Web Hosting Talk : Web Hosting Main Forums : Cloud Hosting : Sudden increase of usage reports on Rackspace Cloud Sites

[Pulsorock]

Have any Rackspace Cloud Sites users have noticed this?

At the beginning of this month they changed the way they calculate usage and now they're reporting a huge increase of usage (this meaning a huge increase of the billing costs too).

Notice the attached graph. The site is getting the same traffic (based on web stats and Google Analytics), but Rackspace is reporting a huge increase after September 3th.

Anyone else noticed this?

[Winky]

maybe they're charging both inbound and outbound whereas before they only charged one way? or, they can be charging for intranetwork traffic...so traffic between your host and the storage server. what were the terms that they changed?

[MikeDVB]

Have you contacted them to ask them what the difference is? Maybe there is some plausible explanation like maybe they weren't previously accurately recording traffic data.

[dazmanultra]

It could be that they were under-charging in the past and have fixed their billing.

[Pulsorock]

That's what they are telling me. That they were "under-charging" and in September 3, they "fixed it" with the actual usage. My disagreement is that they "fix" this without notifying their clients, and that sudden increase scares anyone.

My other concern is that they don't provide a way to see compute cycle usage in detail. There's no way to see what process or pages are making more use of the CC. Unlike the webstats were you can see what pages are making most hits/bandwidth usage, with compute cycles you can't know what pages/process are making more use of it, so the user can optimize and reduce in any way this CC usage.

And with this new calculation algorism, 10,000 compute cycles are too little, wish they could provide maybe extra packages with more bandwidth/compute cycles. The 1 Cent per CC after the 10,000 limit can increase my bill over 400%-500%

[dazmanultra]

What sort of traffic does your site get? If you have a consistently busy site then you may find a dedicated server set up works out better than utility billing (which is better if you have transient periods of business - e.g. a particular day of the week, or particular time of day where the load is much higher than at other times).

[Pulsorock]

We have several sites, but one of them is a heavy visited news/tv site, with an average 30K-50K visitors daily (the graph attached on the first post is from another personal account, I have two accounts, personal and clients).

We used to have dedicated servers for this client (hosted locally in a data center building), but moved to cloud for the ability of scale to our needs in case of a traffic spikes, since our previous setup didn't have any load balancer and during spikes the site was ultra slow/unavailable. Now on cloud sites, we have the ability to scale and use their cloud infrastructure to be always available when we need it. For example, last month during Hurricane Earl on the Caribbean, we had a traffic spike of almost 200K visitors in one day. With our previous dedicated servers setup, we wouldn't be able to serve all that visits, but with cloud we did.

[MikeDVB]

Originally Posted by Pulsorock

That's what they are telling me. That they were "under-charging" and in September 3, they "fixed it" with the actual usage. My disagreement is that they "fix" this without notifying their clients, and that sudden increase scares anyone.

I don't see why they would need to notify you that they've fixed something that wasn't working correctly. I mean if their CPUs were only working at 1/2 speed and then suddenly they fixed that and they were running at full speed you wouldn't be complaining.

I mean I can understand where you're coming from - but the way I look at it - I'd be happy that I got to use more than I was paying for in the past.

Originally Posted by Pulsorock

My other concern is that they don't provide a way to see compute cycle usage in detail. There's no way to see what process or pages are making more use of the CC. Unlike the webstats were you can see what pages are making most hits/bandwidth usage, with compute cycles you can't know what pages/process are making more use of it, so the user can optimize and reduce in any way this CC usage.

I don't know of any providers that do show this sort of detail but admittedly I don't have a ton of experiences in this part of the industry.

Originally Posted by Pulsorock

And with this new calculation algorism, 10,000 compute cycles are too little, wish they could provide maybe extra packages with more bandwidth/compute cycles. The 1 Cent per CC after the 10,000 limit can increase my bill over 400%-500%

So perhaps you should look for an alternative provider or something else that will fit your needs that you can afford?

[Enigma5]

I know this is an old thread, but I am having the same issue with Rackspace. My bill has gone from the simple $100 a month to 3-4x that. I've been searching for more information about this and I'm surprised many people are not complaining because the 10,000 cycles is not enough and people's bills have to be excessive like mine has been.

I e-mailed Rackspace asking them about this and they sent me a Compute Cycle usage report, which shows the Compute Cycle usage for the past 7 days. This does not show complete compute cycles, from what Rackspace has told me this report shows the top 100 urls. I have attached the report to this thread.

This is the message they sent me along with the report..

Hello,

I am attaching two documents to help you better understand your compute cycle usage. The attachment labeled: compute-cycle-report-www.thinkcomputers.org-analyzer.out-2010092222.csv is a compute cycle usage report for the last 7 days. To give you more information about what you are reading, the CC column is the total usage for each page on the site. AVG is the average compute cycle per HIT (page visit).

To view the traffic that is visiting the site, I've included an Urchin stats page for the site (Urchin_2202.csv.). You may also review all stats for your site by clicking on the Urchin 6 button on the General Setting tab of the control panel.

