Linode vs Dedicated Host?

I won't pretend to know anything about Dedicated servers, or the company you mentioned, but I will say this:

You don't get the Linode Platform Manager, allowing you to resize partitions, deploy multiple distributions, manage DNS VERY easily, and manage multiple users on the Manager itself (if you choose to).

Thats really all I have to add.

Edit: Linode is also working on their Backup Service (currently Beta) which will cost substantially less than you will get elsewhere.

Thanks,

Smark

Hello -

I'm not familiar with the company you are contemplating, but a few points to keep in mind:
* Is your transfer rate limited (will you ever be allowed to reach 1.3TB?)

• What kind of storage? Linode uses RAID10.

• Linode support rocks.

• What kind of processor(s)? We hear "My Linode runs circles around my old dedicated server" all the time. Linode gives you 4 CPUs. Compare apples to apples here and get a price for a quad-core Xeon.

• Will they nickel and dime you for tasks easily handled by the Linode Manager such as reboots, OS reloads, etc?

• I won't go into all the advantages of the Linode Manager, ease of upgrade/downgrade, etc.

• Are you sure you need 1GB of RAM?
  -Tom

Another thought is rather than using a single Linode, you can use multiple smaller Linodes and scale horizontally depending on demand.

Yep, I seriously question the quoted price for the Dedicated Server. I would seriously look at what you're getting and the reliability.

I have made the same comparisons as you have and these comparisons lead me to leave Linode in 2006 for ServerPronto's dedicated hosting service.

However, other factors have lead me back to Linode from ServerPronto three years later.

It is true that some aspects of the Linode plans leave alot to be desired from a resources standpoint (allow me once again to lament publicly on the fact that in an age where hard disk space costs less than $100 per TERABYTE, Linode is still offering only 24 GB on their $30/month plan). However, some aspects - such as CPU burst - compare quite favorably even with dedicated hosts.

And then there's the fact that Linode's server management infrastructure is 100x better than what you are going to get with a bargain basement dedicated hosting service. With ServerPronto, you have to CALL THEM ON THE PHONE if you want to have your server manually rebooted. And you get almost no server management tools on their very crummy web account interface. With Linode, the Platform Manager gives you complete control over your server; there is nothing that you would want to or need to do to your server that the platform manager does not allow you to do quickly and easily.

To summarize, I'd say this:

Dedicated hosts give you more resources for a cheaper cost.

Linode gives you a much better server management infrastructure and much better end-user support.

You have to decide what is more important to you. I have done both and in the end, I'm coming back to Linode.

However, I am going to keep bitching about hard disk space on Linode at every opportunity I can get.

Did I mention that Linode needs to offer more disk space on their plans?

@bji:

However, I am going to keep bitching about hard disk space on Linode at every opportunity I can get.

I'll continue to bitch ;-) that a 1U server has room for four physical disks, so there's only room for two disks worth of data with a RAID 1+0 configuration. A quick Google finds 500GB hot-swappable SATA drives for ~$350, giving the cost for 1TB of storage closer to about $700, not considering lifespan, power, or cooling. There's a huge difference between 1TB of disk and 1TB of storage, and that difference is seen when a disk fails.

Throw in the multi-year lifespan of a server, and 640GB (40x Linode 360 @ 16GB each) may well be as much as Linode can reliably sell across the entire network.

Looking at ServerPronto's main page, none of the dedicated plans below $170/mo appear to include plural "disks", so it's not an apples-to-apples comparison. Remember: don't just look at the raw numbers, look at what's underneath the numbers.

(But yeah, more space would be good. Something like S3, but faster and better and local to the Linodes…)

To expand on Tasaro's post, you need to consider the benefits of virtualization.

• "Physical" console access if something goes wrong: Lish

• Ease of rebooting

• Ease of rebuilding servers

• Fewer hardware headaches: if a piece of hardware starts going bad, let someone else worry about it and diagnose it

• Non-data service restoration: if hardware fails in the server, the hard drives can just be swapped to equivalent hardware

• Cloning virtual machines to create a cluster of servers with identical configuration on-the-fly

• The ability to add/remove resources and servers with no notice nor contract

Many of these features encourage trying things out. I have a dedicated server and a number of VPSes with Linode and elsewhere. I am very cautious when upgrading kernels or even rebooting my dedicated server. If it doesn't come up again, I'll have to call my host and likely pay an hourly rate for them to get it running again. On the other hand, I test stuff out on the virtuals all the time. I know that if I break one of the VPSes, I can probably fix it via Lish, overlay it with another, or at worst case rebuild it.

