Comparison and Analysis of Managed DNS Providers

Introduction

This blog post is the culmination of year's effort researching and developing methods for analyzing and comparing managed DNS services. During this time, we provided access to our analysis and solicited feedback from any DNS provider that would listen. We would like to thank our contacts with UltraDNS, Cotendo, Amazon Web Services and NetDNA for the feedback they provided. As always, it is our intent to provide objective, fair and actionable analysis. We have not been paid by anyone to conduct this testing or write this post. If you have feedback, we'd like to hear it.

This blog post is also intended as an introduction to a new monthly report entitled State of the Cloud - DNS available for purchase on our website. The full version of the August 2012 edition of this report is available for free in html format or pdf format (10MB). Future editions of this report will include both free and premium editions where the free version will include everything but some of the more advanced content such as marketshare analysis. This blog post provides an introduction to and summary of the August 2012 edition of this report.

View Full Report - August 2012

What is DNS?

Domain Name System or DNS for short, is the part of the Internet that lets users access websites (and other Internet services) using easy to remember words and phrases called hostnames like amazon.com or google.com. Without DNS, users would be required to use cryptic numeric-based identifiers called IP addresses (e.g. 72.21.194.1). When a user types a hostname into their browser address bar, one of the first steps undergone is to translate the hostname to an IP address. This translation process involves querying a DNS server that has been assigned responsibility for that hostname. These are called authoritative DNS servers. If the authoritative DNS server is not accessible, the browser will be unable to resolve the IP address and display the website.

DNS server software is freely available and is not overly complex to setup or run. The core functionality of a DNS server is simple… the translation of hostnames to IP addresses. It requires only minimal bandwidth and CPU resources to maintain a DNS server. Many organizations host their own DNS servers without much effort.

Managed DNS

Managed DNS is a service that allows organizations to outsource DNS to a third party provider. There many reasons why an organization may elect to outsource DNS hosting... here are a few:

• Simplicity Organizations don't have to worry about setting up and maintaining their own DNS servers. Management of DNS records is also easier because providers enable this using a simple browser-based GUI or API
• Performance Providers that specialize in DNS have often invested significant time and capital setting up global networks of servers that can respond quickly to DNS queries regardless of a user's location
• Availability Managed DNS providers employ dedicated staff to monitor and maintain highly available DNS services and are often better equipped to handle service anomalies like DDOS attacks
• Advanced Features Managed DNS providers often offer features that are not part of the standard DNS stack such as integrated monitoring and failover and geographic load balancing

Whatever the reasons are, managed DNS is a fast growing sector in the cloud.

Enterprise versus Self Service

Managed DNS providers can be generally divided into two categories:

• Enterprise providers typically offer more advanced features, personalized support and account management, and often have larger DNS server networks. These providers typically utilize a formal sales and contract negotiation process for new customers where pricing is variable depending on the customer's negotiating prowess, usage volume and term commitment. Pricing is typically orders of magnitude higher than self service providers. Some enterprise providers offer low volume, low cost introductory packages that are lead-ins to their standard service offerings
• Self Service providers typically offer simple, contract free, self management DNS services. Pricing is often catered more towards smaller organizations with limited budgets. Self service providers usually (but not always) have smaller DNS server networks and offer fewer advanced features. Based on our analysis, these services are generally as reliable as enterprise services

After speaking with multiple enterprise providers, it is our impression that they generally consider self service providers as non-competitors targeting a different customer demographic.

Comparing Managed DNS Services

Comparing DNS services is not as simple as running a few benchmarks and calling it good. There are multiple criteria where comparisons may be drawn. In this post, we'll present some criteria we believe to be relevant, the techniques we have used to implement them, and the resulting analysis. The following DNS providers are included:

• Neustar UltraDNS is one of the oldest managed DNS providers founded in 1999. Their network includes 16 DNS POPs (points of presence) on 6 continents. UltraDNS is a leading provider in marketshare with 403 of the Alexa top 10,000 sites according to our recent analysis
• Dyn has evolved over the years from offering various free DNS services to its current form as an enterprise DNS provider. Although they still support a self service DNS under the DynDNS brand, our analysis includes only their enterprise service. The Dyn network consists of 17 DNS POPs in 4 continents. Dyn's enterprise service is slightly behind UltraDNS in marketshare with 319 of the Alexa top 10,000 sites according to our analysis.
• Cotendo/Akamai Cotendo was acquired by Akamai in 2012. The Cotendo DNS network consists of 29 DNS POPs in 5 continents. Combining Akamai and Cotendo DNS makes them the leading provider in marketshare for Alexa top 1,000 sites according to our analysis. Akamai's DNS service, Enhanced DNS currently utilizes different DNS infrastructure from Cotendo and is presented separately in this post
• AWS Route 53 is part of the Amazon Web Services suite of cloud services. It launched in 2011 and is the newest service included in this post. Route 53 uses a self-service, low cost model. The DNS network consists of 33 DNS POPs in 5 continents. Route 53 marketshare has grown significantly in 2012 according to our analysis. It currently lacks many of the more advanced features offered by enterprise providers including DNSSEC and integrated monitoring
• easyDNS is a smaller, self-service provider founded in 1998. Their network consists of of 16 DNS POPs in 3 continents
• DNS Made Easy is another smaller, self-service DNS provider founded in 2002. Their network consists of 12 DNS POPs in 3 continents

