We're still here

Despite being off for a while and getting locked out of my password manager its time to post an update. 

As you would have gathered from the last post the farm was off from November 2024 due to the heat. Now we're in a new year and its still summer here in Sydney so still hot. We had a couple of days of cooler weather a fortnight ago so I had the Pis running for two days and then this week we had a severe thunderstorm and strong winds but it did cool things down a bit to be able to get the Pis back up and running again.

Marks Rpi Cluster is back to running Einstein work at the moment. I did try a burst of Asteroids but they had an app update to 102.21 that apparently missed out on optimizations so they take longer than the previous 102.20 app. Until they fix that I am not going to run any Asteroids work on the Pis.

In the last couple of weeks the Raspberry Pi foundation released a 16GB model of the Pi5 for 120 USD. They are the same speed as the other models. I don't need the extra memory for running BOINC projects - the 4GB model is enough for that. They might be useful if you use it as a NAS.

Current cluster setup:
24 x Pi4 (8GB) as compute nodes
4 x Pi5 (4GB) as compute nodes
2 x Pi5 (4GB) as support nodes

 

Marks Rpi Cluster is off at the moment. Its hot today so everything is off. As usual it has been running the Einstein@home BRP4 app, mostly on the Pi5's with the occasional burst from the Pi4's when its cooler. Its reached 105M credits for Einstein so far.

The cluster currently consists of:

24 x Pi4 8GB as compute nodes
4 x Pi5 4GB as compute nodes
2 x Pi4 2GB as support nodes
1 x Pi5 4GB as support node

I haven't heard from BitScope as to if/when the Edge Cluster 12 will be updated to support the Pi5.

 

 

8th of September

Marks Rpi Cluster is on and off. As I type this its on. We had a few weeks of hot weather where I had the Pi's idle and when it gets a bit cooler have them going again.

I have reduced the cache on all of the Pi's so that I can finish off what work they have quickly. With the reduced cache they will fetch additional work only when they are about to complete the 4 tasks they have at the time.

In other news I got a few of these:

They are ASUS C2500's. They are a USB type A to RJ45 network adapter that gives 2.5GbE network connectivity. This allows me to improve the network speed on the support Pi's. I have a 2.5GbE switch on the way as well.

18th of August

Marks Rpi Cluster continues to run 24/7. The last few weeks saw two updates that interrupted processing. When applying updates that require a reboot I finish off the onboard work, apply updates and reboot. The Pi4's can take up to 3 and a half hours to finish off their work.

As usual we're running Einstein BRP4 work after I let the Pi5's have a week running Asteroids work. The Asteroids app seems to run fairly quick on the Pi5 relative to the Pi4's.

The cluster consists of:

4 x Pi5 (4GB) as compute nodes
1 x Pi5 (4GB) as support node
24 x Pi4 (8GB) as compute nodes
1 x Pi4 (2GB) as support node

I will likely swap out the Pi4 support node with a Pi5. It doesn't need the extra processing power as its a time server with a GPS, but all the Pi's hit it regularly so there is a fair bit of network traffic.