So close I can taste it…

Shared web hosting is a changing animal. In the mid 90′s it was nothing more than some server space with a few tools to manage your files and FTP access. Today shared hosting can be as simple as a placeholder page for a newly purchased domain name to advanced back end systems with multiple databases and encrypted voip traffic. One thing that hasn’t changed is that 1% of the users cause 95% of the server issues and 95% of the headaches that all users experience in relation to server performance and downtime.

This is going to change very soon and I couldn’t be more excited!! For years I have been writing about kernel issues with linux when it comes to disk I/O tracking/shaping and monitoring. I have complained about MySQL’s inability to monitor REAL CPU and REAL I/O by user and individual databases historically or in realtime. I have vented that the linux kernel had no real metrics to break down cpu and I/O by individual user or by process over a historical period or in realtime.

Up until now we have dealt with each of these issues in a variety of ways. Some of our solutions were elegant while others were clumsy and somewhat inaccurate. Times are changing. For the first time ever I feel like I have REAL solutions to every one of the problems mentioned above. More importantly I can track and find these issues in realtime as they are happening without putting any load on the server to glean the information.

This might not seem so important to those of you reading this entry, but in my opinion it is the single greatest thing to happen to shared hosting since its inception. Thats a bold statement. Shared hosting by definition has been shaky and unstable at times for the simple reason that so many people are using the same resources at the same time. For nearly every hosting company the strategy has been to track what you can, but fix most of the problems caused by users minutes or hours after the problem has already occurred.

It doesn’t have to be this way anymore. I know because we have many tools that find these problems within a few seconds. In fact, we have the tools to find everyone of our common ailments in less than 5 seconds. I just have to finish having the tools written to act on this information to relieve server problems in realtime. That is the easy part – that is why I am so excited. I have personally spent thousands of hours researching, testing, coding, testing, researching, testing, and testing some more to solve many of these issues. It can be done.

Most of these problems can be attributed to less than 1% of our users (Actually about .5%). Most of the problems aren’t intentional on the part of our users but are created by them nonetheless. We are light years ahead of the competition in shared hosting when it comes to tracking and solving these issues. In a few months time (Most in less than 30 days) as these changes are rolled out across our servers you will notice SUBSTANTIAL increases in both uptime and server speed.

We have invested a huge amount of money and time with in house developers, kernel developers, and custom tools to address these issues. So far many of these changes have been good, but soon you will see what we are capable of and what drives us to be better. Other shared hosts won’t be able to compete with the changes we are putting in place. Blood, sweat, and tears and a lot of sleepness nights can’t be duplicated by other hosts in a few months time. Thanks for sticking with us, and be on the lookout for MUCH better server performance!

Matt Heaton / President Bluehost.com

15 Responses to “So close I can taste it…”

  1. John Lynn says:

    Great to hear. I hope you’re also keeping the end user in mind too and giving us access to the CPU utilization in CPANEL too. It’s always great to go into the bandwidth graph and know how much bandwidth I’ve used and how it’s grown over time so I can plan accordingly. Hopefully you can do the same for CPU, etc so that the 99.5% of us can know we’re safe.

  2. Robert@PNG says:

    Hi Matt,

    I have been a Bluehost customer for around four months now and am impressed to say the least. I don’t know about what horsepower or specs you have running in the backend – nor honestly do I really care.

    At the end of the day it is quality of product and service that a company offers that make it stand out in the pack. Bluehost on both counts and in my experience is miles ahead.

    I came to Bluehost from a competitor – it was taking months for their support staff to respond to my queries – since jumping ship and coming on board with Bluehost I have needed to contact your support staff on at least 6 occasions. On all counts they responded promptly,professionally and we were able to find resolution in all cases.

    Whatever it is that you got cooking there – all I can say is…

    Bring it on man!!

    R

  3. deuts says:

    We’ll watch out for those positive changes!:D

  4. adam says:

    this sounds good, although i am concrned that i or my clients will find ourselves in that 1%

  5. “This might not seem so important to those of you reading this entry…”

    Doesn’t matter, Matt (even though I do get its importance)! You’re excited about it, and that’s the best thing a customer like myself could hope to see from the President of the company. :)

  6. Homer says:

    What can I say but thank you Matt for all your dedication and hard work (and it certainly seems like you have worked very, very hard to make BlueHost the world’s best shared-hosting provider).

    I must admit that at times I have been rather concerned about claims of certain “unlimited” services, since the over-subscription model invariably seems to fail eventually, but you’ve somehow made it work, so kudos to you for that.

