Knowledge Base - FAQ
Dedicated Servers vs. AWS Cloud Hosting
Several people have asked us how our dedicated servers compare to the larger instances offered by Amazons EC2 Cloud. (AWS)
One of the things people really like about Amazon's EC2 is that you pay by the hour for the server. There are no long term obligations, contracts to sign, or commitments to make. Thus we would like to proudly point out that Kattare has had a similar approach since 1997. We have always prorated service starts and stops and we have never forced anyone into a long-term contract to provide service. At Kattare, you have the same flexibility to leave whenever you need to. However, we're confident that with our excellent service, you'll want to stay.
For the following tables we've taken each of our dedicated server options and done a one to one comparison as best we could. AWS EC2 pricing data came from http://aws.amazon.com/ec2/pricing/ and was current as of March 2012. AWS EC2 instance sizes and configurations came from http://aws.amazon.com/ec2/instance-types/ and was current as of March 2012.
We have not compared the t1.micro instance because, well, a Nokia N900 cell phone has more processing power. And in another article, they estimate a t1.micro to have 0.35 ECU(1), which is roughly a 350 Mhz CPU: http://huanliu.wordpress.com/2010/09/10/amazon-ec2-micro-instances-deeper-dive/.
We have not compared instances larger than the m1.xlarge because we do not offer pre-configured options for very large specialty servers. If you are interested in a very large server, please contact us. We are always happy to put together a quote, and we are confident the pricing will compare favorably, much like you see below.
As you can see, for the average dedicated server customer, moving to the cloud makes little to no sense financially or in terms of performance. There is a limited subset of situations in which cloud hosting genuinely makes sense, for example:
Any questions? We'd love to hear from you. Contact Us.
Interested in ordering? We'd love to show you what we can do. Sign up today!
Footnotes:
(1) Amazon somewhat hides the speed of their CPU resources by calling them EC2 Compute Units. In practice, 1 EC2 Compute Unit is roughly equal to a 1.0 GHz 2007 Intel Xeon processor. Thus 2 EC2 Compute Units would be a 2.0 GHz CPU, etc. The frustrating thing about compute units and estimates is that it's all best case. They don't tell you up front, but at peak times Amazon EC2 instances experience "CPU Steal". This is a varying percentage of CPU power stolen from your instance and given to another instance on the same host server. At peak times CPU steal peaks, and this is typically when you also need it the most. This is an important point - because what this means is that you will frequently find yourself purchasing more instances or larger instances than you really should need, just to make up for resources stolen by other users instances.
(2) POOF storage means that when the underlying server hardware crashes or reboots, you lose everything you've stored. You also lose everything you've stored when they do software or firmware patches, or if you stop (pause) your server. Apparently, that's just "The Cloud WayTM". The only way around this is to use slower EBS storage volumes, for which you pay $0.10/mo/GB.
(3) When you use EBS, which you have to do if you want persistent storage, ALL OF YOUR DISK IO FLOWS ACROSS THE NETWORK via iSCSI. This is one of the core reasons MySQL and other disk intensive services do not function well in EC2, the network restriction throttles disk access, and adds latency to every single disk operation. To speed things up perceptually, EBS also does some nasty tricks, like telling your OS things have been committed to disk before they actually have, rendering ugly corruption and data loss in crash events.
(4) Nightly Backup calculations assume you're using nightly S3 snapshots of your EBS volumes at $0.125/mo/GB and keeping 30 days worth of snapshots.
One of the things people really like about Amazon's EC2 is that you pay by the hour for the server. There are no long term obligations, contracts to sign, or commitments to make. Thus we would like to proudly point out that Kattare has had a similar approach since 1997. We have always prorated service starts and stops and we have never forced anyone into a long-term contract to provide service. At Kattare, you have the same flexibility to leave whenever you need to. However, we're confident that with our excellent service, you'll want to stay.
For the following tables we've taken each of our dedicated server options and done a one to one comparison as best we could. AWS EC2 pricing data came from http://aws.amazon.com/ec2/pricing/ and was current as of March 2012. AWS EC2 instance sizes and configurations came from http://aws.amazon.com/ec2/instance-types/ and was current as of March 2012.
We have not compared the t1.micro instance because, well, a Nokia N900 cell phone has more processing power. And in another article, they estimate a t1.micro to have 0.35 ECU(1), which is roughly a 350 Mhz CPU: http://huanliu.wordpress.com/2010/09/10/amazon-ec2-micro-instances-deeper-dive/.
