amuck-landowner

GVH suspended after minor use of 100TB plan

jarland

The ocean is digital
Sure, but a normal host should not be monitoring what processes you are running unless you get an abuse report.

I guess you can think of it this way, for someone buying a 4GB RAM server, it should be able to handle 3GB of RAM, so my usage to test the RAM seems legitimate.  For someone who buys a server with 250GB of space and tries to transfer over all their files from another host would need to transfer at 4MB/s for 18 hours in order to start using the service, so downloading a 100GB file at 4MB/s seems legitimate to me.
Situational, but a host should absolutely be monitoring processes. Not like snooping on a VPS but when you get an alert, which could be from anything not necessarily you, you start looking at what is slamming the node. If you're up there, damn right they're going to see the process on OpenVZ. You're not hidden from top and iotop, your processes are right there and they need to know what is running hard when the node is tanking. If you fit the bill as a part of the problem, expect to hear it, unless you're on a provider that doesn't care. Some may even choose not to talk to you about the specifics, but rest assured on OpenVZ...they see them.
 
Last edited by a moderator:

drmike

100% Tier-1 Gogent
To be fair, there's a difference between utilizing your VPS and abusing resources for the sake of abusing resources. I've terminated and refunded many people who wget cron test files to fill up disk space then rm -rf it, or consume large amount of network for no valid reason.


If you're going to use your resources, use them properly. Don't be a dick. I will say that 100tb plans are practically begging for people to do that though.
^--- that's spot on.  

You of course know I am rabble-rouser  and recommended certain tactics to utilize your paid for resources (i.e. reserving disk for later and the 100TB challenge).  Fine line in areas between testing and just setting something on fire to watch it burn.  

Important on the provider side to have a level and somewhat slow to smack hand as a provider.  There are lots of creative folks, some that are just being on poor behavior before nap time and others that didn't read the ToS or slid around them.  Sometimes even the best management and tools snags a false positive.    That's possibly what may have happened with the OP.

When you sell the plans that we do, it'll be more beneficial to have extra RAM and extra cores. Primarily for the extra RAM though, we'd prefer the Dual L5420.
54xx series is capable, but that's not suitable for these plans at this point in time.  Shared hosting, cPanel, yeppers fine.   Hosting for local/regional/real companies not in the active tech space, suitable.   Pitting a 54xx against the 56xx and newer stuff, meh, maybe if you are getting extra low pricing or buying a rack full for a song.  Not advised though.

 ColoCrossing is currently having issues with space in their facilities and we're waiting on them to get the things that we need to get done.
Officially, I've heard from customers, 3+ weeks wait time for CC servers.  No one knows why.  Power?  Liquidity tapped?   This is the hazard of renting and why businesses tend not to do rentals like folks in the lowend* market do.   If rentals, businesses pre-buy, pre-rack and are paying months ahead of deploying.   Just-in-time style fulfillment of rentals stinks other than for limited cash flow stretching.
 

GVH-Jon

Banned
Jade,


We're not going to be using RAM as an excuse for more overselling. If you look at the plans we sell and do the math, the extra RAM capability of the Dual L5420 is needed to maintain a low RAM overselling ratio. I understand they use up a lot of power, but that's not our main concern. Our main concern is our customers.
 

SkylarM

Well-Known Member
Verified Provider
Jade,

We're not going to be using RAM as an excuse for more overselling. If you look at the plans we sell and do the math, the extra RAM capability of the Dual L5420 is needed to maintain a low RAM overselling ratio. I understand they use up a lot of power, but that's not our main concern. Our main concern is our customers.
On second thought, go for it!.....
 
Last edited by a moderator:

Jade

NodeServ
Verified Provider
Jade,


We're not going to be using RAM as an excuse for more overselling. If you look at the plans we sell and do the math, the extra RAM capability of the Dual L5420 is needed to maintain a low RAM overselling ratio. I understand they use up a lot of power, but that's not our main concern. Our main concern is our customers.
I would think if your main concern was your customers you'd be more worried about performance than doing the math on mainaining a low RAM overselling ratio by packing more RAM into a box.
 

DomainBop

Dormant VPSB Pathogen
When you sell the plans that we do, it'll be more beneficial to have extra RAM and extra cores. Primarily for the extra RAM though, we'd prefer the Dual L5420.
When you overload nodes the way you do you need to get something like online.net's 40 core/80 thread 4 x E7-4870 with 1TB RAM

When you sell the plans you do you also need more processing power than an aging dual L5420 will offer so go for something like a dual E5-26xx, X56xx/L56xx, etc.
 

drmike

100% Tier-1 Gogent
I'll say I've seen plenty of ratios of real RAM to sold RAM...

