What if they didn’t?

We have long offered a plan called “bandwidth buckets” designed to let people who knew in advance that they would be using more bandwidth. However, this program was never terribly popular, partly because it’s pretty complicated and partly because it requires a lot of micromanagement and a fair degree of prescience to make the most of it. We’ve killed it. (Naturally we’ll honor all previously-purchased bandwidth buckets.)

As of February 1st (midnight UTC), we’ve begun tracking the total bandwidth used by all the sites in each personal bandwidth account. The 1GB/$1 pricing will remain unchanged. The change comes on the second gigabyte. The short version for nerds like me: from now on, the number of gigabytes you can buy with $1 increases from 1 GB to 5GB proportional to the base-10 log of the number of gigabytes your account has used. If you think logs are for sawing, here’s a breakdown using nice round numbers:

1 GB = 1GB / $1
10 GB = 2GB / $1 ($0.50 / GB)
100 GB = 3GB / $1 ($0.33 / GB)
1000 GB = 4GB / $1 ($0.25 / GB)
10000 GB = 5GB / $1 ($0.20 / GB)

But here’s the cool part: those numbers are examples, not tiers. As soon as you’ve used that first gigabyte of bandwidth, your prices start coming down. As of now, the number of bytes you get for a penny is displayed on your account page. After that first gigabyte goes by, this number starts to climb. And it climbs fast. Take a look at this breakdown, which shows the first few gigs:

1GB = 1GB / $1
2GB = 1.30GB / $1
3GB = 1.47GB / $1
4GB = 1.60GB / $1
5GB = 1.70GB / $1

That’s right, you’re now getting about 30% more for your bandwidth dollar after your second 2GB, and 70% more after just 5GB. This is an automatic discount; you don’t have to do anything to get it. But you don’t even have to get to 2GB. If you’re on your account page and you’ve accrued more than a gig of transfers, you can actually hit reload a minute later and see your bytes-per-penny increase. You can watch your hosting getting cheaper in real time. I’m actually kind of obsessed with watching mine, probably for the same just-a-few-more-points reason I’ll never be allowed to play World of Warcraft.

These discounts aren’t monthly or anything; the more you transfer, the cheaper your service gets, no matter how long it takes. It’s also permanent for the life of your bandwidth account and applies immediately to new sites you create on the same account. We’ll be posting information about this on our public site over the next couple of days.

This is absolutely not designed to compete with the bajillion-gigs-for-$9.95-a-month* plans out there. Those plans are based on overselling, not actual cost of services. We’ll never go down that road. Instead, we’ve been waiting a long time to get the purchasing power and volume discounts needed to make hosting even more affordable for our members, and it’s very exciting that after six years, we’re finally here.

So your mission, should you choose to accept it, is to get out there and use more bandwidth. Hey, you can afford it!

*Bajillion-gigs-for-$9.95-a-month offer not available for websites containing graphic files, videos, scripts, large downloads, small downloads, text, HTML documents, or websites that anyone might want to visit.

UPDATE: We’ve added a Bandwidth Calculator to our public site to help you figure out how much certain bandwidth costs. Turns out the formula is given as:

Unfortunately, this equation is problematically difficult. Thank goodness for numerical approximation! Since both bandwidth and currency are quantized, not continuously variable anyway, I believe this to be sufficiently accurate for estimating purposes.

On October 12th, our pricing for all the top-level domains we support will increase from $7.50/year to $7.99/year for all domain registrations, renewals, and transfers. All other pricing, including RespectMyPrivacy.COM, will remain the same.

It is possible to renew domains for multiple years (up to 10 total), so if you know you’re keeping your domain and you want to get those years at the best possible price, we urge you to renew before October 12th. You won’t lose any current registration on your domain; renewals will add right on to the end of the current expiration date.

The sky is not falling, and the world will definitely not end if you don’t renew your domains before the price increase, but I know a lot of our members like to optimize their costs down to the last penny of savings.

• About 25% of our members have ever interacted with our support system.
• About 14% of our members have opened more than one support issue.
• Over 50% of our support issues come from under 5% of our members.
• Providing support is our single largest recurring monthly cost line item. That means it costs us more than we pay for bandwidth.
• The cost of providing support is growing at a faster rate than any other cost.

In other words, 75-95% of our members are paying to subsidize the support requirements of a small minority. Although that raises some serious issues about fairness, it isn’t really unusual; most hosting companies operate exactly that way: they build a certain amount into their monthly fee to each user to subsidize the cost of providing support to the small portion of their userbase that requires it.

