OVH Forum – Turnover on dedicated servers | Hacker News Reader

OVH Forum – Turnover on dedicated servers

forum.ovh.co.uk

67 pointsjjoe69 comments

It's not a "failed pricing strategy" per se, the problem is that they want offers that are really attractive, they are able to provide them, but those offers are so good that too many people use them to upgrade. Which you can't blame them for, but it puts OVH in a position where it wants people to not upgrade too quickly, but still provide those good deals.

The obvious solution is to re-instate installation price, but Oles has never been a big fan of it as it increase the entry cost for new comers. Remember, they don't want to penalise new installs or growing stack in anyway, only (too) quick rent/upgrade cycle.

An alternate solution might be a "un-installation" price if you stop a server AND grab a new more/recent one to replace it after only a few month of renting, but that would be too impractical and hard to explain.

The long term fix boils down to increasing prices, adding setup fees, or adjusting pricing of older equipment. Either way pricing needs to be adjusted (pricing failure or lack of pricing foresight). There's no way to successfully enforce retaliatory measures against those who cancel / re-order every few months. Because you can always use a different CC / account / payment method.

So I really think this hiatus was put in place so OVH can accumulate enough revenues over the next few months to be able to purchase new servers. They're just not able to get in the black or even amortize older equipment.

Reducing prices on used servers seems to be the best solution. That is what iweb does (they have a 'clearance' section with cheap, slightly older servers) and it allows them to clear out their old stock.

Generally you don't need the newest hardware unless you have huge amounts of traffic.

The problem is that the price they're charging is predicated on the customer staying a customer for a minimum period of time. When that predicate fails, because leaving earlier is too attractive, the pricing model falls down.

The obvious crude solution is lock-in, but that doesn't quite cover it, because they're happy with a smaller-than-current fraction of their customers leaving early - and, probably more important, they're happy that customers feel the freedom of being able to leave.

Your suggested "un-installation price" is a bad idea: That gives the competition an edge to poach off your customers - you don't want to charge your customers extra to stay with you.

Something that might work is to raise the price of new servers, then have it drop automatically after 6 months (or so). After all, new servers are worth more, that's why people want them. Maybe the price could be lowered further automatically every n months, perhaps by a faster rate if you've held the server for a long time. That makes upgrading less attractive without making existing customers feeling shafted.

I'll admit that when I seen the new prices and quickly upgraded my servers. I got a 400% increase in hardware for only double the price.

That would probably be illegal as well in France.

It sounds like they need a standard cell phone style two year service contract with a breakage fee.

Or, more simply, charge enough for upgrades to cover the actual costs.

How to discriminate fast upgrades versus newcomers is the gist of this issue, yes.

I guess this "no installation fee" thing is killing their bottom line because for well-managed shops with automated software deployment there is hardly any disincentive to replace a rented server with a new one if prices or specs change.

These days, I ensure any personal servers that I rent can be scripted as much as possible with ansible so I can do this myself - not just work servers.

If a new provider comes up with a great offer, I can place an order, run ansible, rsync the remaining data, change IPs, proxy any remaining services I need for a little bit and off I go!

Show no loyalty, get the best deal you can.

What's experience has Hacker News had with OVH?

Personally I got a £4 a month ($6.30) dedicated sever from them not long ago when they came out (I think they sold out within a couple of days). I had to buy 3 months together as a minimum but it was still very cheap for a dedicated server (2GB RAM and 500GB harddrive). However my initial experience with OVH wasn't amazing. The website promised 24 hour delivery when I ordered it. However after 4 days no word from them so I sent a support ticket. I got a quick response saying that "As stated on our forums and website, the server will be delivered between 1 to 10 days". Kind of strange as the website had clearly said 24 hours when I placed the order (although it had since changed to 10 days) I'd covered this in my ticket but this wasn't mentioned. Anyhow I did get my server after 20 days (no word as to the hold up but at least they started the 3 months from when I actually got access to the server rather than when I'd ordered it). This is the only experience I've had with them and it wasn't terrible but not very promising. I've always had good experience with VPS on Digital Ocean so I think I'm likely to go with them instead when these 3 months run out. Alternatively I might keep the dedicated server just to backup files on to. The terms and conditions did not seem to forbid this. Has anyone else had experience doing this? Are they likely to destroy my server if I use it for my personal backups? Has anyone else used OVH? What experience did you have?

"The terms and conditions did not seem to forbid this."

Why would they forbid that? Do you realize that you are paying basically nothing for a dedicated server with a 500GB harddrive and an unmetered 100Mbps connection?