We have not compared instances larger than the m1.xlarge because we do not offer pre-configured options for very large specialty servers. If you are interested in a very large server, please contact us. We are always happy to put together a quote, and we are confident the pricing will compare favorably, much like you see below.
| Kattare Dedicated 1.1 | AWS EC2 m1.small | |
| CPU | Dual 2.8 Ghz (5.6 ECU(1)) |
Single 1.0 Ghz (1 ECU(1)) |
| RAM | 2 GB | 1.7 GB |
| Disk | 74 GB (SATA RAID 1) | 160 GB (POOF(2)) |
| Base Monthly Fee | $150.00 | $57.60 (US-EAST) |
| First 200 GB Data Transfer | $0.00 | $23.88 |
| Persistent Storage (74GB) | $0.00 | $7.40 (EBS(3)) |
| Nightly Backup (30 day, 20 GB) | $0.00 | $75.00 (S3(4)) |
| Nightly Backup (30 day, 74 GB) | $0.00 | $277.50 (S3(4)) |
| The Bottom Line | (w/ 20GB S3 Backups) | |
| Cost | $150.00 | ~$163.88 |
| Performance | ~5.6x More CPU Cycles ~2x Higher DISK IO(2) |
| Kattare Dedicated 2.1 | AWS EC2 m1.medium | |
| CPU | Dual 2.8 Ghz (5.6 ECU(1)) | Single 2.0 Ghz (2 ECU(1)) |
| RAM | 4 GB | 3.75 GB |
| Disk | 146 GB (SATA RAID 1) | 410 GB (POOF(2)) |
| Base Monthly Fee | $200.00 | $115.20 (US-EAST) |
| First 300 GB Data Transfer | $0.00 | $36.00 |
| Persistent Storage (146 GB) | $0.00 | $14.60 (EBS(3)) |
| Nightly Backup (30 day, 20 GB) | $0.00 | $75.00 (S3(4)) |
| Nightly Backup (30 day, 140 GB) | $0.00 | $525.00 (S3(4)) |
| The Bottom Line | (w/ 20GB S3 Backups) | |
| Cost | $200.00 | ~$240.80 |
| Performance | ~2.8x More CPU Cycles ~2x Higher DISK IO(2) |
| Kattare Dedicated 3.1 | AWS EC2 m1.large | |
| CPU | Quad 2.6 Ghz (10.4 ECU(1)) | Dual 2.0 Ghz (4 ECU(1)) |
| RAM | 8 GB | 7.5 GB |
| Disk | 146 GB (SATA RAID 1) | 850 GB (POOF(2)) |
| Base Monthly Fee | $250.00 | $230.40 (US-EAST) |
| First 400 GB Data Transfer | $0.00 | $48.00 |
| Persistent Storage (146 GB) | $0.00 | $14.60 (EBS(3)) |
| Nightly Backup (30 day, 20 GB) | $0.00 | $75.00 (S3(4)) |
| Nightly Backup (30 day, 140 GB) | $0.00 | $525.00 (S3(4)) |
| The Bottom Line | (w/ 20GB S3 Backups) | |
| Cost | $250.00 | ~$368.00 |
| Performance | ~2.6x More CPU Cycles ~2x Higher DISK IO(2) |
| Kattare Dedicated Corporate |
AWS EC2 m1.xlarge | |
| CPU | Hex 2.6 Ghz (15.6 ECU(1)) | Quad 2.0 Ghz (8 ECU(1)) |
| RAM | 16 GB | 15 GB |
| Disk | 292 GB (SATA RAID 10) | 1690 GB (POOF(2)) |
| Base Monthly Fee | $300.00 | $460.80 (US-EAST) |
| First 500 GB Data Transfer | $0.00 | $60.00 |
| Persistent Storage (292 GB) | $0.00 | $29.20 (EBS(3)) |
| Nightly Backup (30 day, 20 GB) | $0.00 | $75.00 (S3(4)) |
| Nightly Backup (30 day, 250 GB) | $0.00 | $937.50 (S3(4)) |
| The Bottom Line | (w/ 20GB S3 Backups) | |
| Cost | $300.00 | ~$625.00 |
| Performance | ~2x More CPU Cycles ~2x Higher DISK IO(2) |
As you can see, for the average dedicated server customer, moving to the cloud makes little to no sense financially or in terms of performance. There is a limited subset of situations in which cloud hosting genuinely makes sense, for example:
- If you are running a large cluster of servers, all of which have a copy of the same data, and none of which you care about when that data goes POOF(2). This makes some sense because then you do not have to pay for EBS storage or S3 backup snapshots.
