I run a small community forum and a personal homepage, both from the same Windows 2000 Gold hosting plan (100Mb/2Gb/£150pa) using www.hostinguk.net

My personal homepage has negligible activity, however the the forum (a community of digital camera users) has around 500 members, with about 125 visits a day. Activity is still sufficiently low that an Access 2000 database provides acceptable performance.

My problem is reliability. It seems that I have about 3 or 4 cases a month of server unavailability. Anywhere from 2 minute outages to 2 to 3 hours of either unavailable service, or crippled (unusable) performance.

It looks like about 30% of these cases are unplanned critical failures (server corruption, crashes, denial of service etc), 10% are planned maintenance (patching IIS) and the rest appear to be caused by other users on the same shared server writing and executing bad code. (usually ASP - not closing objects, misuse of session variables, global.asa problems etc)

In my last email to hostinguk.net I explained that I felt that their level of downtime was unnaceptable, and got the following response:
Judging by the number of times you are contacting us we seem to be unable to provide you with the hosting you require, and I would suggest that you need to be hosted away from amateur programmers, perhaps on a dedicated server. I think perhaps this may be the best way for you to go in order that you can be hosted in such a way as to not be affected by the actions of others.

Obviously you may choose not to do this with us, which we of course understand.This makes it obvious to me that they are unable or unwilling to provide the level of reliability I want for the price I am prepared to pay, so my question to the panel is:
Is it this simple? "Shared Hosting=Unreliable?" £150 a year seems to be a lot of money to pay for a service with no availiabilty guarantees...

If reliability on shared hosting is just a pipe-dream, then I might as well learn to live with it, since I cannot afford a dedicated server...

By the way - this is not an attempt to discredit hostinguk.net - My only complaint with them is reliability, in every other way they have been helpful, courteous and professional.

All opinions welcome!

Peter

Naw reliabilty is not a problem on a shared server we have had 5 months now of 100% up time on one server with 237 accounts on it.

Your reliablity problem is easy to understand "Windows 2000" :D

Switch to Linux

I am adding another thought to this post. There was a thread somewhere on here a few months back about the longest up time, there were a few post discussing Windows servers and their uptime. Maybe search for that thread and see if any of those guys posting about windows servers have uptimes you feel you could live with, or at least maybe contact one of them and see if they can give some stats on shared windows hosting.

Originally posted by Monte
Your reliablity problem is easy to understand "Windows 2000" :D

Switch to Linux
Totally agree :d

OK, I know about IIS's shortfalls - I'm sure (judging by market share) that there will be twice as many messages to "ditch Win2K/IIS" as any others.

I appreciate (& mostly agree) with your views, but right now I'm committed to ASP, and a number of ASP server-side components (otherwise Chillisoft ASP would be an easier move for me), so I would be even happier to hear from hosts who can give specific comments on their successfull approach to reliable Win2K/IIS hosting!

Thanks!

You should not have such problems with your shared hosting package. For the amount that you are paying, you should be able to find a provider that can offer 99.9% uptime on their accounts. I would suggest shopping around and possibly getting setup with a trial on each prospective company to check their uptime percentage.

Of course, moving to linux would be the best bet, but that would most likely mean converting to PHP which you may or may not want to do!

Originally posted by Monte
Naw reliabilty is not a problem on a shared server we have had 5 months now of 100% up time on one server with 237 accounts on it and only 10 minutes down 2 others in that same time frame.
Hope they're not 237 reseller accounts....:D :stickout

Originally posted by Aussie Bob

Hope they're not 237 reseller accounts....:D :stickout


hmmm not sure what kind of server you would need to handle 237 reseller accounts, I don't think there is one:rolleyes:

I didn't say they were all my accounts only that there were 237 on the server;)

I know there are enough flame wars about what OS is best. However, you really don't want to run Windows with IIS, it's just asking for trouble. There's too many problems and Windows is not made to offer someone the control they'd need to keep things in check, not have the system overload or do the things you should need to do. I really don't want to get into an argument with any Windows hosts about it, but it's just true and I'm not going to beat around the bush about it. I don't know that you'll find a better ASP hosing solution on anything other than a Windows system, because those other's aren't really perfected yet.
However, running IIS, Access, etc., it's just not a good idea. Access isn't really built for much more than simple things and shopping lists (in a lot of people's opinion and it's definitely not efficient compared to any other DB type of software). I'll say this is my opinion and leave it at that, and not get into all the issues I have with Windows servers. It's just hard to believe anyone would run a Windows server with any very well trafficked site.

Although I am UK based, I used to use www.crystaltech.com in the States for my ASP hosting. They are excellent and I would use them without hesitation now apart from having made the move to PHP/Linux etc.