End-User Performance

There are many factors that affect DNS performance. When a user types a hostname into a browser address bar the path taken for resolving that hostname to an IP address varies between users. Generally, the first point is the user's ISP DNS resolver. These are specialized DNS servers that cache lookups and are used only for DNS resolving. If this DNS server does not have an answer, it will query the next level up (often another ISP DNS server), and this process continues until the authoritative server is queried. This process is referred to as a recursive DNS lookup. From an end-user's perspective, DNS performance is the total lookup time including the entire recursive chain. To improve performance, managed DNS providers typically deploy mnay DNS servers globally and use an Anycast network to reduce latency and thus reduce lookup times. A well designed DNS network will generally provide better and more consistent performance for end-users globally. However, because DNS lookups are almost always cached, after the first lookup, DNS generally has little impact on website performance.

To measure end-user DNS performance, we developed a browser based test that measures the time difference between downloading a small (4 byte) file using both cached and non-cached hostnames. We use a special type of DNS record called a wildcard name that allows the test to generate random hostnames that are guaranteed to require an authoritative DNS lookup. Multiple measurements are taken during each test and the median is recorded. We've run this test using thousands of unique users globally every month. We also utilize a Geo IP database to determine where the user running the test is located which allows us to generate the region specific analysis for each DNS service show below.

The following charts display the results of our end-user DNS performance analysis for July 2012. The bars in each chart represent a managed DNS service and a specific geographic region. The vertical axis is the median DNS lookup time in milliseconds (1000 milliseconds equals 1 second) for all users and tests in that region (typically hundreds or thousands of unique users). The line spanning horizontally on the chart represents the median lookup time for all regions. In general, DNS lookup times in the 50-200 millisecond range are very good.

In North America, bandwidth and connectivity are relatively simple and affordable. All providers performed generally well in these regions with limited variation between them.

Performance in Europe regions was only marginally slower than North America and we observed slightly higher performance variation between providers.

Bandwidth and connectivity are significantly more costly and complex in Asia, Oceania, and South America. We observed slower performance and higher variation between providers in these regions.

Synthetic Performance

Another method of measuring DNS performance is to query authoritative DNS servers directly (bypassing recursive lookup chains). This method utilizes test agents located in data centers and provides a more repeatable, consistent and controlled testing environment. However, because these tests are conducted by just a handful of servers located in data centers, and bypasses recursive lookup chains, it is less relevant to the performance an end user would experience. For our testing, we are utilized a network of 110 servers (57 US, 28 EU, 25 AsiaPAC and other locations) conducting tests every 5 minutes. The charts below use the same format as the end-user charts.

Due to the lower cost and easier deployment of bandwidth and connectivity, performance in North American regions was generally good and showed minimal variation between providers. Akamai uses a combination of Anycast and Unicast (non latency minimizing) networks which is likely the reason for their lower performance in this analysis

Performance in Europe was also generally good. Many of our test agents are likely located in close proximity to the DNS servers (i.e. in the same or a nearby data center)

In Asia and Oceania, where bandwidth and connectivity is more expensive and complex to setup, we observed lower performance and higher performance variation between providers

Availability

Managed DNS provider networks consist of many DNS POPs (a single POP consists of 1 or more DNS servers) distributed globally. When used in combination with Anycast routing, these DNS servers can be configured to automatically failover to other POPs. Additionally, the DNS protocol can include more than one authoritative server and incorporates retry logic such that DNS clients will try to query multiple DNS servers until a response is received. These factors enable DNS to provide continual availability as long as at least 1 DNS POP is available.

