50%+ of nodes can now be remote rebooted freely.
Documentation and software work on that is on-going, but the hardware is now fully in place and scaleable.
To fully document this, most of the nodes needs to be hard rebooted once. Approximately 12.5% of total. This will happen gradually over time, and we should only need to do it once.
Older platforms will be upgraded by end of Q1/2024 to the platform with hardware remote reboot capabilities, they are missing that and they require a full platform physical upgrade to do that. Same servers, just the infrastructure hardware around them needs to be replaced.
---
In other news, we received a batch of nodes capable of supporting 2x NVMe drives. We will try to get some of these online by end of November. We received almost 100 new nodes last week and work is starting to get them online.
We are currently also researching NVMe drive cooling options, and have some new parts in order which should allow us to install cooling for all NVMe drives in fraction of the price and cost as before. Prior versions it took quite a bit of time to install the cooler, and the coolers were rather expensive. Next version cooler install is almost neglible time, and almost neglible cost. Please let us know if you are monitoring your NVMe drive temperatures, or if you have seen temperature warnings. Same goes for CPU temps.
That being said, we are yet to see a single case of throttling or beyond warning level temps. There is a NVMe drive model which has over 10C lower warning temperature threshold than others and we've had 1 report of warning threshold being reached. We are going to upgrade those first.
Most amazingly, we are expecting to more than double the total number of MD series nodes by end of the year. Yes, more than double.
Finally on hardware side of things, we are almost over with version 2 platforms, now at version "2+" (manually doing some changes) and shortly all new nodes should be on platform version 3 which simplifies manufacturing these quite a bit, estimated time savings per node is significant. We are currently doing manual fabrication work to upgrade version 2 to "version 2+", drilling, tapping etc. but the next batch to be ordered will have the fabrication shop robots do these steps, along other upgrades in thermal management.
---
Software has been roadmapped quite decently, from end to end. Progress has started on that, but we think visible changes to end user will start happening around December. Our primary focus is still on the backend, making sure we can efficiently manage these with least possible time consumption. Less human effort involved means lower price for Your Server.
There is a rescue mode already if you need it, and you can install any distro via rescue mode.
יום שלישי, אוקטובר 10, 2023
