Most of the times I manage to run two NBs per Google account, but I also get (randomly) the message "Too many sessions", and then I can only run one.
I am currently running CPU-only NBs on free-tier accounts (using a total of 5).

Hmmm... That's interesting. I'm currently running 3 concurrent CPU-only notebooks under a single account (free-tier).

Maybe it is looking at the total number of sessions originating from a single IP, regardless of the accounts?!

What is the maximum number of concurrent sessions anyone has ever attempted?

[QUOTE=axn;561687]

Maybe it is looking at the total number of sessions originating from a single IP, regardless of the accounts?!

[/QUOTE]

I guess that´s the case, yes. I may have as much as 10 sessions (2 per account) running concurrently from the same computer, when the gods are in a good mood.

Also, as the Colab VMs have two cores available, I don´t think there´s big value in running more than 2 CPU intensive NBs per account. No hard figures for that, tho.

[QUOTE=lycorn;561689]Also, as the Colab VMs have two cores available, I don´t think there´s big value in running more than 2 CPU intensive NBs per account. No hard figures for that, tho.[/QUOTE]

Negative. Each NB is getting its own VM.

It appears that compared to one core, you don't get very much from two cpu cores on Colab. For my instance, running two workers gives a ~7% gain relative to one worker. It's still a "free" gain for LL/PRP, but as of P-1 the gain will be at least somewhat cancelled out by the fact that each worker has less memory to use in Stage 2.

[QUOTE=axn;561748]Negative. Each NB is getting its own VM.[/QUOTE]

Ah, ok then. I supposed each [B]google account[/B] would get a VM, and then all NBs fired up under that account were to share the VM (2 CPU cores + mem) resources. I will give it a try someday. Thanks for your clarification.

[QUOTE=Ensigm;561791]It appears that compared to one core, you don't get very much from two cpu cores on Colab. For my instance, running two workers gives a ~7% gain relative to one worker. It's still a "free" gain for LL/PRP, but as of P-1 the gain will be at least somewhat cancelled out by the fact that each worker has less memory to use in Stage 2.[/QUOTE]

Colab VMs don't have two cores - you just have one hyperthreaded core (i.e. 1c/2t). Therefore, you should use 1 worker 1 thread, and if you want to see if hyperthreading helps, enable the option of "use hyperthreading for LL" (please lookup the correct option for this setting), which will use both threads for a single test. That means you get the benefit for P-1 as well.

[QUOTE=lycorn;561838]Ah, ok then. I supposed each [B]google account[/B] would get a VM[/QUOTE]

So did I, until a couple of weeks back I tried to run another NB in parallel (just for the heck of it) and discovered the fact. In fact, currently running 4 in parallel, all in their own VM, all under a single free tier account.

I don't know how far I'm gonna push my luck, though.

I have never been able to get 2 to run in parallel on a single account (on purpose).

[QUOTE=axn;561840]Colab VMs don't have two cores - you just have one hyperthreaded core (i.e. 1c/2t). Therefore, you should use 1 worker 1 thread, and if you want to see if hyperthreading helps, enable the option of "use hyperthreading for LL" (please lookup the correct option for this setting), which will use both threads for a single test. That means you get the benefit for P-1 as well.[/QUOTE]
Thank you for the suggestion. The option is [C]HyperthreadLL=1[/C]. I should be able to see if there is any improvement in an hour.