I'm not saying dedicated servers don't have their uses, but I think most people should avoid them. Application developers don't need to bother themselves with hardware.

@hoopycat:

(But yeah, more space would be good. Something like S3, but faster and better and local to the Linodes…)

The day Linode gets SAN storage at reasonable per-GB prices will be the last day I even glance at dedicated server providers. Not that I glance much that way anyway…

@hoopycat:

@bji:

However, I am going to keep bitching about hard disk space on Linode at every opportunity I can get.

I'll continue to bitch ;-) that a 1U server has room for four physical disks, so there's only room for two disks worth of data with a RAID 1+0 configuration. A quick Google finds 500GB hot-swappable SATA drives for ~$350, giving the cost for 1TB of storage closer to about $700, not considering lifespan, power, or cooling. There's a huge difference between 1TB of disk and 1TB of storage, and that difference is seen when a disk fails.

Throw in the multi-year lifespan of a server, and 640GB (40x Linode 360 @ 16GB each) may well be as much as Linode can reliably sell across the entire network.

Looking at ServerPronto's main page, none of the dedicated plans below $170/mo appear to include plural "disks", so it's not an apples-to-apples comparison. Remember: don't just look at the raw numbers, look at what's underneath the numbers.

(But yeah, more space would be good. Something like S3, but faster and better and local to the Linodes…)

I don't know what world you live in but plenty of 1U, 8 disk harddrive with 16 DIMMs exists.

Here's an example of just one:

http://www.siliconmechanics.com/i6091/d … server.php">http://www.siliconmechanics.com/i6091/dual-xeon-server.php

You can fully load the server with 64 GB RAM, 8 TB disk for less than $5,000. (These are the servers powering Wikipedia)

@hybinet:

@hoopycat:

(But yeah, more space would be good. Something like S3, but faster and better and local to the Linodes…)

The day Linode gets SAN storage at reasonable per-GB prices will be the last day I even glance at dedicated server providers. Not that I glance much that way anyway…

Actually, local disk might be preferred by some due to lower latency.

@BarkerJr:

To expand on Tasaro's post, you need to consider the benefits of virtualization.

• "Physical" console access if something goes wrong: Lish

• Ease of rebooting

• Ease of rebuilding servers

• Fewer hardware headaches: if a piece of hardware starts going bad, let someone else worry about it and diagnose it

• Non-data service restoration: if hardware fails in the server, the hard drives can just be swapped to equivalent hardware

• Cloning virtual machines to create a cluster of servers with identical configuration on-the-fly

• The ability to add/remove resources and servers with no notice nor contract

Many of these features encourage trying things out. I have a dedicated server and a number of VPSes with Linode and elsewhere. I am very cautious when upgrading kernels or even rebooting my dedicated server. If it doesn't come up again, I'll have to call my host and likely pay an hourly rate for them to get it running again. On the other hand, I test stuff out on the virtuals all the time. I know that if I break one of the VPSes, I can probably fix it via Lish, overlay it with another, or at worst case rebuild it.

I'm not saying dedicated servers don't have their uses, but I think most people should avoid them. Application developers don't need to bother themselves with hardware.

Wouldn't all of the above be gain if you simply ran your dedicated server with VM on top of it?

@timjk:

I don't know what world you live in but plenty of 1U, 8 disk harddrive with 16 DIMMs exists.

Here's an example of just one:

http://www.siliconmechanics.com/i6091/d … server.php">http://www.siliconmechanics.com/i6091/dual-xeon-server.php

You can fully load the server with 64 GB RAM, 8 TB disk for less than $5,000. (These are the servers powering Wikipedia)

That one there looks to be a 2U chassis, but I will concede that it is possible to stick 8x 2.5" drives in a 1U chassis:

http://www.supermicro.com/products/chassis/1U/?chs=113

http://www.siliconmechanics.com/i22978/ … 500-1U.php">http://www.siliconmechanics.com/i22978/8-drive-xeon-5500-1U.php

I don't believe an cost-effective set of speedy 8x 2.5" drives would offer substantially more storage than a cost-effective set of speedy 4x 3.5" drives, however.

