View Full Version : The perfect server
Assuming that some (or many) here have their own servers, I am always surprised to see the server specs at many web hosting companies. Powerful CPUs, lots of memory and so on - but if you build a business, why do so many businesses do not build their servers with basic failover technology?
Example: All servers I maintain at work (I am not a web host but a system administrator for about 100 servers) do have dual power supplies and either RAID 1 or RAID 5 setup. That's standard for all servers we have. We also throw teamed NICs into the mix where necessary. Spare hard drives of all sizes that are in use, are available.
A web host builds up a brand and tries to establish credibility. Then the hard drive on a server fails. Yes, there is a backup drive - but it is not configured for failover. No spare hard drive is available and/or the server needs to be rebuild and the configuration and data needs to be restored. Result several hours of downtime and many angry customers. Some customers might go public to forums and bash this web host. Result: new sign-ups might slow down and eventually potential customers go somewhere else and/or existing customers move away.
RAID Controllers are cheap, hard drives are cheap - but the basic web hosting industry does not use this simple technology.
Why? Is the upfront cost too high - compared to the marketing damage that a failed hard drive can create?
Would like to hear what everyone else thinks about it?
Christoph
Chris,
Glad to see there is someone else with sense.
It comes down to cost. EV1 sells tower servers with redundant nothings, unless you want to cover additional costs.
It is also the mentality of the client looking for services. Most of the people on forums are looking for the cheapest server around that they can in turn resell and oversell. They think this is how business is done. They see what EV1, Server Beach, The planet ect all have done and feel that they need to try an compete with them on the same level.
What cracks me up even more is those people and "companies" that resell servers from the same ones I have mentioned.
It costs more upfront to offer dual nics, dual power supplies, extra drives.
Basic idea buy dirt cheap and sell just above that and try an make up for it in volume.
We on the other hand have spent the extra costs. We have bought blade servers and have them connected to a SAN via fibre channel HBA's. We have just ppartnered with a leading company in the data protection industry to be able to provide our clients and their servers the ability to recover from a complete server crash in less then 30 minutes.
Why?
I believe in providing quality of services and not quatity. Yeah, I have less then 200 dedicated server clients. But, I have not had to format a box due to a crash or do ANY major other repair work except OS updates in the 2 years we have been offering managed solutions.
Why?
It goes back to believing in my business and believing in my clients businesses to provide them with quality hardware, software and customer service.
Originally posted by nsusa:
Assuming that some (or many) here have their own servers, I am always surprised to see the server specs at many web hosting companies. Powerful CPUs, lots of memory and so on - but if you build a business, why do so many businesses do not build their servers with basic failover technology?
A very good question. One I often wonder about. It can't just come down money and cutting corners alone. An extra HDD costs $70. Let's say RAID controller comes for free if you choose the right MB. So that's maybe 10% extra on the price of the server (price of the metal alone, not including $X000 in "real world" costs and SCO fees, etc).
Let's say cost of machine is not the major cost. Bandwidth, building, electricity and above all, the salary that your webhost intends to realistically pay himself must eclipse the price of the hardware?
So your $5/month hosting with extra added RAID 1 is going to cost ...say ...$5.20?
We could all stand that.
There must be some other reason. Technical? I doubt it. Let me throw a bag of rattlesnakes in amongst you here: Do hosts like to lose all the data, wipe HDDs and rebuild occasionally 'cos it allows them to wipe all traces of their bad stewardship? No of course not... :) (but I certainly love wiping my HDD at home periodically)
Does it come down to lack of imagination? Possibly. Imagination is not a quality you find in abundance in this industry. I almost fell off my seat one day last week when I saw two posts in quick succession offering something new. One was tranz and his 30 min restore. The other was offering ability to resell Linux or Windows plans from one account (by jojohost I think, I didn't read up on it but it certainly looked impressive at first glance).