You got in on the server deal of the year and yet you are still thinking about canceling when the three months are up?

You come very close to being an example of a pathological customer.

Haha. It is a very cheap server. I was thinking more along the lines of how some unlimited plans actually just end the contract with you if you use "too much". I've never heard bad things about OVH but I've never really heard good things either. I just felt a little uneasy when I had a slightly bad first impression of them. A cheap server is good but I'd rather pay more if they're going to be bad. (Not that OVH were).

I was just interested to know what experiences others had had with them.

I've been with OVH for a few years now. Historical when they have advertised the server will be ready in 1 hour it is. The problem with the new range was that they were simply too popular. So much so that OVH's hardware partners could not deliver the hardware fast enough. They got something like 15,000 orders for just one type of server in about a day. That is 10% growth in their total server count in just one day! Sooo maybe you could cut the a little slack?

Actually the cheap servers are suppose to be for personal use. They lack hard disk raid, proper support, and have the lowest priority for hardware replacements.

Yeah, I think there were even a bunch of non-technical people over at Hot UK Deals who ordered the new low-end server. Must've been a nasty shock for OVH all round.

I knew they were popular but I had no idea there were that popular. Yes, I'll cut them some slack here, I wasn't actually in that much of a rush for the server anyway. It would have been nice if I'd been kept informed however (I'd be uneasy recommending them to a client if they need a server the next day).

My first purchase was a £16/month Kimsufi (cheap) dedicated server, when that range launched in August. I am pretty happy with the server now it has been delivered, but the launch of that range was a series of the worst mistakes I've seen a company make. I recommend not using OVH for any serious work.

The company cancelled all orders by PayPal, and didn't inform customers they were cancelled for a week.

The company took up to 2 months to deliver servers it expected to deliver in 1-24 hours. Tech support staff refused to give details.

The company refuses to communicate with customers directly, forcing them to negotiate refunds, and discuss billing issues in a public forum, which seems to be check by staff about once a day.

My big problem with OVH was the lack of communication. We don't really know why they cancelled all those orders, why they took so long to deliver the servers, or why they stopped providing billing and incident support via email. We do know that the technical support staff have not been told these things either.

What little they did say, mostly tweets in French by the CEO, turned out to be false reassurances. I think something went really badly wrong at OVH, and not knowing what damages my confidence in them.

Had a similar experience with a similarly priced server; it's simple though: too many of us bought one at the same time and they ran out of stock and people to install it. Fair enough, they're all personal servers anyway.

Why would they prevent you from using your server for backups? (Assuming the files are legal ofc)

OVH Canada recently had a major price increase, no more cheap servers as of a month or so ago.

Well I take that back, they still have entry level but they are very weak.

They used to have a $60 server that was usable but now it is $80, so $240 more a year, not trivial.

OVH Europe has better pricing/configurations than Canada.

I don't really think $20 increase is 'major' considering the hardware is brand new compared to the older SP model with a much faster CPU (i7 vs i5 class) and the other upgrades that came with it.

It is the overpriced Intel CPU. You aren't going to get much more done with that $80 server than this $30 server they sold to European customers (but not US/CA customers directly).

http://www.ovh.co.uk/dedicated_servers/kimsufi_16g.xml

4 real cores (HT is pointless in linux anyway) and runs "turbo" at 3.1ghz on all cores. 100mbps unlimited pipe. Same as the $80 server. Same 2TB software raid. Only the extra 16GB on the $80 is better, and $240 a year is a lot to pay for that extra 16GB

If you needed a dedicated server for years, I'd much rather make the $30/mo one work itself to death than pay $500+ more a year for the $80. Especially if you needed several of them.

HT is not worthless on Linux. It's not a true core but it does increase performance quite a bit, especially on a busy server.

> OVH Europe has better pricing/configurations than Canada.

You could always rent an OVH server in Europe.

They even sell the canadian servers to european customers for lower prices than to north american customers.

It's a pricing model that makes markets pay what they are used to. I had hoped OVH would have brought down prices in this market but it is not happening.

As I understand it, OVH Europe goes to quite a lot of effort to ensure non-Europeans don't order servers from them.

Oh, I assumed they would operate like a usual company. I guess in that case I wouldn't want to be their customer.

What are you comparing this to? I just got an OVH Canada server (the $79 model to compliment some AWS and Digital Ocean assets) and there is absolutely nothing remotely in the ballpark. Just to compare with AWS, the comparable (but still inferior) m2.2xlarge would run $600 per month, and wouldn't include transfers. OVH just destroys the offerings from GoGrid or Rackspace.