1-3 to 1-5 are the range most folks run.

GVH container per server count I suspect (no I don't know firsthand) is laughably low comparatively, has to be.  But that is offset and made bad by two things:  1. High RAM per VPS sold;  and, 2. Likelihood the abusers are using the resources.

Again 54xx's will work, but mass packing on them no way.   Sexy e-penis competition, they come up short.
 

Jade

NodeServ
Verified Provider
I'll say I've seen plenty of ratios of real RAM to sold RAM...

1-3 to 1-5 are the range most folks run.

GVH container per server count I suspect (no I don't know firsthand) is laughably low comparatively, has to be.  But that is offset and made bad by two things:  1. High RAM per VPS sold;  and, 2. Likelihood the abusers are using the resources.

Again 54xx's will work, but mass packing on them no way.   Sexy e-penis competition, they come up short.
It seems the only reason why he wants to go with the 54xx's is to pack them. I may be wrong, but that's the information I gather from what he says. You want to add more RAM to your box so you can put more clients on it... However adding more RAM to a box that you'll never be able to use that much RAM on before the I/O is done seems a wee bit pointless if you ask me. Just my thoughts.
 
Last edited by a moderator:

drmike

100% Tier-1 Gogent
Spot on @Jade.

If my upstream couldn't deliver gear in weeks, I'd be making other plans.  Colo is one option, if their DC has the space, power, etc.  Heck if I honestly know.  Someone does.

Any time a provider wants to buy some gear I am all for it.  Shows a bit of proper thinking often.  Leasing has it's allure if you have the networth and proven track record.  But even that needs done correctly or quickly squandering income.
 

devonblzx

New Member
Verified Provider
To be fair, there's a difference between utilizing your VPS and abusing resources for the sake of abusing resources. I've terminated and refunded many people who wget cron test files to fill up disk space then rm -rf it, or consume large amount of network for no valid reason.


If you're going to use your resources, use them properly. Don't be a dick. I will say that 100tb plans are practically begging for people to do that though.
I've stayed out of this thread since the beginning but who decides what proper use is?  A VPS is a virtual private server, hence, as long as they are using the resources they are given in the package, then it shouldn't ever matter to the host how they use it.  The host should not be concerned with how their customers use their service as long as they aren't breaking the acceptable use policy.  If you don't want users to use the resources you are giving them, then stop overselling so much.

Abuse is relative.  In a plan providing 100TB of bandwidth and 250GB of disk space, then using 4MB/s or even 7.2MB/s shouldn't be considered abuse.  As seen from the dd tests posted here, the disks are taxed and therefore 4MB/s was causing the system load to go up.
 

Jade

NodeServ
Verified Provider
Spot on @Jade.

If my upstream couldn't deliver gear in weeks, I'd be making other plans.  Colo is one option, if their DC has the space, power, etc.  Heck if I honestly know.  Someone does.

Any time a provider wants to buy some gear I am all for it.  Shows a bit of proper thinking often.  Leasing has it's allure if you have the networth and proven track record.  But even that needs done correctly or quickly squandering income.
Really reseller business's don't have a viable source of income. If their upstream just up and vanishes that leaves them to crash hard. Business's that invest heavily into their own hardware/equipment will be around a lot longer than business's who resell services.
 

drmike

100% Tier-1 Gogent
Really reseller business's don't have a viable source of income. If their upstream just up and vanishes that leaves them to crash hard. Business's that invest heavily into their own hardware/equipment will be around a lot longer than business's who resell services.
This is true.   It's hard to differentiate the nature of deals at some of these providers/DC's.  

Unsure about GVH's arrangement with upstreams.  If really a straight rental (HVH is known for month-to-month rentals) or something funkier.  Assuming rental in this muck though.

My rule with dealing with upstreams is to know who owns what, how financially viable they are and balance the idea of them selling/buying out.  If a company is pitifully small/self funded startup I might deal with them, but likelihood they default on financial obligations, leases, etc. is high.  So I probably would deal with colo in a place like that.

CC for all their finger in the beehive trying to rob some honey behavior and recent inventory issues is in things for the long haul and positive on cash flow.   I've talked about what I perceive their exit plans to be and timeline on that is out in future with IPv4 allocation is gone.  Even if they go tits up, there is a collection of stuff there is worth enough that someone would gracefully buy and continue running.   In reality outside of BUFFALO, all other locations are pods, racks, limited gear, so costs known, manageable, etc.  Seamless transition on a buyout / acquisition and something I've long watched for :)

Now that other white label reseller racket... that's another beast / doom model.  It's first couple months in good idea and for folks who have customers but don't want to necessarily be a provider.   Those arrangements are too odd/stressful for me to ever consider.   See people done wrong on them all the time.
 

