Beautiful summertime is passing by and due to closed borders members of the flespi team are forced to stay in Belarus during the traditional vacation period. This is not so bad on the other hand — our nature is great and COVID19 gives us a chance to discover it.
But as a good engineering team, we use a lot of load balancing and high availability techniques even for vacations. Part of our team is working from home, another part is already back to the office, and yet another part from the countryside which is available almost immediately for any emergency case.
Vacation time and concentration mostly on internal things like refactoring, optimization, performance tuning, and so on allowed us to finish July with perfect 100% uptime in both regions. The second time in a row and this is a great achievement. In general, if you will look at the chart with uptime history for the last 12 months you may notice that once we launched the RU region and finished all hard reconfiguration of the services architecture including networking, everything started operating very smoothly. And I hope we will keep up at this extremely high level.
- As I already mentioned, in July we were mostly concentrated on internal things and released not too many features. For those who are using flespi analytics, there are a couple of new features: the possibility to limit the maximum duration of the active interval to regenerate a new one and the possibility to control routes.
- Also, we are working on the replacement of the REST API service that is a gate to all our database mirrors — it is called mdbctl on this page. The replacement service is rewritten from scratch — it is faster, better, more efficient, and will work in parallel to the old one for some time. This is a 4-month development work and we are very close to launching it in the production. Our target is to make the launch absolutely invisible to our users. By the way, it is the best practice of IT nowadays — to deliver a new fresh replacement of a component seamlessly and with zero effect on the users.
- The same applies to our 2-year work on the replacement of our protocol parsing and devices communication engine — the pvm. A few months ago we installed its first version and started big work on transitioning all existing protocols to the new engine. 29 out of 64 protocols have already migrated and we had only one problem so far noticed by our users — with settings control for the concox protocol. Immediately after that to prevent similar cases in the future we started to notify about the protocol transition to the new engine with a corresponding changelog record for each protocol.
- We published only three blog articles in July but each of them is quite an interesting reading. One is about calculating storage used by the device in day-to-day operations to estimate your possible flespi costs. Another is related to scooter sharing projects experience — this article appeared as our reaction to the boom of scooter sharing projects we had recently. And the last one was authored by our user — how to combine video telematics, AI-based face recognition, and deliver all these into the 3rd party analytical application. All these articles are a great read and may inspire you to new ideas in the telematics field.
In August we will continue working on all our internal projects, let other parts of our team relax on vacations and one interesting project we just started to work on is a catastrophic datacenter survival. For each region separately we plan to launch an additional datacenter that will operate as a distributed RAID-1 system with a full copy of services, data, and even IP addresses. It means that even if we lose our datacenter for some reason with all racks and servers in them, we will still be able to provide smooth uninterrupted operation as the whole load will be mitigated by another datacenter that will operate in parallel. Our target is zero seconds downtime in such a case and we have some thoughts on how to achieve this. We plan to launch such an installation next year in the EU region and after a certain time and growth, we will reproduce this in the RU region. Datacenter survival architecture will be an integral part of flespi and will be available to all our users automatically without additional charges.
Enjoy the end of summer and you are welcome to contact us in the chat (top right corner of flespi.io) for any discussion.