acceptance criteria are concerned, we have to mention the following features:
Smooth maintenance and advanced upgrade procedure: Additional test cases focused on maintenance procedure impact on all components. Improved upgrade procedure significantly reducing impact on applications by using live migration.
Focus on failover scenarios: Ensuring HA/clustered set up for all components and test failover procedures in complex environments.
Real-life scenarios: We introduce testing scenarios closer to real telco workloads, using “tuned” applications with carrier grade VNF features
It is also necessary to focus on the Enhanced NFVI testing, because we have implemented advanced set of touchstone test suits to be part of continuous testing process.
API testing: A set of API tests was developed to ensure key API functionality, stability, and performance. We focus on stable Nova, Glance and Heat responses while ensuring that new SDN controller API (Neutron API) performance meets our KPIs.
Networking performance: Set of tests on L2 and L3 covering multiple contexts using different available acceleration technologies to find the best design for different scenarios (standard VMs and carrier-grade optimized VMs) ensuring stable performance.
Storage performance: To benchmark storage performance and make sure, that we will have stable improvements in Boron, CIT KPIs are used for storage as well as peak performance measuring for Beryllium. We are focused on reaching stable higher performance in the new release.
Enhanced tests are used to validate and prove our design and configuration decisions during the development to make sure the code will reach production only when improvements are stable and proven by tests.