To this actual submission and OVH's dilemma, the solution is clear -- constantly reduce the price of servers as they age. If people end up becoming suckers of sorts by becoming locked into obsolete hardware at its original, shiny and new price, of course they're going to constantly chase the market forward. Of course the flip side of this is that if you can't project the residual payments forward for years, the initial period may be more expensive.

Oh they are good value but would be better value @ $20/mo cheaper like they were ;-)

Europe gets the same server hardware for $40/mo

Been meaning to write up an OVH survival guide. Like moving/removing their ssh keys, installing your own kernel, turning off their monitoring ping so they don't reboot your server if you are doing something that blocks the ping for too long, fixing the cpu governor so you get advertised speeds out of it, how to tune the software raid, joining the unlisted mailing lists, etc. etc.

Please do write this, that'd be awesome!

>Like moving/removing their ssh keys

/root/.ssh/authorized_keys

>installing your own kernel

option during installation. You can also configure your own ssh key and set a post-install script under the expert check box.

>turning off their monitoring ping so they don't reboot your server if you are doing something that blocks the ping for too long

option in the control panel

That reminds me of dealing with Fasthosts. They used to reset iptables to their own defaults during the boot sequence...

Why would you compare a budget unmanaged dedicated server provider like OVH to a high end managed server provider like Rackspace or a cloud provider like GoGrid? It's an apples to oranges comparison.

Hetzner and Burst.net would be more appropriate direct competitors in the budget space. They may still not be quite as cheap as OVH, but they're much closer than the other options you mentioned.

Why would you compare a budget unmanaged dedicated server provider like OVH to a high end managed server provider like Rackspace or a cloud provider like GoGrid?

GoGrid offers dedicated servers as well. That RackSpace manages the servers is anathema to most of HN, but it is often brought up in regards to dedicated servers.

Yes, they are competitors to the people who look for servers, and whether one is "high end" while the other is "budget" comes through in the product and the service, not in the price, and these often come up in the same conversations.

Low end and high end is just an extra dimension secondary to them being different products for different markets. Unmanaged dedicated hosting to managed hosting to cloud hosting are as different as a car rental to a limousine to a public transit service.

>Unmanaged dedicated hosting to managed hosting to cloud hosting are as different as a car rental to a limousine to a public transit service.

I've lost track of what you're trying to argue. People do compare car rentals to limousines to public transit to cabs to trains to airplanes. If your goal is "get somewhere" (or "online computing resources"), they are in competition.

Sure, and people will make the argument that taking the bus is much cheaper than taking a limousine, when comparing methods of transportation. However, people aren't generally comparing pricing between limousine company A and city of X public transit, and wondering why one is so much more expensive than the other. They're going to compare limousine company A with limousine company B instead. If all you need is public transit, then by all means use that method of transportation, but there are times when a limousine service may be called for.

Likewise, if all you need is an unmanaged dedicated server, and not a managed server then great. Yes, it's going to be much cheaper than a managed hosting company, as it doesn't include server management. Just as a car rental doesn't include a chauffeur. But if you're honestly shopping around, you should be comparing one unmanaged dedicated server provider to other unmanaged dedicated server providers. And you may not, but I would also make the distinction between budget unmanaged providers and premium unmanaged providers. There's little point in comparing OVH to SoftLayer. Just because OVH is much cheaper doesn't mean SoftLayer offers poor value, not that I'm arguing it doesn't. OVH's network has limited transit options, has frequent congestion issues, and much higher latency transport links. They also use desktop grade hardware, which are much more prone to failure. Anyone heavily involved in the hosting industry is well aware not to host anything mission critical on OVH's service, and those that do are well aware of the risks involved.

I wonder if this explains why, when my current server failed and I decided to replace it with two more powerful ones for half the price, my order was mysteriously on hold for a week until the servers I ordered were out of stock, and then the payment was cancelled after it was too late to question it :(

(They did at least give me the old server back with the failing drives replaced...)

They could make everything more expensive, and offer significant discount for "reserved" instance for 1 to 3 years, like Amazon does.

I got the first dedicated server offering in their Canada DC. Since then, OVH offered better servers at lower prices. I ask OVH support if they can lower my monthly payment to match a similarly spec'd server. Keep in mind, they no longer offered the one I was using afaik. They said "No, cancel the server and order a new one". So that is exactly what I did, I was disappointed because it was more work. If they would just offer a price match , they could have lower turn over.