To calculate availability, we utilized the 110 monitoring agents querying provider DNS servers every 5 minutes throughout the month. Every authoritative DNS server was queried during each test. Two metrics are were captured... availability of at least 1 DNS server and availability of all DNS servers. The former is the more important as it reflects what an actual user would generally experience. All providers offered very high availability. The following table represents the results on this analysis for July 2012:

DNS Propagation Latency

Another comparison criterion is the time required to push a DNS record update to a provider's DNS network. We refer to this metric as DNS Propagation Latency. This criterion is more relevant to an organization making frequent or automated DNS updates, where those updates are critical to some functionality. To measure DNS propagation latency, we used 110 global test agents. At the exact moment a DNS change was submitted, those agents were instructed to directly query provider DNS servers and record the amount of time that change took to complete. Testing was conducted for both primary and secondary DNS (secondary DNS is a replication-only service). The tables below show the results of this analysis (Propagation Latency is a median value for all 110 test agents and all provider DNS POPs):

Secondary DNS is a replication service, DNS propagation times are longer because the secondary service has to wait for the primary service to replicate first. During our testing, DNS Made Easy was the primary DNS service. Not all providers offer secondary DNS service.

DNS Provider Marketshare

Marketshare is a good indicator of the robustness of a provider DNS network. Providers that manages DNS for popular websites have a demonstrated capability to effectively support very high DNS volume.

To capture DNS provider marketshare we determined which DNS providers are used for the top 10,000 Alexa sites (Alexa publishes a list of the most popular websites). Some DNS providers allow customers to mask their DNS servers using custom hostnames (a featured referred to as vanity DNS servers). To include these, we correlate vanity DNS servers using IP address matching (matching class C IP addresses to those of known provider DNS servers). To verify correlations, we perform authoritative lookups of the hostnames using actual provider DNS servers.

We also track provider marketshare distribution changes by comparing provider marketshare this month to that of the previous month. The following are the results of this analysis for the month of July 2012:

Top 20 Provider Alexa 10,000 Changes - July 2012

Confirmed Alexa Top 10,000 Changes - July 2012

Because the makeup of the top 10,000 Alexa websites changes from month to month, the marketshare change analysis may not represent actual provider change. The change metrics above, represent the number of actual confirmed provider changes between July 1 and Aug 1 2012. During this time, we observed significant usage growth for both Route 53 and Dyn.

Features

Advanced features are one of the biggest ways that DNS providers distinguish their services. The following is an overview of a few common features and their associated support with each DNS provider:

Health Checks - DNS Failover

DNS failover involves dynamic DNS resolution based on the availability of target hosts. These hosts are monitored continually by DNS providers using ICMP (ping) or more advanced methods such as HTTP content monitoring. If a primary target host fails a health check, DNS resolution automatically change to a backup target host.

UltraDNS	Dyn	Cotendo	Route 53	DNS Made Easy	easyDNS
Yes	Yes	Yes	No	Yes	Yes

Health Checks - DNS Load Balancing

Like DNS failover, DNS load balancing monitors target hosts. However, with load balancing there target hosts are not considered active or failover - instead, all hosts receive a even (or weighted) distribution of traffic. If a target host goes down, the DNS service will stop sending traffic to it by ceasing to resolve that IP address.

UltraDNS	Dyn	Cotendo	Route 53	DNS Made Easy	easyDNS
Yes	Yes	Yes	No	No	No

Location Based Routing (Geo IP)

Location based DNS routing allows a DNS hostname to resolve dynamically depending on the geographic location of the user (or more specifically, location of the user's DNS resolver). To accomplish this, the geographic location is determined using Geo IP databases like Neustar IP Intelligence (formerly Quova) or MaxMind. This location is then run through custom, user-defined DNS rules that may affect the IP address the hostname resolves to. For example, a hostname might resolve to a server in Singapore for users in Asia, and a server in the US for others thereby improving webpage load times.

UltraDNS	Dyn	Cotendo	Route 53	DNS Made Easy	easyDNS
Yes	Yes	Yes	Yes1	No	No

¹Route 53 provides a unique feature called Latency Based Routing, where DNS resolves to a target host with the presumed lowest latency to the end user (the target host must be in one of 7 AWS data center regions)

Zone Based Routing (Anycast)

Zone Based Routing is functionally similar to Location Based Routing, but instead of using a Geo IP database to determine the user's geographic location, it uses the DNS server that the user is querying. In Anycast networks, this DNS server will typically reside in the same general geographic region as the DNS client. In practice, this limits the number of location specific rules to the number of Anycast zones in the provider's network (typically in the single digits). For example, Dyn's network consists of 7 Anycast zones, thus allowing up to 7 location specific target hosts. Because of this, zone based routing is more limited relative to location based routing. Additionally, zone based routing can be problematic in some geographic regions where Anycast networking is less predictable (i.e. Asia) or when POPs are taken down for maintenance