@timjk:

@hybinet:

@hoopycat:

(But yeah, more space would be good. Something like S3, but faster and better and local to the Linodes…)

The day Linode gets SAN storage at reasonable per-GB prices will be the last day I even glance at dedicated server providers. Not that I glance much that way anyway…

Actually, local disk might be preferred by some due to lower latency.

Which works well and is preferred for the OS and a few other things, but how about mounting NFS from an internal Array for data eh?

You also need to know what are you going to use it for. If you only going to use it lightly, a VPS is fine. But if you do something that's intensive like doing lots of database queries or handling very heavy traffic, then you would need a dedicated server. If your using your server lightly and you have a dedicated server, you wasting your money where you can have the same thing at the lower cost. Plus there are features on a VPS that dedicated servers don't have, like a manager.

@techman224:

But if you do something that's intensive like doing lots of database queries or handling very heavy traffic, then you would need a dedicated server.

Or a different architecture. :-) A VPS can handle a considerable amount of traffic without breaking too much of a sweat, as long as you minimize time spent waiting for disk I/O and use the available memory smartly. If you can arrange the problem such that network throughput and CPU are the bottlenecks, you're going to need a lot of cash to match a 360's performance.

@timjk:

Wouldn't all of the above be gain if you simply ran your dedicated server with VM on top of it?

But that would still be limited to the dedicated server that you have, and you are responsible for managing it (on top of the VMs you have created). Consider the scenario that you need to quickly clone and deploy the same app onto multiple DC – it would be difficult to do with dedicated servers.

What I like about VPS is the speed of deployment (especially now with an API for it). If I get an idea that I would like to test out, and assuming I am already a Linode customer, I can just

• Log into Linode Manager

• Add a new Linode to my account

• Pay for it (most painful part :))

• Choose a DC, choose a distro, and viola! A shining new box to test out some ideas, within minutes of coming out a new idea!

And as usual my ideas are crap and ended up not working. I can then remove the Linode and get pro rata credit back to my account. Anyone know whether I can do that with a dedicated server provider?

@timjk:

Actually, local disk might be preferred by some due to lower latency.
Actually I'd prefer SAN disk because of ease of failover; linode69 explodes? shrug just migrate the SAN disks to linode287 and reboot. That's a BIG win. But it's expensive. I very much doubt linode could maintain their price point with such a backend.

@scotty:

And as usual my ideas are crap and ended up not working. I can then remove the Linode and get pro rata credit back to my account. Anyone know whether I can do that with a dedicated server provider?

Yeah one of the main things I love about Linode is the flexibility to provision new nodes for testing ideas. Plus the great support too of course :wink:

@sweh:

@timjk:

Actually, local disk might be preferred by some due to lower latency.
Actually I'd prefer SAN disk because of ease of failover; linode69 explodes? shrug just migrate the SAN disks to linode287 and reboot. That's a BIG win. But it's expensive. I very much doubt linode could maintain their price point with such a backend.
I might prefer a SAN once they've produced a bit of a better long term track record with providers. So far there have been some spectacular SAN failures at the root of major outages at host providers, so I'm still waiting to wait a bit for things to become better.

Even one of the more recently opened SAN-based VPS providers that I was watching while also evaluating Linode has had some SAN related outages that were a bit more intrusive and impactful than I'd feel comfortable with.

I'm sure it'll get better, but the SAN concentrates a lot of dependence on relatively little control hardware, so I'm not sure they're quite there yet, at least from watching some providers that use them. I suppose in most respects it shouldn't be that much worse for a single customer, but then again, I tend to see outages that can be more prolonged and affect many customers than a single machine's hardware issues would otherwise.

– David

@db3l:

@sweh:

@timjk:

Actually, local disk might be preferred by some due to lower latency.
Actually I'd prefer SAN disk because of ease of failover; linode69 explodes? shrug just migrate the SAN disks to linode287 and reboot. That's a BIG win. But it's expensive. I very much doubt linode could maintain their price point with such a backend.
I might prefer a SAN once they've produced a bit of a better long term track record with providers. So far there have been some spectacular SAN failures at the root of major outages at host providers, so I'm still waiting to wait a bit for things to become better
(Quote snipped)

But those hosts ain't Linode, so they'll have major outages and other spectacular inconveniences even without SAN :P

Anyway, I agree with u that total reliance on SAN is probably a bad idea. I'd prefer to have BOTH a moderate default allowance of local storage (for system files, databases, etc) AND an optional network-based share (for large amounts of data that must be available to several linodes at the same time) as an add-on. And obviously at prices much lower than $2/GB that Linode currently charges for extra local disk space.