It sounds like they aren't able to offer discounts. I did the same with my provider after a few years. They halved the price.

It's interesting to note that they've decided to go and upgraded all existing older SP servers which previously had 100Mbit ports to 200Mbit for free. They've been upgrading all the switches the last 3 weeks.

I think they could get less churn by doing some of these upgrades but also by lowering server prices after reaching a certain milestone.

Personally I know I was interested in moving to a newer SP server which had a faster CPU and more RAM combined with 200MBit port for only $20/more a month until I heard they were going to upgrade my port for free. But that wasn't advertised until I emailed them and it hasn't happened weeks later. If they were proactive in emailing an upgrade notice and also maybe even upgrading the ports first before the new servers came out they would prevent a lot of wasted 'upgrades'.

They most likely did not email you because they actually don't have the capacity to handle the extra load caused by everyone's cap being doubled.

You can clearly see their capacity with the OVH Weather Map and it really doesn't show that being the case for at least North America (I know it's not peak time but I've been keeping tabs on it at peak almost every day for the last few months):

http://weathermap.ovh.net/usa

I don't know what the situation is in Europe.

That's never stopped OVH in the past...At one point that had 150x more sold bandwidth than available, they rely on high contention

I just got an email from them saying "We would like to remind you about the near expiry date of your services at OVH". Apparently my server will expire in "less than 60 days' time".

I wonder if this is part of there new strategy to keep people on their current plans. I only ordered my server for 3 months so they're very eager.

I think the real issue here is PayPal. I just opened an OVH account not that long ago and was surprised they only accept PayPal in Canada and USA.

So I opened a business PayPal account even though I have no intention of becoming a "paypal merchant". To my surprise, PayPal holds the money 8-10 days. Excellent.

It took about 20 days to get the payment through, so at least they started the billing cycle from that point.

Having heard so many PayPal horror stories I would not be surprised if they got screwed over somehow.

I think I'd have stuck around if their interfaces weren't so awful.

There were literally 5 different control panels I could use, it took me a support ticket to actually find where my server was residing (only in a single control panel version). Love that they're making a sixth control panel to fix that.

It's really the first time a company that big halts all its commercial offerings! We are running a few OVH servers and they're clearly a great value for money.

Not true... Softlayer when they were servermatrix suspended sales in 2005. IIRC it lasted about 6 months (but that's from my foggy memory.. so I could be wrong).

Found this WHT thread:

http://www.webhostingtalk.com/showthread.php?s=&threadid=453...

The orquestration software like Ansible, Puppet, and Chef are killing them. You you scripted your server, the cost to move to another server is almost nil.

A kickass cloud offering could be the solution.

Interesting problem and read (if you're able to get past the Frenglish).

======================================================

Dear Customer,

The server deals that we offer are based on certain principles which make renting a server from us very appealing: - no installation fees - no commitment - monthly payment - regular release of new features

Up to 100, 000 servers in our DCs, we haven't had any particular problems with these principles. To manage growth, it was sometimes necessary to add a limit of 2 new servers per month. Sometimes we had added installation fees for a few months, but overall, the model functions very well.

Between 100,000 and 150,000 servers, we felt that these principles began to create problems in terms of turnover.

What is turnover? Turnover is when our existing customers that rent servers decide to take new ones from us to replace the old. This if not a problem if the old server has been rented for 2-3 years, or if we can re-rent the old server as a Kimsufi for another 2-3 years. But with 150,000 servers, the number of customers renting a new server again after 3 to 6 months has become the majority. And the old servers are becoming comercially obsolete too quickly.

Today we have 170, 000 servers and since the range was updated at the beginning of August, the turnover has sped up even more to become too high. The majority of deliveries are basically just turnover. We are no longer delivering enough servers to keep up with new projects or new customers. The delivery time for everyone has become too long and we can say that the mechanics isn't working any more.

As for the 170K servers already leased out, there's no problem to report and no change to expect. Everything is running perfectly well and we're not changing anything. The problem only concerns new orders for new servers.

How do we fix this turnover issue?

---------------------------

We think we can find a better way of working which will enable us to deliver servers in 1hr, while managing 200,000 or even 300,000 servers. But such a way of working requires serious consideration.

We have started to make adjustments on certain elements such as IP Failover, but clearly it is necessary to go more in-depth. We have multiple avenues to follow, but if we start making changes every day, a piece here and a piece there, nobody will understand anything any more.

And so even if we have more than 10,000 servers ready to be delivered (all ranges and DCs mixed up), we have thus decided to "sell out" all dedicated server offers in France and the European subsidiaries, as well as in Canada and the USA. Orders already placed will be delivered, but we are not accepting any more new orders.