Imagination I think. Vision. A lot of hosts actually have a block when it comes to putting themselves in the position of customer and seeing their needs. Some like the "control" aspect. Others like the "me too" aspect. Others the "tupperware party" (best desc I could find, sorry ;) ) aspect. Too few seem product or even business oriented.
A lot of (I'd like to say most) hosts operate far away from the standards you see in an average mid-sized "real company", to go back to the point made in the opening post. (reminder to self: must update that other thread.)
IMO :)
Interesting responses. Actually I am a little bit surprised by the lack of responses. 33 views and only 2 responses.
Cost could be a factor in the very beginning of an enterprise. But once the business is established, I would assume the web host tries to protect his investment and his business. Surprisingly - this does not seem to happen. Is it ignorance? Following this path - what happens once a server reaches the end of its life-cycle. The moment it dies - it is already too late. Is anyone planning for this and puts a time limit into production for the servers? Example: Once the server reaches 4 years of production, it will be retired and replaced with newer hardware to maintain stability.
Christoph
Larger guys have these plans in place, we have these plans in place and we replace servers every 2 years, and we own all of ours.
The difference is that you are used to enterprise solutions and the thinking that goes along with them.
To be honest most of the people in this industry have never had the opportunity to work, design and, or implement an enterprise solution.
One with thousands of workers, larger amounts of servers, auto tape loaders, SAN solutions with terrabytes of data.
Most of them started out with a reseller account and learned as they grew. So, the level of thinking is total different and its one of, "just enough to get by"
Also, most companies dont last long as they are here to make a few buck and move on to something else. The ones that want to have a long run DO have the correct thinking.
I mean you see posts about people wanting to open a DC and sell colo ect, when all they have had in the past is a few dedicated boxes.
The biggest thing I see that is the cause of a great deal of the issues within the industry is the people running the businesses and their approach to business.
BTW, this isnt pointing at anyone so no flames please, you'll just point yourself out as one of the people I'm talking about.
Rincewind
05-09-04, 06:47 PM
For a small host just putting it's foot tentativly into the water, the cost of additional hard disks and backup services is one that they try to avoid. We have an industry where people try to build businesses out of reseller accounts that cost less than $10 per month. In order to provide for those protencial customers, the hosts at the budget level cut corners down to the absolute minimum.
It's not only a lack of imagination and risk taking in the hosts. But also in the customers. Too many customers look at the bottom line price and forget about the quallity. That is untill after the server goes down. Even then, I see many requests in the forums for new web hosts "cause the current one is down" but demanding the exact same, or better, spec and price.
So people and businesses offering junk and people knowing are willing to buy junk.
Thus, the cycle continues after all.
I don't think a web host has to offer redudant systems right from the start. But once the business generates revenue the host should develop a plan to offer these options on his second or third server and thereafter. What the industry shows currently, is ignorance of proper planning in many cases.
Christoph
I would disagree. It might as well be a standard configuration. IMO. Just like RAID servers are deployed in a business without a second thought. We all know that in project management terms retro-active actions generally cost an order of magnitude more than what they would cost at the first stage and loss of homogenity would also entail a corresponding hit on reliability and effectiveness of support. Hopefully the time will come. I'm just talking about RAID 1 here. There probably would be further relatively cheap enhancements that could be considered standard as well though.
Rincewind
05-11-04, 01:14 AM
RAID is not the only fallover solution. You could set up your dns to use a fallover server. That is, if one server goes down, the site is servered form a completely different server. This would offer more protection than RAID since you could located the servers in different datacenters and so it would protect against network and system problems rather than just hard disk failure. Such a setup would be best used for static, non dynamic sites so that there is no requirement to sync data between the two servers. You would also be able to use the two servers for load ballancing for high traffic sites too.
As well as fallover service - very few advertisers talk about the backup services that they offer to customers. Whither that is just to a second hard disk, via ftp to a remote server or tape based storage system or something more advanced.
vBulletin v3.5.4, Copyright ©2000-2012, Jelsoft Enterprises Ltd.