    Certainly the pricing structure and technical capabilities of your services are absolutely astounding.

    My only wish is that you’d also offer dedicated and/or Cloud Computing services, since I’d love to be able to offload some of my high-workload projects onto a remote host (e.g. a distro buildsystem). Maybe one day, eh?

    Anyway, thanks again.

  7. I have been with bluehost.com (with several domains) and quite a few referrals for over two years now. I had to say it, but I have seen the service getting increasingly worse.

    In the case of my church (hopebiblechurch.org) the site has gone down a dozen times over the past few months with no warning. The “live chat” comments have been “it’s down, it will be up soon” with no explanation. Just this week it has gone down twice for no real reason.

  8. Will says:

    Matt,

    I have been a customer for over a year. Overall I have been happy with the service until the last few “outages” that I have encountered. Some of the issues are simply due to a Shared Hosting environment and to be expected (runaway scripts) etc..

    However, all too often when I notice my site is not accessible, I call in to find out it is due to a “maintenance” issue.

    (Please hear me out… there is a bit of complaining, and then a positive “solution” to these circumstances)

    One day the site was offline for over 30 minutes because your admins were “installing a SSL certificate for a customer” and had to take the entire server offline for this. If I recall, this was in the middle of the day. Why not schedule all of these miscellaneous installations for 3 am ?

    Another day my site was inaccessible for bout 10 minutes before I called in only to find that the hard drives were offline (corrupted) and will be 6 to 8 hours before it’s back up and running.

    Another time it is simply routine maintenance on the MySQL server that will have the site down for about 1 to 2 hours.

    And probably as may times as I can count on one hand, the admins take it upon themselves to install, or do other server maintenance such as upgrading versions of software that end up making my site inaccessible.

    Ever heard of IonCube? Some people do use this… It’s been compatible with BlueHost for quite sometime, I’ve used it since I started my account at BlueHost for over a year. However, as I stated, no less than 5 times has my site gone offline because of some modification in the server that breaks my site.

    I guess my point is.. Can you PLEASE start having your admins email us (or only the people that wish to opt into these type of these email updates) so that we can take proactive measures to lessen the outage for OUR customers? If I had this alert pop up on my blackberry, I could have taken proactive measures (or reactive measures) to ensure I get my site back up as soon as possible.

    I can understand when there is a unplanned event that occurs and you don’t have time to warn us. But, maybe after it’s been offline for 5 to 10 minutes you can shoot out an email to those affected on the server? This can be an automated process (of course there is a human in the loop to initiate the alert) and doesn’t have to go to *everyone* on the server (only the people that opt into these types of updates).

    To say nothing about PLANNED events such as Operating System Changes.. etc.. REALLY, anything… anything at all. And I do mean ANYTHING that the admin is going to do to the environment has a chance of messing up certain sites on the server; even if they don’t think it will cause any issues.

    As I said, I have had about 5 issues regarding ionCube becoming disabled, one because Linux changed the Linux version from 5.1 to 5.2 and I had to hard code in the IonCube loader for version 5.1 in the php.ini because the dynamic loader was not compatible with BlueHost at the time (based on what I could determine). When I woke up in the morning.. it took me about 1 hour to realize you changed Linux versions.. and I simply had to change the loader file from 5.1 to 5.2.. I’m sure the admin didn’t foresee this issue, but it affected my site. This is what I mean by alerting us (that want the updates) of *anything* at all that is happening with the server we are on.

    I don’t know how popular this “opt in” list would be.. But it would certainly be helpful so that I can be aware of outages, and react accordingly. It would mean the difference of my site being unnecessarily offline for 5 to 8 hours (until I’m notified of it being down) and it being offline for 5 minutes so that I can address any problems immediately.

    I would assume a large portion of customers don’t do anything except straight HTML (I could be wrong of course). It’s probably only the customers of yours that use MySQL, IonCube, and many other features (the more links in the chain that can fail) that would probably be most interested in these updates..

    I would even be willing to pay EXTRA for these types of updates. Furthermore, I would even be willing to pay EXTRA for higher uptime. I would say triple or quadruple your current price is a price-point that I personally would agree to for better performance and redundant servers/clustering.

    Once again, it’s about respecting my customers.. If my site is offline, I look like I have egg on my face if I didn’t even know it was down.

    In Summary: Give us some warning when planned server maintenance or issues are about going to occur or are being addressed. One to two days notice for planned issues would be nice, but really.. 30 minutes would be *helpful*. For unplanned issues.. depending on seriousness, an email within a few minutes of determining that an issue exists would be helpful as well because.