Francisco

Company Lube
Verified Provider
PqjOAvj.png

Saw this on LET, maybe this is how he can get 18 staff members.....

Mun
As I said, he likely uses an outsourcing group which does all of his support. That company probably has 30+ employee's. He doesn't have 18 dedicated to him, simply because his bottom line doesn't cover that kind of a staff.

As for the L5420's, you might be able to get 64GB on them if you can find 16GB sticks or something, but they aren't cheap.

You'll be doing 32GB RAM which an E3 can do. Granted, the L5420 w/ 24GB RAM can be had for a couple hundred, where as an E3 is going to run you $500+.

Francisco
 

Jade

NodeServ
Verified Provider
As for the L5420's, you might be able to get 64GB on them if you can find 16GB sticks or something, but they aren't cheap.
But how much of that 64 GB will you be using before everything starts going crazy on a 5420..?
 

Francisco

Company Lube
Verified Provider
But how much of that 64 GB will you be using before everything starts going crazy on a 5420..?
I don't know :)

We use L5420's for a while with 24GB ram and it was a stretch.

Granted, our customers have always been monsters when it comes to treating the nodes nicely.

Francisco
 

Jade

NodeServ
Verified Provider
I don't know :)


We use L5420's for a while with 24GB ram and it was a stretch.


Granted, our customers have always been monsters when it comes to treating the nodes nicely.


Francisco
I just don't see their vision of using the 5420's to satisfy customers more, I see it as an overselling gimmick, but that's just me. If you're going to put over 24 GB+ of RAM on a 5420 you might as well put it on a 56xx or better. I'm not here to rag on other providers or anything just giving my two sense.
 
Last edited by a moderator:

drmike

100% Tier-1 Gogent
PqjOAvj.png

Saw this on LET, maybe this is how he can get 18 staff members.....

Mun
And this is why someone advised Jon to cool it :)  To take a long run, read a book, do some yoga, etc.  Put down IRC, stop the social stuff in the industry, and regroup.

It's nice to see people helping or trying to (Jon), but deals like this are why I raise holy f'n hell on providers and especially this group.

I just read Biloh a bit ago on LET coming out and saying the HVH "About Us" page is going to be upgraded to define the relationship a bit more.  I doubt what gets spat out will do much to quiet folks and stem the bloodflow/distrust.   Thinking this and other Jon posting gems might have contributed to Biloh getting pinched and hand forced.

To be clear, HVH seems to pass all billables through to CC, all credit cards, banks, PayPal bear CC's name, not HVH.  Refunds seem to be all CC too.   We saw that eons ago with a credit card statement and this week with a botched refund HVH won't approve.

HVH has their own support staff that has varied and rotated over past year.  Separate from CC.   They certainly boost their support heard count via outsourcing (who doesn't?).  If they aren't proud of that revelation, then oh well hire employees and deal with the ugly financial implications of that in today's America.
 

Francisco

Company Lube
Verified Provider
To be clear, HVH seems to pass all billables through to CC, all credit cards, banks, PayPal bear CC's name, not HVH. Refunds seem to be all CC too. We saw that eons ago with a credit card statement and this week with a botched refund HVH won't approve.
Alright?

HVH has their own support staff that has varied and rotated over past year. Separate from CC. They certainly boost their support heard count via outsourcing (who doesn't?). If they aren't proud of that revelation, then oh well hire employees and deal with the ugly financial implications of that in today's America.
Mhmm?

I just read Biloh a bit ago on LET coming out and saying the HVH "About Us" page is going to be upgraded to define the relationship a bit more.
OK.

Wild guess but could be on the mark. If this is a 100% come clean, it's possible CC will mark HVH as their fully managed brand and "sell" off the unmanaged (read: LE priced) customers to address possible conflict of interest concerns.

The managed part is a likely, where as the budget users? who knows.

Peoples problems with Jon and CC is that there's just so many lies and when things finally leak/get confirmed, they reply with "Well we never denied it, people just didn't ask the right way". Again, another lie because I had a conversation with Jon during our first few months at CC over the phone where he denied any involvement in LE and that it was strictly a client<>host relationship.

Francisco
 
Top
amuck-landowner