Folding and HP Cloud Service

[toTOW]

As some of you may know, HP has run FAH as a burn test for their Cloud solution : http://folding.extremeoverclocking.com/ ... =&u=601623 which has shed light on the private beta testing of their service.

I’ve been following this beta testing with HP representatives and it looks like they had a great response from the folding community resulting in a significant number of applications to their beta program.

Unfortunately, this situation went out of control, and folder usage is stealing too much resources from other use case of the cloud system.

As a result, HP will no longer accept applications from folders until their service exit public beta and they start charging for it. Please spread the word to the community at your respective team forums.

HP is also reducing the resources allowed to their beta tester on February, 27th. If you are running folding clients on the cloud, please exit it nicely with –oneunit flag so that the FAH project won’t suffer from WU losses before terminating your instances.

Thanks for your understanding.

[zodac]

Got a link to where HP said they won't be giving access to Folders?

[DR650SE]

If HPCS is allocating a specified number of resources to each tester (20GB RAM limit) , can't they use those resources however they deem fit? I mean they are meant to be used right? Just trying to get a handle on the situation before I do anything. I know a few individuals who in the private beta application specifically stated they would be running F@h and were accepted for the testing.

Hopefully they will allow folding. it would be nice to get a lot of WU's done for the science.

[Jesse_V]

If HPCS is allocating a specified number of resources to each tester (20GB RAM limit) , can't they use those resources however they deem fit? I mean they are meant to be used right? Just trying to get a handle on the situation before I do anything. I know a few individuals who in the private beta application specifically stated they would be running F@h and were accepted for the testing.

Hopefully they will allow folding. it would be nice to get a lot of WU's done for the science.

Right. I totally agree that running F@h on HPCS is a fantastic idea. Just look at their productivity! I heard there were violations of both F@h's and HPCS's EULAs that resulted in this occurring. Once HP became aware of this, they tried to stop it. HP is gambling that their testers will use a certain amount of resources (and they probably try to cut costs by sharing) and F@h would probably be more CPU intensive than they anticipated. Sounds like HP doesn't explicitly deny F@h, and as stated they have run it on their cloud. It is a free service, and I think it should be respected. Use it fairly and legally. When things get out of control and become unethical, this sort of action makes sense.

[ChasR]

I got my acceptance email this week. It states the following:

We are very pleased to welcome you into the HP Cloud Services private beta, free for a limited time. This is your chance to experience and test HP Cloud Compute and HP Cloud Object Storage. We look forward to your feedback.

Please be aware that we are requiring all private beta customers to:

1. Limit your HP Cloud Compute usage to 20GB of RAM total for all instances within an availability zone. Multiple availability zones are available

2. Limit HP Cloud Object Storage usage to 2TB of total data across the account

3. Only use the private beta for testing purposes. No production workloads should be hosted on the system at this time


We are implementing these limits during the free private beta to accommodate a wider variety of private beta customers. Greater diversity in our private beta phase will help us plan for the massive scalability we intend to offer at public beta.

Folding at home would likely be considered a "production workload" and contrary to Item 3 of the acceptance letter.

[zodac]

That's the same email that was sent out weeks ago; points #3 was included, but when in chat with one of their tech guys, several of my Folders asked if Folding was acceptable, and they said it was.

So I don't think that email is proof enough that they'll no longer grant access to Folders.

[7im]

...
So I don't think that email is proof enough that they'll no longer grant access to Folders.

Correct. Not proof either way. And a comment from "tech guys" is not proof they will or won't accept access to new folders either.

Let's wait and see how it plays out. HP could publish an additional statement, or they might end the free beta tomorrow, or something in between.

[DR650SE]

I guess just to play devils advocate I'll throw this out there. They are using individuals as beta testers. One way to test the system would be to use the resources. In our case it would be testing the CPU's to ensure that they are stable.

I just hopped on chat on the HPCloud site, and asked flat out if running F@h was acceptable. I'm sure I can't post private correspondence here, but I'll sum it up.

I asked if F@h was an appropriate way to test the system for stability. Answer was Yes it is fine to run F@h. The individual stated that there are many people running folding@home on the system, and it's perfectly fine.

I then asked if it was ok to run it, given the fact that it's CPU intensive, and could be looked at as hogging resources, and stated that I didn't want to do anything that would result in my termination from the beta testing program. I was told that it was fine, so long as I'm not doing anything illegal, I will be fine. He also stated that it would have been specifically stated in the documentation if F@h was not to be run on the system.

That corresondence summed it up for me

[7im]

Unless your chat session was with the CEO of HP, or the director of that HP Cloud site, I question whether the "tech" you chatted with has the authority to make that call or not. The IT geeks will respond, Ya, burn it up. We can take it." While IT Management would respond differently, like sending out that email ChasR quote.

I wonder what the chat response was to having multiple HPCloud accounts.

[Punchy]

What upsets me even more than the folks with multiple accounts running as many as 160 instances is the folks from the early beta stages that did not terminate their 8p instances and are simply going to keep them running until HP terminates them. Of course this means those WUs will be lost, time out, get resent, and delay the work.

[MtM]

What upsets me even more than the folks with multiple accounts running as many as 160 instances is the folks from the early beta stages that did not terminate their 8p instances and are simply going to keep them running until HP terminates them. Of course this means those WUs will be lost, time out, get resent, and delay the work.

I'm not bothered with people using 160 free accounts if those accounts all would be in line with eula and would run like normal clients ( meaning not dumping wu's when the resources dry up ).

If people have 160 4 core instances, and HP does not contact their users directly or release a public press release, I can't say they have to stop. That changes the moment f@h officially posts these instances are causing troubles or hp makes public this usage falls outside the scope of the free beta trail.

I did already paste the OP with a link to this thread on my team's forum, but I must say I haven't seen any point drop on those I know to be using hp cloud.

[toTOW]

I don't know why we are debating whether FAH is allowed on HP cloud or not ... No one ever said it was forbidden ...

I said that FAH is one use case of the system that has been valided by enough people, which explain why they won't accept more folders. Don't forget that a cloud system is based on shared resources, so if one use case has been validated, they need to save resources for other use cases.

Multi accounts and people with 100+ instances is another story ... and only HP can decide what to do with these accounts.

[MtM]

HP is also reducing the resources allowed to their beta tester on February, 27th. If you are running folding clients on the cloud, please exit it nicely with –oneunit flag so that the FAH project won’t suffer from WU losses before terminating your instances.

Well you have to admit this wasn't that clear, at least not to me and I guess others as well. Left room for interpretation, saying 'please exit it nicely..' could very well be interpreted as saying: you have to stop your instances.

[toTOW]

Those who are involved in the beta test know what it is about. For the others, here is the explanation :

- with original limitations, people where able to run two 8 cores instances and one 4 cores intance before reaching limitations.
- with current limitations, the 8 cores instances are no longer available, so to comply with the new rules, they should be stopped and replaced by something else. That's in this case that people need to shut down their clients nicely.

It is also apply to those who won't want to continue running instances with the new limits ...

[MtM]

Thanks for adding the explanation, I'll add this addendum to the repost I made on the team forum.