However, as you know, we don’t have a monthly fee. That means the support subsidy has to come from somewhere else. At NearlyFreeSpeech.NET, it’s currently coming out of the R&D budget. So while our entire member base may not be paying a fee for support they may or may not use, you most definitely are paying for it in terms of reduced functionality and delays in rolling out the new features and services you want.

This is particularly frustrating since we’ve always presented ourselves as a no-frills, do-it-yourself host with limited support that does a lot of R&D that you simply don’t get anywhere else.

Obviously, we need to make a change to get back to our core values.

Effective immediately, we are changing our standard support hours and implementing an optional “Extended Support” plan.

Our new support hours for standard issues are:

• Monday through Friday, 10am - 6pm
• Saturday and Sunday, 12pm - 4pm, Pacific Time

With Extended Support, hours for standard issues are:

• Monday - Friday, 8am - 8pm
• Saturday - Sunday: 11am - 5pm, Pacific Time

Standard issues are those typical questions we receive that don’t pertain to downtime or other serious problems with the service. We’ll be continuing to do our best to address emergency issues as quickly as possible, regardless of our stated hours.

Wider hours are not the only benefit to Extended Support:

• Members with extended support will have their standard issues prioritized over other standard issues.
• To ensure timely and fair support to all members, we may occasionally delay response to standard issues raised by those members who submit an above-average number of standard issues. Members with extended support who submit an above-average number of standard issues will encounter shorter delays than they otherwise might.

The cost for Extended Support is $1.00 per month. Unlike most of our services, it’s purchased one month at a time and billed in advance. Once activated, it can be set not to renew at any time without affecting any time remaining in the current billing period. Since for most moderate users, the bulk of their support issues arise in the first month while they are getting set up, they can add Extended Support if they wish, and then remove it once things are humming along smoothly.

The revenues we generate from Extended Support will go directly to pay the people who provide support. Once we have enough Extended Support members to cover the costs of the current extended hours, we’ll start widening the extended hours. Should we reach 24×7 extended coverage, we can evaluate widening the scope of the support we provide.

I want to emphasize that Extended Support is optional. If you don’t use our support, this doesn’t affect you at all, and no one is required to sign up for it. If you value our support, and you want to show it, this is an excellent way to do so.

This is definitely a sort of social experiment, to see if the people who interact with our support system find enough value in that system to cover the associated costs. If it is successful, we will be able to extend and improve that system in a way that would otherwise be impossible. If not, we will have to evaluate our support offerings and look for ways to constrain it to a manageable cost.

However, as time goes on, MySQL is becoming more of an issue. Not only is each successive release more demanding in terms of CPU and memory, but as time goes by, the things our members are doing with MySQL become more intensive as well.

One good example of this is InnoDB tables. When we started, the current version of MySQL was 3.x and there was no such thing as InnoDB. Even when it became part of the distribution in 4.x, it wasn’t used very often (although its mere presence causes MySQL RAM usage to skyrocket). Now, some applications, like MediaWiki for example, will use InnoDB tables by default. Similarly, we would like to start making MySQL 5 available, but it promises to ratchet up the resource requirements once again.

Originally, we planned to extend the storage-based model to MySQL, but what we’ve found is that not only is this not a great model for MySQL, but the accounting mechanism would waste a lot of disk I/O, which would slow down everyone’s databases. By “not a great model” I mean that our MySQL servers don’t share a storage pool the way our web cluster does; each MySQL server has an independent local RAID to maximize performance. Due to the size of disks these days, these servers always run out of I/O capacity or CPU or memory well before they run out of disk space. MySQL is becoming progressively more expensive to maintain, and we need to find a way to recover the cost of its servers that is based on the resources that actually do run out.

Therefore, we are implementing the following pricing scheme for MySQL, effective January 1st, 2007:

• Your first MySQL process will be charged at a base rate of $0.01 per day.
• Your second (and successive) MySQL processes (if any) will be charged at a base rate of $0.02 per day.*
• If a MySQL process has InnoDB tables enabled, it will be charged an additional $0.01 per day.
• If a MySQL process is in the top 10% of MySQL CPU users for a given day, we will apply a $0.01 surcharge for that day.**

This means that roughly 90% of our members will see a MySQL fee of $0.01 per day (about $3.65 a year over existing prices). In the worst-case scenario, a member’s second MySQL process that has InnoDB enabled and is in the top 10% every single day would top out at $0.04 per day (about $14.60 per year), which is not unreasonable for a MySQL process that’s obviously seeing consistent high-volume usage.

*Once these changes are implemented, we will remove the restriction of one MySQL process per member.

**It is unlikely we will have the mechanism to charge for this in place by 1/1/2007, but if not, it may go into effect at any time after that.

If this is the only post you see in a category, it just means the category hasn’t been assigned any real posts yet.