A few reasons, from my point of view:

Ability to scale down below a single machine: I'm running a personal VPS for $20/month that does everything I need.

Ability to add redundancy inexpensively: I could add another slice and create a redundant/failover configuration for another $20/month, still half the price of your cheap server. I'm actually not in a huge rush to do this though because I expect Linode will move faster to deal with a hardware problem that effects dozens of customers than I expect a cheap dedicated host to move on a problem that effects one customer.

Ability to scale up quickly: If I get a sudden surge of traffic, I can basically ramp up RAM, CPU share and bandwidth in under 15 minutes. When the traffic has passed, I can scale back down again. No need to change anything about my application architecture, other than tweaking a few resource consumption limits.

It is, of course, very easy to provision new VPSs as well. I'm taking advantage of that at work to make disposable staging environments for major new releases of a web app. Create a new VPS, configure it and deploy our app, test and then switch over DNS. If everything still looks good, I archive off the old VPS and then get rid of it.

With Linode it's never really felt to me like there is anyone else on the same physical box. Obviously having "root" access without being able to see other users on the host would give this impression anyway, but I mean in terms of overall system resource utilisation.

With normal shared hosting many companies seem to put too many people on one box (happens with VPSes too), and whilst with dedicated systems you get a whole box to yourself… There's not really any benefit IMO. At least none justifying the cost difference.

You may not actually NEED the extra resources a dedicated host provides. Perhaps being on a VPS may make people think more about optimising their setup, and ultimately if you can get the best out of a smaller system it's probably better than not making much use of a bigger system.

And finally there's always the "green" factor. If you're sharing a host with other people you're not adding another set of disks, processors etc. and consuming hardly any additional power.

@Silver Blade:

And finally there's always the "green" factor. If you're sharing a host with other people you're not adding another set of disks, processors etc. and consuming hardly any additional power.
If I turn off my VPS for Earth Hour, will it decrease power consumption?

@BarkerJr:

@Silver Blade:

And finally there's always the "green" factor. If you're sharing a host with other people you're not adding another set of disks, processors etc. and consuming hardly any additional power.
If I turn off my VPS for Earth Hour, will it decrease power consumption?

No, but there won't be another commodity Celeron D server sitting on a shelf in some data center eating up power 24/7 while idling through most of the day.

Scenario #1:

Your uptime continues whilst you can feel warm and fuzzy knowing that you are saving energy by not having a whole server to yourself.

Scenario #2:

All Linodes get powered down during Earth Hour. :lol:

I think we need a vote functionality. Each resident of a specific host can vote if the entire host gets powered down for the hour.

I think a vote is unnecessary; there's bound to be at least one person on each host who hasn't drunk the Kool-Aid and thinks the whole thing is stupid.

If I vote no, I don't want my host shut off because I live in New York…

I regularly read the Linode forums, but rarely post. I could not let this thread go by without a comment.

I read all the reasons for using a Linode virtual host over a dedicated server. All posts were technically accurate and very helpful. But I have to tell you, disregarding all the reasoned comparisons, the fact is dedicated servers are just not as reliable as a Linode host. We had a dedicated server at The Planet for several years, during that time we experienced 2 disk failures, one power supply failure, and in May of 2008, there was an explosion in the data center that took out our server for a week. We moved to Linode, purchased 3 Linode hosts in 3 different cities, and have not had one minute of downtime since. And we pay less for all three than we paid for one dedicated server.

It's all about reliability, and Linode wins hands down.

@icljt:

Ithe fact is dedicated servers are just not as reliable as a Linode host.

News flash: Linode runs on the same style hardware that your dedicated was on. The difference is the quality of hardware that was used. If you had written a spec equal to what Linode uses in their hosts…

@icljt:

We had a dedicated server at The Planet for several years, during that time we experienced 2 disk failures, one power supply failure, and in May of 2008, there was an explosion in the data center that took out our server for a week.

Poor hardware, poor hardware, and that stuff could happen to anyone.

@icljt:

We moved to Linode, purchased 3 Linode hosts in 3 different cities, and have not had one minute of downtime since.