My problem now is to find a really good Linux host who can meet my specific needs, however I know it will be far easier to find a top quality Linx host company than a top quality W2K hoster.

Thanks for your comments - I would still love to hear from Windows 2000 hosts that feel they can offer reliability on a shared server... anyone out there?!:confused:

IIS = Headaches

Enough said.

Originally posted by Monte
Naw reliabilty is not a problem on a shared server we have had 5 months now of 100% up time on one server with 237 accounts on it.

Your reliablity problem is easy to understand "Windows 2000" :D

Switch to Linux

I am adding another thought to this post. There was a thread somewhere on here a few months back about the longest up time, there were a few post discussing Windows servers and their uptime. Maybe search for that thread and see if any of those guys posting about windows servers have uptimes you feel you could live with, or at least maybe contact one of them and see if they can give some stats on shared windows hosting.

Monte, not to fuss with you, but how did you upgrade apache without the reboot. the big security hole was only 4 months ago.

also, the thing about unix uptime which will always beat out win uptime is that unix services can be started and stopped with out any real problems and the upgrades can be done that way. while in windows you need the reboot.

so I never really trust those netcraft numbers. if a HTTP request can not be done then the server is unavalible in my view. that's downtime to me.

OK - no Win2K host was prepared to stand up behind ANY kind of reliability claim. So how do you sleep at night, knowing that your main source of revenue is based on an risk of failure that is completely outside of your control?

Aren't there even any recommendations? a way of sandboxing users writing bad code? a way of isolating users to their own processes? Knowing the lowest impact time to apply critical security patches, and giving fair warning? More proactive measuring of single server processes, not just a straight "If my monitoring software can ping the web server, I will assume everything is OK, until my customers tell me otherwise"?

What about flexible hosting plans that include solutions that address these risks - all hosts seem to follow the same boring approach "Heres one server, with ALL the toys installed" - "now tell me how much you want to pay, and I'll tell you how much bandwidth & disk space you get" This approach must surely be one of the most significant contributing factors to this unreliability - customers start downloading scripts & components and hacking code with no idea of the basics, since they never needed to make a commitment to one technology or scripting language.
What about rewarding low-risk customers by moving them to lower contention ratio servers with other 'mature' customers, so that reliability risks outside of your control are reduced?

I saw a thread recently calculating 12-24 months as the average time a customer stays with one host - I believe that is due to neglect, plain & simple. And based on the low margins from simple hosting plans, compared to the high level of support, this doesnt seem to make for profitable business...

The unimpressive response to this thread means that as a customer, I now appreciate that none of the Windows hosts so far are taking reliability seriously. - strange, since they have to make a higher investment in license costs than the linux/unix/apache/mysql users - you would have thought they would be more interested in whether its paying off...

*SIGH* I guess I should start looking at Chillisoft ASP as soon as possible, and see whether there are any good server components for Chillisoft (SunOne) ASP that can replicate my current functionality... *thinks* must buy some shares in Sun microsystems...

P3TER,

You make excellent points about the problem with a great deal of hosts. However, this problem is very much existent on *nix servers, just as much as Windows servers. Of course there are pros and cons and Windows are by default (barring any exploits in the OS's services) more secure and stable for someone that doesn't really know much, than a *nix system that's set up and ran by someone that doesn't know much.

I won't get into all the issues I have with Windows hosting or how much *nix is better, I just wanted to say that these problems you mentioned are a very accurate portrayal of the issues with many hosts, no matter what the OS and few seem proactive or skilled enough to either want to bother or know how if they do want to. Luckily there are providers that are aware and attentive to these aspects and hopefully you won't have too much trouble locating one.

Greetings:

"Shared Hosting=Unreliable"

It depends on how you define reliable.

Most shared hosting companies today offer 99.5% up time or greater.

Given 365 days in a year, there are 8,760 hours in a year.

99.5% up time translates to 43.8 hours of down time per year or 3.65 hours of down time per month.

99.99% up time would provide 8.76 hours of down time per year or 0.73 hours of down time per month.

Some providers offer 99.999% up time which means only 0.0876 hours of down time per year or 0.0073 hours of down time per month.

Furthermore, some thoughts...

1. Microsoft Windows tends to require more patches than any other server-based operationg system to keep it secure.

Last year Microsoft released 60 security patches (not including recommended or general patches).

This year (with over 3 months to go before year end), Microsoft is up to 50 security patches. I have a feeling that they are going to break last year's record.

2. Microsoft Access is a DESKTOP database engine. It IS NOT a client server database engine.

Microsoft engineers who are know Access and its limitations will tell you point blank you should never use it on a production Web site because it is a DESKTOP database!

3. The Microsoft Access driver has both a connection resource break point (usually 20 to 40 simultaneous connections) where it will bog down; AND, it has a per box limit (I forget whether it is 128 or 256).