Finally, I would like to make you aware of an enhancement that was made to the usage reporting system for Cloud Sites. On September 1, 2010, we began displaying and invoicing usage information from a stable and accurate usage reporting system. This new system has provided stable Urchin traffic reporting and raw logs for processing web statistics since May 26, 2010.

Please let us know if you have any questions. As always, we are here to assist you 24 hours a day via our live chat, tickets, or telephone support. I have also included a few FAQ's on the the enhancements that were made. Have a great day and thank you for being a customer.

I have talked to support again this month after getting my bill and I asked for a full compute cycle report.

[Pulsorock]

That exact thing happened to me. I searched online and was amazed too that didn’t found more threads about it, considering the huge bill increase this would be to customers.

We started using Rackspace about 3 months ago, so our base price is $149 which includes 50GB space, 500GB bandwidth (BW) and 10,000 Compute Cycles (CCs). Before we registered, a Rackspace representative told me via chat when I asked him about how did Compute Cycles worked and if it was enough, that "only about 4% of our thousands of cloud sites customers exceeds the 10,000 cycles each month". The first month we used under the 10,000 CCs and under 500GB of BW. The next month the increase in calculations came in, and we saw a increase of almost 60,000 CCs and 600GBs of BW. That meant an increase of over $500 each month. Paying $149 and then over $650 for about the same usage and traffic is outrageous.

I told them this via a ticket that was escalated to an account manager, and basically they told me this: "[...] on September 3rd we did make some changes to the usage system. As a result of rapid growth in the Cloud, the previous usage reporting system was not able to continuously process the amount of data that it needed to record. There were periods of time during which the system could not complete the processing of data, and usage data was lost. The lost data resulted in a lower amount of usage being reported and invoiced. A stable environment is in place to accurately report usage on a daily basis. As of September 3, 2010, we began invoicing and displaying usage from this stable environment."

Then they proceed to provide me a CSV document just like yours, that is a sample of 7 days of the top 100 Files that generate the most compute cycles for one of my sites with the most traffic. With that file I did optimized some of the files, reducing to about 52,000 CCs usage this last month, but still I'm paying over $430 in CCs overage fees alone.

I could understand that this is their way of calculating the usage and billing, but 10,000 Compute Cycles is NOT enough (unless you want to host a blog that receives 500 visits daily) and I doubt with that small traffic someone would consider Cloud Hosting. This Cloud Hosting service is targeted to large sites that want to have the Cloud infrastructure to support large visits and traffic spikes, and this 10,000 CCs is not enough. Also, being charge 1 cent for each extra Compute Cycle is way too expensive. I wish they could offer bundles of discounted packages if you go over their baseline limit, but they don't.

Hope this helps!

[Enigma5]

Thanks for the quick reply, Yes I am still amazed that there isn't much more on this from anyone else using the service. After doing a little more digging I found this thread where Chad who is the Product Manager for cloud sites says this...

How many compute cycles will my applications use?
Since web applications vary so greatly, it's hard to make a perfect guess. However, there are some guidelines that can help. First, you can think of 10,000 compute cycles as being about the same processing power as you'd get from an average dedicated server or Amazon EC2 instance. For example 10,000 compute cycles would power:

about 2.1 million page views using a database-driven content management system
about 11 million page views of rackspacecloud.com
about 25 million requests for a static 15KB image

Another quick note; only approx 5% of our entire customer base exceed the standard 10,000 compute cycle allotment. Typically extremely high traffic web sites.

Seems like bs to me, according to google here is the pageviews for my 3 sites for the last billing period.

Site 1: 566,464 Pageviews
Site 2: 6,009 Pageviews
Site 3: 1,919 Pageviews

Anyways I'm waiting for the full report I asked for if they decide to sent it to me. I'll try and keep you updated. Do you plan on changing hosts?

[Enigma5]

Well the basically sent me the same report....not a full compute cycles report. I asked again for a full report to see the compute cycles I am being charged for and they said this to me...

Hi Bob,

I believe this is the only data we capture in terms of time period, etc... I will double check with support and see if they can assist with your inquiry.

Any update on your end?

[plumsauce]

It would seem as if rackspace calculated the price point and usage metrics using their old system, then changed the system. That would mean that your allowances are based on the old system metrics and not the new system metrics.

In all fairness, they should have adjusted the included allowances to reflect the new system.

[Enigma5]

Originally Posted by plumsauce

It would seem as if rackspace calculated the price point and usage metrics using their old system, then changed the system. That would mean that your allowances are based on the old system metrics and not the new system metrics.

In all fairness, they should have adjusted the included allowances to reflect the new system.

I would have to say so. My main site only does about 100-120k uniques a month and it goes way over the 10k CC allowance.

I would have to think many people are experiencing the same thing.

[giorgio79]

Do you guys use any kind of caching, which writes out php/mysql pages to html so static content is served to the visitor?

I assume this would decrease the compute cycles usage massively.