Well, now you're cheating. You're using the amazing flexibility and affordability of Linode to create a more available service.

Just doing my part to enforce rule #4

Hope I'm not rehashing too many old points, but just want to also note that dedi also will take more time to maintain.. I've gone back and forth myself and depending on the situation each can make sense, though for my stuff i only use Linode now.

One example I've been through quite a few times on various cheap dedi providers – when that cheap power supply or NIC or hard disk fails, you end up with something like this on a dedi:

* If you were lucky, you set up or bought a monitoring service that alerts you when connectivity is lost. Otherwise, you discover when trying to connect that your server is down, or another user tells you.

• At this point, you don't really know if there is a hardware problem, a software problem, some intermediate network problem or data center problem like a power outage. The dedi providers I have used have nothing close to the Linode highly active forums where we see nearly instantly when these sorts of broader problems happen.

• You contact your dedi server provider and ask them to remote reboot, which they try, and the server never comes back up, because its power supply is dead

• At this point you open a ticket with them saying your server is not pingable. Depending on the level of support of your provider, they might respond within half an hour or so, but you need to be monitoring the situation and making sure they respond, escalate if they don't, etc. WIth the cheaper providers, there will be few techs and a continuous flow of breaking servers and the response times won't be great.

• Eventually a tech will go physically examine your server and diagnose why it's not powering on. Once they figure out its the power supply, they'll have to find a new one. Usually they will have spares around. Once i had it happen that the power supply was "obsolete", so they had to switch the chassis, which required getting a new server shipped. All kinds of crazy things can happen here, and if you care about your uptime, you need to stay on top of the process for hours and hours – sometimes days.

• After you get the working power supply swapped in, they'll boot the server, and it will hopefully come back up.

• It might not come up, if, for example, you recently had modified some startup configuration option without testing it by rebooting. In that case, you' d need to get a remote console, which sometimes costs extra money and can also take some time to set up. It's often not like Linode where you click a button on the web page and have a console, whenever you want it. It's fundamentally easier to do this in a VPS like Linode, since it can be done all in software, as opposed to fancy hardware VGA-emulating virtual KVMs for physical servers.

Keep in mind that if some failure happens while you're on vacation or something, your server will be down until you can react. I've never had a non-managed dedi provider that actually fixed some hardware problem without me first pointing it out. Of course it's also the case that if your Linode goes down while on vacation you're also in trouble, but I've found this very infrequent compared to dedi provider issues especially considering Lassie reboots etc.

The only time I've found it appropriate to have a dedi provider is when I had an actual in-house team to keep them alive, and i had 10 or so physical servers and needed the power and predictability of a dedicated server Since the team had other lower-priority things to do while the servers were up, it wasn't really much more of a burden for them to handle keeping the dedi servers up, considering the extra processing power, space, etc. for the price. I've also opted for a colo rack in a similar situation – could tolerate some downtime, needed huge, huge CPU and bandwidth capabilities, had a team to keep it alive.

Having said that though, I've also had points where I had a company get 20+ Linodes, many of which were for a sophisticated test environment that really needed a lot of server nodes all hooked together, and some dev tools like subversion, wiki, etc., by a private vpn. This would have been ridiculously expensive and unreliable with dedi servers and the power of even the cheapest dedi server would have been more than really needed.

And as others have pointed out it's extremely convenient to just push a button to upgrade/downgrade your Linode level as resource needs go up/down, and Linode's prorated pricing makes this very affordable. On a dedi server, switching to a different server is risky and time consuming because OS installs on physical hardware can be very sensitive to hardware changes and things can break if you're not careful.

And one more factor, which basically elaborates on earlier comments on Linode support: I've found dedicated providers to be very shifty and unreliable -- they are always buying and selling companies and switching business tactics. I once bought a dedi server for a good recurring price by spending a higher amount up front, ie $90 for the 1GB memory upgrade instead of an extra $5 a month or something. Within months, the place announced they were increasing prices 25% because they were supposedly going to offer better support, which they didn't. Can anyone imagine Linode ever doing that? Linode has been rock solid in its business ethics and is gives us a seemingly constant stream of free memory and disk upgrades, along with cool extra administrative features. In general ,after dealing with many types of purely internet-based IT service providers over the years, the business practices of Linode really stand out and are without a doubt the best I have seen.