This "sold out" action will enable us to take time to analyze the sales of the last few months in greater depth, and understand the behavior of each customer contributing to the turnover.

We will then put the servers back on sale in limited quantities (batches of 100 servers) and new principles to then re-analyze what has happened, and so on and so forth until finding the principles which will allow us to:

- deliver the servers to you quickly - offer very competitive offers - increase customer loyalty to their servers

At the same time, we're speeding up the rewriting of the API in v6 in terms of domain names, emails, web hosting and dedicated servers. The same goes for Manager V6 which will replace V3 and V5 before the end of the year. Our order system will be reviewed before the end of the year for all the services that we offer you. We will be able to offer bank card auto-renewal in Europe and Canada. In short, we are making the most of this " sold out" period to focus internal resources not on growth, but rather on consolidation of our backoffice and on careful consideration of the mechanics of our offers.

In the meantime, if you have new projects which oblige you to rent new servers, please don't hesitate to contact us so that you can place an order manually. We have the server stock but we don't want to deliver it and continue this high turnover cycle.

We're counting on your feedback to help us to refine the mechanics, which must be convenient for you and allow us to do our job without having to change the rules of play every 6 months because we have 200K or 300K servers in our DCs. This is basically a good time to discuss and exchange, and then to come up with the idea that nobody has yet come up with…

Best wishes,

Octave

"Interesting problem and read (if you're able to get past the Frenglish)."

His French is just as bad. It's kinda refreshing to read stuff that isn't passed through PR consultants for polishing. OVH brutal honesty is very rare.

It's true that some people just have two second languages, and some barely a single.

You never get fired for choosing AWS.

If you're just an employee in a big corp choosing a vendor, of course you choose AWS - it's not your money, so who cares?

But for those who are in charge of a startup, every cent that goes towards cost counts. AWS can be 10 times more expensive than a regular dedicated server(for the same amt of resources)

You'll get fired for making a devastatingly bad call, and AWS could be one of those. It's difficult to work with, can be expensive if not used effectively, and exposes you to completely different kinds of risk than your team might be used to.

We're a small company (30+ people) and been happily using AWS for several years now. We've found them price competitive for the same features - and they are really easy to use if you're looking for scale.

A few reasons for why I think they're now the 600-pound gorilla in that space. ;)

Just curious, do you just use EC2 and do you dynamically scale the number of machines you use frequently? All the math I've done for a constant number of servers/instances shows that unmanaged dedicated or colocated hardware comes out far ahead in price.

I'm not saying it can't be done, but if you do it wrong you can really mess up your business.

And you have the same risks if you pick any other cloud provider.

AWS has different risks. Many, such as Linode, will bend over backwards to fix your systems if they've suffered a hardware failure or data corruption issue. In using them for six years, with 50-60 instances at any given time, I've only permanently lost one instance.

AWS is a purely do-it-yourself model, where hardware will blow up from time to time and data will get lost. Any and all infrastructure on AWS is to be considered temporary. Safety only comes with redundancy or a service layer like EngineYard that handles durability and persistence for you using their tooling.

In three years of using raw AWS and hosting about a dozen instances, I permanently lost six instances, had several others corrupted and rendered unusable. You need to be prepared for parts their infrastructure to completely evaporate at any given time. Everything has to be persisted to S3 and off-site relentlessly.

As much as it's true that you need backups for any provider, it's especially true for raw AWS. You're a tiny boat in a big ocean of trouble if you only have a few instances, no redundancy, and infrequent backups.

been using OVH for years and I've look into cloud based alternatives frequently but I have yet to find a provider that is even comes close to OVH in terms of value for money.

If you haven't already, I would recommend to give a shot to Digital Ocean also.

Digital Ocean isn't even remotely in the same business as a dedicated server provider. And, unless you are heavily IO/bound (few people are), costs at minimum double what I'm paying for an OVH dedicated server.

You will if you don't have a backup/redundancy strategy other than "it's AWS, we don't need one!" Seems like there's an AWS outage every week.

Though, to be fair, you need a backup strategy regardless of where you are. If you don't have one, it's not anyone's fault but your own. AWS in particular is built with the mantra of "expect and embrace failure".

Redundancy is dependent upon your usage case (and budget), but if your livelihood is on the line, you'll want a redundancy strategy as well regardless of host. Again, you are ultimately responsible for catastrophic failures that result from poor planning.

If you have the money yes :).

Reply on news.ycombinator.com