So it is very possible in a shared environment to get NO connections because other sites on the box are using the connection pool to the max.

If you want to increase your reliability:

1. Ensure you are with a host that offers a minimum of 99.5% up time.

2. Ensure you are with a host that lets you know in advance of scheduled maintenance.

3. Ensure you are with a host that honors down time credit requests when their performance guarantee is not met.

4. Ensure you are with a host that has a monitoring system in place to help them keep their service level agreements. Monitoring systems don't prevent down time; however, a good one will help keep down time to a minimum by properly notifying the correct personnel for the type of problem, and escalate automatically if the problem is unresolved for a given period of time.

5. Move from Microsoft Access to Microsoft SQL Server, mySQL, or another client server database engine.

6. Move from Windows to Linux, Sun Solaris, or another reputable UNIX operating system.

Thank you.

Originally posted by dynamicnet
Greetings:


1. Microsoft Windows tends to require more patches than any other server-based operationg system to keep it secure.

Last year Microsoft released 60 security patches (not including recommended or general patches).

This year (with over 3 months to go before year end), Microsoft is up to 50 security patches. I have a feeling that they are going to break last year's record.

2. Microsoft Access is a DESKTOP database engine. It IS NOT a client server database engine.

Microsoft engineers who are know Access and its limitations will tell you point blank you should never use it on a production Web site because it is a DESKTOP database!

3. The Microsoft Access driver has both a connection resource break point (usually 20 to 40 simultaneous connections) where it will bog down; AND, it has a per box limit (I forget whether it is 128 or 256).

So it is very possible in a shared environment to get NO connections because other sites on the box are using the connection pool to the max.

Thank you.

Before this turns ugly,
security patches for a redhat web server also doing mysql and dns and win2k server with simular setup are neck to neck this year. ( 48 average security and general fixes and or service releases )

but the point is more along the line like this:
MS installed end user base large and shrinking
linux user base small and growing.
whom are they going to look for as the better hacking target
M$, if you make your bones on a new M$ secirity risk, you got yourself a history and a sure chance of a job.

at the end of the day, it's the system admin that makes or breaks the server. I know that we spend alot of time reviewing log files and trying everything and anything to prevent long term server problems.

Mike

Originally posted by P3TER
OK - no Win2K host was prepared to stand up behind ANY kind of reliability claim. So how do you sleep at night, knowing that your main source of revenue is based on an risk of failure that is completely outside of your control?

Aren't there even any recommendations? a way of sandboxing users writing bad code? a way of isolating users to their own processes? Knowing the lowest impact time to apply critical security patches, and giving fair warning? More proactive measuring of single server processes, not just a straight "If my monitoring software can ping the web server, I will assume everything is OK, until my customers tell me otherwise"?



Peter, do a search for some of my post,
the toics you need to look for is

1) uptime
2) server high avalibility
3) server tuning
4) domains per server ( thier are 2 schools of though on this )

I covered those topics before ( win platform base only ) and the fustration of having to deal with lower priced competitors that will keep there cpu's and drive i/o untuned. Can you imaging having to deal with someone that places 600 domains ( active ) on a server that does not even get tuned. How the heck to I compete when everyone thinks we are all the same. And my prices are fair.

The only way I get business is by word of mouth and small advertising in the same field ( I would like to say at this time that I have a few unix host that have windows plans with me, just to keep up on the secuirty updates, I beieve that if they ever disclose themselves I will have very positive reviews ).

Well I had to vent today,
Verio is on the rag and there seattle routers went boom all night long

Mike

ADEhost: About Upgrading without a Reboot:

Unix and its derivatives like Linux are great operating systems - I'm sure you know. One of the coolest features is flexibility.

To Upgrade Apache without Downtime:
Assuming your current install lives at /usr/local/apache_x.x.x/ and you have a symlink from /usr/local/apache to that dir.

1) Get new Apache tarballs, might as well upgrade PHP et al at the same time.
2) Install them to /usr/local/apache_x.x.x (where x.x.x is the new version.)
3) Set up the conf files to ONLY listen on one new IP address, an IP that your existing apache is not configured on. Test your install on that IP.
4) Shut down your new apache, add ALL of your virtual hosts to its conf file.
5) Shut down old apache, move your symlink, start new apache. 10 seconds of downtime here if things go well - if not, shutdown new apache, move symlink back, start old apache again and figure out the problem.

You can make 6 attempts at upgrading like this and only have a total of one minute of downtime. Downtime in small increments like this (10 seconds) will only really harm people who are doing large HTTP downloads as they will be broken at each shutdown and start (even on a HUP signal downloads are lost.)