- If you are anticipating very large up and down swings in traffic to your site, in which case you can program your site to auto-scale instances up and down based on your traffic. Note that programming your site for EC2 auto-scaling is a daunting and error-prone undertaking. You will need significant programming resources to do it correctly.
- If you are a research institution that has a lot of computing that needs to be done and you need it ASAP. EC2 is excellent for this. You can spin up 1000 instances to get a job done, then shut them all down a few hours later.
- If you need to run a database with any significant number of INSERTS or UPDATES. Performance of high disk IO applications in the cloud is abysmal. Even if you stripe 4-5 EBS volumes together, you will always hit a fairly low performance bottleneck because ALL disk IO is network bound. Check out this article on Agile Testing: An ode to running a database on bare metal.
- If your application depends on any one server for operation. IE, if your application goes down when any single server in your cluster becomes unavailable. Per-instance failure in the AWS Cloud is more than 4x that of a dedicated hardware server. This due in large part to the bad-neighbor effects of having a lot of virtual servers on the same piece of hardware. Before you move to the cloud, you MUST re-write your application to not depend on any one server.
- If you depend on persistent storage. AWS's persistent storage solution (EBS) has a 200% higher failure rate than a single dual-disk RAID 1 setup. (yes, EBS is TWICE as likely to fail!) They publish it right on their site. (http://aws.amazon.com/ebs/) They claim a 0.5% annual EBS failure rate, and they claim a 4% failure rate on commodity hard drives. What they fail to mention is that server hard drives are not commodity drives, AND for a simple RAID 1 setup of two hard drives to fail you have to calculate out the risk of both drives failing at the same time. This article on wikipedia (http://en.wikipedia.org/wiki/Standard_RAID_levels) does the math for us, clearly indicating a 0.25% chance of failure of a RAID 1 setup over a THREE year span. (vs 0.5% rate on EBS over a ONE year span.) And we're not even getting into the volume corrupting software/firmware issues EBS has had over the years...
Any questions? We'd love to hear from you. Contact Us.
Interested in ordering? We'd love to show you what we can do. Sign up today!
Footnotes:
(1) Amazon somewhat hides the speed of their CPU resources by calling them EC2 Compute Units. In practice, 1 EC2 Compute Unit is roughly equal to a 1.0 GHz 2007 Intel Xeon processor. Thus 2 EC2 Compute Units would be a 2.0 GHz CPU, etc. The frustrating thing about compute units and estimates is that it's all best case. They don't tell you up front, but at peak times Amazon EC2 instances experience "CPU Steal". This is a varying percentage of CPU power stolen from your instance and given to another instance on the same host server. At peak times CPU steal peaks, and this is typically when you also need it the most. This is an important point - because what this means is that you will frequently find yourself purchasing more instances or larger instances than you really should need, just to make up for resources stolen by other users instances.
(2) POOF storage means that when the underlying server hardware crashes or reboots, you lose everything you've stored. You also lose everything you've stored when they do software or firmware patches, or if you stop (pause) your server. Apparently, that's just "The Cloud WayTM". The only way around this is to use slower EBS storage volumes, for which you pay $0.10/mo/GB.
(3) When you use EBS, which you have to do if you want persistent storage, ALL OF YOUR DISK IO FLOWS ACROSS THE NETWORK via iSCSI. This is one of the core reasons MySQL and other disk intensive services do not function well in EC2, the network restriction throttles disk access, and adds latency to every single disk operation. To speed things up perceptually, EBS also does some nasty tricks, like telling your OS things have been committed to disk before they actually have, rendering ugly corruption and data loss in crash events.
(4) Nightly Backup calculations assume you're using nightly S3 snapshots of your EBS volumes at $0.125/mo/GB and keeping 30 days worth of snapshots.
Do you allow cron/at jobs?
Absolutely. CRON and AT jobs can be specified on the command
line on all accounts that support shell access.
If you already have an account, there is an FAQ Entry specifically about how to use CRON.
If you already have an account, there is an FAQ Entry specifically about how to use CRON.
Do you provide dial-up service?
We no longer provide any dial up services, sorry.
Do you support Java Servlets?
Yes we do!
We're currently using Jakarta-Tomcat as the default servlet engine. (Though there are more JVM options with the Corproate Plus account.) All servlet accounts get their own JVM so you can start, stop, and configure your Tomcat servlet engine to your heart's content.
Do you support PHP?
Yes, we've been supporting PHP and all of the popular modules since 1997.
How can I get my website set up with it's own IP address?
Our standard website configuration is called name-based virtual
hosting. This means that several sites usually share an IP
address. If you need your own IP address for your site,
email support@kattare.com
and let them know. They'll set you up. The cost is $5/mo per IP address.