    Might I suggest.. a Level ranking…
    For instance, if the RAID array goes out and the server my site is on will be down for 6 to 8 hours that might be considered a Level 1 alert because of the length of time and seriousness. If MySQL server is going to have a patch applied and be offline for 5 minutes that might be a Level 3 alert.

    The purpose of putting “LEVEL-1″ or “LEVEL-2 or 3″ in the email is so that I can key in on that “string” to program my blackberry email filter to key in on whether or not my phone should wake me up at 4 am. However, if it is a lesser important alert then I might wait until I wake up in the morning and notice the alert to see if anything has negatively affected my site.

    Overall, your support is one of the best in the industry… I usually get right through on the phone, etc.. I just can’t say quite the same about server performance. Hopefully the changes you outlined in your blog will address the server performance… to add the “icing” to the cake so to speak, the email alerts would be very helpful.

    Thanks for listening
    -Will

  9. Moe Turk says:

    What you meant to say is that you are light years ahead in the b/s department with all due respect Matt. You’re a man who accepts criticism. Those 0.5% are the real hosting customers. The other 99.5% are kiddies who believe what they read on your frontpage and believe that a server has no limitations (while every single server does).

    Some of them don’t notice that you cap upload speed and your agreement contract they have to agree give you rights to cancel their accounts if you want to and alot alot alot more things to prevent you from even yusing 1% of the space, or in this case, unlimited space provided.

    This tool is only going into effect now because BH can’t handle the overselling and therefore wants to pipedown the real users.

    Whether its 5$ hosting or 500$ hosting, the consumer wants what’s advertised.

    Take care !

  10. Shannon Cranston says:

    Yeah, I’d just like my Bluehost site to…you know…be up. Maybe, hopefully sometime this week?? SO over the whole downtime the past few days. NONE of my sites work. I can access my email. OH WOO, I can get a free email anywhere. “Thanks.”

  11. Daryle says:

    Glad to hear this…

    And if I fall into that 1% (I HIGHLY doubt it), let me know and I’ll do my part to increase everyone else’s experience ;)

    - Daryle

  12. Marian Marinov says:

    Hello Matt,

    I’m working for a different hosting company and we already did most of the things you are mentioning. It took me almost 3 months in development so the users can actually see their CPU utilization and we plan to soon have the option to give to our customers MEMORY and I/O utilization charts.

    It was very nice to hear that we are not the only one that is implementing such tools.

  13. Ray says:

    What about Solaris Zones? I think the technology has really evolved. Will you at any point begin to offer zone hosting? Zones offer all of the resource control you want and allow you to run various OS types in completely isolated environments. In theory, this would prevent the outages caused by the .5 percent bad apples out there.

  14. Scott Ault says:

    When Bluehost is working, it’s great. I left my previous provider because of constant lengthy unexplained outages. Actually, not so much the outages, as it was the lack of support during the outage. I wanted to type everything Will said above. Fortunately, he already typed it so I don’t have to. We have a way to monitor the server status, but every time it hasn’t been working (four significant times in the last year) it reported no issue. Actually, that’s not true, once it said scheduled maintenance, but that didn’t show up until after I called.

    It can’t be that hard to send out an email to notify of maintenance or 10 minutes into an issue. And why does maintenance occur in the middle of the day?

    For the last 24 hours, the transfer speed to my box has been about .9 kb/s. The replay to my ticket was, “Currently I can see notes for your server that say:
    We are undergoing network maintenance for this problem happening. We have dedicated our internal team of server administrators to bring everything back up and running as soon as possible. They have no ETA time right now when everything will be up and running. We do estimate that all problems should be worked out by tomorrow morning. We deeply apologize for this inconvenience and want you to know we are doing our best to prevent this from happening in the future.”

    Morning was eight hours ago. It’s still happening. My ticket says resolved and the only way to get updates is to submit another ticket or reply to my closed one. Why can’t I see that note? Why do I have to email or sit on hold for an hour to get a status?

    When it works it’s great. When it doesn’t work (Which I understand. Things happen) it is incredibly frustrating trying to make plans to serve my customers not knowing what is happening or when it might possibly be resolved.

    The problem isn’t the problem, the response to the problem and its communication is the problem.

  15. [...] – bookmarked by 6 members originally found by mischamezerano on 2008-11-12 So close I can taste it… http://mattheaton.com/?p=146 – bookmarked by 5 members originally found by library20 on 2008-11-02 [...]

Leave a Reply