Most people use software for resuming large downloads, so thats fine. I doubt many people would even NOTICE that there was any downtime at all.

Thanks Refcom,
I just handed it over to the guy that does all the unix stuff.

I just thought that there was not way without the reboot. LOL

that's what I like about unix, you really can get away with just about Zero downtime over a long term period.

more niffty than anything else

Mike

Well, If you want something interesting - I have been looking (in my endless search for better uptimes) for a way to do KERNEL upgrades without a reboot. In theory it can be done, but spilling all of my technical knowledge here would be beyond the scope of this board. It could be feasable though, and I think it can be done with only about 10 seconds of downtime with a similar model as that mentioned above. This would be a kernel update withOUT a reboot at all... I mean patching the kernel as it is in memory in REAL time...

THere are problems - other software for example, that is running on a real-time modified kernel, but - for all intents and purposes of this board I am stopping here :D

Do some searchs on google, some people, especially the FreeBSD folks, have done a lot of work and tech talk on this.

Originally posted by refcom
Well, If you want something interesting - I have been looking (in my endless search for better uptimes) for a way to do KERNEL upgrades without a reboot. In theory it can be done, but spilling all of my technical knowledge here would be beyond the scope of this board. It could be feasable though, and I think it can be done with only about 10 seconds of downtime with a similar model as that mentioned above. This would be a kernel update withOUT a reboot at all... I mean patching the kernel as it is in memory in REAL time...

THere are problems - other software for example, that is running on a real-time modified kernel, but - for all intents and purposes of this board I am stopping here :D

Do some searchs on google, some people, especially the FreeBSD folks, have done a lot of work and tech talk on this.

all matters aside, I think a load balanced system would stop this update headache of downtime in a heart beat LOL. atleast then you know if you made it or not.

in looking at clustered systems on MS ( slightly different from load balancing ) I have come to the conclusion that downtime has to happen ( even if it's 4 seconds ), the only way to escape that problem is by load balancing servers and round robin the entire server base.

and your right, it's way outside the scope of this forum on how to keep a system hot, I think that out of this entire forum I have found only 5 or 6 people whom have the same uptime concern, and server concerns as I. Damm shame that they are the only ones and that we have to compete against the regular people.

mike

That's quite true - of course, make sure your load balanced system looks like this:


Internet
|
Head 1 ---- Head 2
|
[multiple servers here]


My method of load balancing is to keep TWO front end servers, and I make them visible to the internet by having a serial connection between them - if one can't connect to the other it uses ARP Spoofing to steal is MAC address and take over.

Nifty system eh?

Originally posted by refcom
That's quite true - of course, make sure your load balanced system looks like this:


Internet
|
Head 1 ---- Head 2
|
[multiple servers here]


My method of load balancing is to keep TWO front end servers, and I make them visible to the internet by having a serial connection between them - if one can't connect to the other it uses ARP Spoofing to steal is MAC address and take over.

Nifty system eh?

just send you a PM about something

Originally posted by ADEhost
Peter, do a search for some of my posts,
the topics you need to look for is

1) uptime
2) server high avalibility
3) server tuning
4) domains per server ( thier are 2 schools of though on this )

I covered those topics before ( win platform base only ) and the fustration of having to deal with lower priced competitors that will keep there cpu's and drive i/o untuned. Can you imaging having to deal with someone that places 600 domains ( active ) on a server that does not even get tuned. How the heck to I compete when everyone thinks we are all the same. And my prices are fair. Mike,

I looked at a few of your more recent posts, I will certainly spend more time looking at the others - You seem to be one of the few hosts that has an understanding of what causes Win2K to become unreliable and how to build your business around the limitations.

This is all I'm looking for - and, to maybe answer one of your questions "How the heck do I compete when everyone thinks we are all the same?" I think you, and any other hosts who think reliability and quality is important, should make a point to sell your unique benefits. Dialup & Broadband ISP's have now started making 'contention ratios' a unique selling point - If you start doing the same, quoting your quality processes and your low server utilization rates as your business differentiator, I believe you will find customers choosing you for just that reason, and ultimately, you will be able to charge them more. (While the "me too" hosts lose their customers in 2 years through lack of care & lack of quality.)
But dont come right out & say "we are more reliable" - this just becomes a stick for your customers to beat you with - but focusing on your quality practices shows you care, and if nothing else gives your prospective customers a list of really hard questions for to ask your low-price competitors!

I'm in IT Management myself, and I guess what I can see is that the right blend of commercial sense and technical skills makes for for a good hosting company - It seems many hosts are too much 'techie', not enough businessman. - this leads to the approach of 'keep adding stuff till it breaks, then start firefighting' (just like my OWN attitude to putting crap on my home PC!)

Thanks to all for their contributions!