How long have you been around as a company?
Kattare Internet Services opened it's doors officially in
June, 1997. We have grown steadily since then and have
strived to maintain the same personal service as the day
we opened. We currently have the tech support and system
administrators to make Kattare the best all
around Web Service that you can trust.
The question many ask is, should you trust Kattare Internet
Services?. We think you should. The reward is great service
for a great price. We are
very proud of Kattare and the service we provide to our
customers, feel free to contact any of our customers on
either our business page or personal pages and ask them how
they like our service. Our customers are our best salesmen!
How many times has Kattare been ranked by Netcraft?
Netcraft started tracking Kattare in late 2006. Since then Kattare has been
in the "Top 10 Most Reliable Hosts" ranking 14 times.
January 2007
February 2007
May 2007
February 2008
May 2009
December 2009
June 2010
July 2010
January 2011
March 2011
May 2011
June 2011
October 2011
April 2012
We're at a little bit of a disadvantage, as we're a West Coast host and Netcraft's monitoring points are Atlantic-centric, but we definitely have a good track record.
January 2007
February 2007
May 2007
February 2008
May 2009
December 2009
June 2010
July 2010
January 2011
March 2011
May 2011
June 2011
October 2011
April 2012
We're at a little bit of a disadvantage, as we're a West Coast host and Netcraft's monitoring points are Atlantic-centric, but we definitely have a good track record.
How often do you backup your systems?
As of this writing (Apr 8, 2003) our backups work as follows:
Core File Servers: (users home directories)
- backed up nightly
- backups are kept 90 days
- restores necessary for user error in this category are $10. - restores necessary for sysadmin error are free.
Core Mail Servers: (users mail/pop3 inboxes)
- backed up nightly
- backups are kept 30 days
- restores necessary for user error in this category are $10. - restores necessary for sysadmin error are free.
Core Database Servers: (users MySQL/Postgres databases)
- backed up nightly
- backups are kept 60 days
- in addition, we now replicate the database servers in real-time, in case of system failure. You can read more about this here:
http://www.kattare.com/readnews.kvws?id=134
- restores necessary for user error in this category are $15. - restores necessary for sysadmin error are free.
Web Servers: (apache/tomcat/jboss/resin front end servers)
- backed up nightly
- backups are kept 60 days
- software on these platforms do not change often
- restores are free of charge, as they're practically guaranteed to be sysadmin errors.
Dedicated Servers: (servers we build and maintain)
- backed up nightly
- backups are kept 60 days
- restores necessary for user error in this category are $15. - restores necessary for sysadmin error are free.
Core File Servers: (users home directories)
- backed up nightly
- backups are kept 90 days
- restores necessary for user error in this category are $10. - restores necessary for sysadmin error are free.
Core Mail Servers: (users mail/pop3 inboxes)
- backed up nightly
- backups are kept 30 days
- restores necessary for user error in this category are $10. - restores necessary for sysadmin error are free.
Core Database Servers: (users MySQL/Postgres databases)
- backed up nightly
- backups are kept 60 days
- in addition, we now replicate the database servers in real-time, in case of system failure. You can read more about this here:
http://www.kattare.com/readnews.kvws?id=134
- restores necessary for user error in this category are $15. - restores necessary for sysadmin error are free.
Web Servers: (apache/tomcat/jboss/resin front end servers)
- backed up nightly
- backups are kept 60 days
- software on these platforms do not change often
- restores are free of charge, as they're practically guaranteed to be sysadmin errors.
Dedicated Servers: (servers we build and maintain)
- backed up nightly
- backups are kept 60 days
- restores necessary for user error in this category are $15. - restores necessary for sysadmin error are free.
What type of SSL encryption does Kattare use in it's secure pages?
A 1024-bit RSA encrypted key and a secure socket layer (SSL) certificate signed by Geotrust. (Geotrust is a division of Equifax.) The SSL is RSA-encryption based. Our secure server software is Apache-SSL.
What versions of MySQL do you support?
As of this writing (Oct, 2010) we're supporting MySQL 4.0.x and 5.1.x. Because of
compatibility issues with various java connectors, we'll likely be keeping
4.0.x around for years to come.
[Submit a Question]
Indicates a question that only Kattare members can view. (For security reasons, sorry.)
You are at: Kattare Hosting: Knowledge Base: Pre Sales
Comments or feedback? Email: support@kattare.com
Legal: Terms of Service | Privacy Policy | DMCA Procedure | Trademarks
Comments or feedback? Email: support@kattare.com
Legal: Terms of Service | Privacy Policy | DMCA Procedure | Trademarks