UltraDNS	Dyn	Cotendo	Route 53	DNS Made Easy	easyDNS
No	Yes (7 regions)	No	No	Yes (4 regions)	Yes (4 regions)

DNSSEC

DNSSEC (Domain Name System Security Extensions) is a specification for securing DNS records. DNSSEC was designed to protect clients from forged DNS responses by digitally signing DNS responses. By checking the digital signature, DNS clients can verify the authenticity of those responses. Usage of DNSSEC is growing, but due to complexity and lack of support is still relatively low. Provider or User Managed DNSSEC Manual generation and management DNSSEC certificates can be cumbersome. Some providers simplify this by generating and deploying certificates automatically thereby eliminating many of the administrative complexities.

UltraDNS	Dyn	Cotendo	Route 53	DNS Made Easy	easyDNS
Yes (provider managed)	Yes (provider managed)	No	No	No	Yes (user managed)

Pricing

Enterprise DNS providers, generally do not disclose pricing publicly. Part of this likely has to do with the negotiable nature of their pricing. In order to determine pricing for these services (where it was not available publicly), we contacted each provider for a pricing quote. Actual pricing may vary depending on a customer's ability to negotiate.

DNS Query Pricing (monthly)

Provider	1 million	10 million	100 million	1 billion	10 billion
AWS Route 53	$0.50	$5	$50	$500	$2,750
UltraDNS	$50-$1951	$865-$1,2002	$2,200-$3,0002	$5,1252	$17,5002
Cotendo	Not offered3	$500	$1,000	$5,000	$10,0000
Dyn	$604	$295 (10 QPS)	$600 (40 QPS)	$2,250 (400 QPS)	$5,495 (4000 QPS)
DNS Made Easy	$2.505	$55	$2187	$1,5207	$7,3707
easyDNS	$9.956	$20	$2008	$2,0008	$20,0008

1. $50 plan includes US and EU DNS POPs only
2. UltraDNS discounts query pricing by up to 50% when bundled with advanced features list below. The prices provided here are based on some bundling
3. The lowest usage tier for Cotendo DNS is 10 million queries/mo
4. Dyn Enterprise DNS Lite - includes 1.2 million queries/mo
5. Must prepay annually
6. Enterprise Plan - Pricing is for 5 million queries/mo
7. Based on $1500/yr corporate membership (includes 50 million queries/mo)
8. Based on Enterprise plan and published overage rate - discounts may be available

Advanced Feature Pricing (monthly)

Provider	Health Checks - Failover	Health Checks - Load Balancing	Location Based Routing (Geo IP)	Zone Based Routing (Anycast)	DNSSEC
AWS Route 53	NA	NA	$0.25/million queries10	NA	NA
UltraDNS	$2251	$5632	$5003	NA	Included
Cotendo	$1304	$1304	Included	Included	NA
Dyn	$1005	$2005	$4006	$2006	Included
DNS Made Easy	$0.427	NA	NA	$558	NA
easyDNS	Included9	NA	NA	Included9	Included

1. UltraDNS refers to this feature as Sitebacker. Price is based on 10 million query bundle pricing (25% discount) with 2 monitored target hosts. Each additional target host is $113/mo ($150 without bundle discount) up to 5, then $90 up to 10, then $68
2. UltraDNS refers to this feature as Traffic Controller. Price is based on 10 million query bundled pricing (25% discount) with 3 monitored target hosts. Each additional IP is $188/mo ($250 without bundle discount)
3. Pricing based on 10 million query bundled pricing (25% discount) with up to 5 target hosts. Each additional target host is $100/mo
4. Per hostname with up to 10 monitored target hosts
5. Priced based on # of monitor samples per month, $3 per 1000 samples - pricing provided is based on 3 target hosts, 1 monitoring node and 3 minute monitoring intervals. Pricing reduces to $0.30/1000 for 10 million samples/month
6. Per hostname
7. Must prepay annually - the Business and Corporate plans include 3 and 10 failover hostnames respectively and up to 5 monitored target hosts using 2-4 minute monitoring intervals
8. Requires $60/yr business plan. $1500/yr Corporate plan includes 1 geo-targeted hostname at no additional charge
9. Only 15 minute monitoring intervals are supported
10. Query surcharge for latency based routing - only target hosts in AWS data centers are supported. Query pricing reduces to $0.125/ million queries for volume above 1 billion queries/mo

Summary

There is a lot to consider when comparing managed DNS providers. We've included what we believe to be a few relevant, objective and comparable evaluation criteria in this post and our new DNS report. Organizations should consider the criteria that are most relevant to them when evaluating providers. If advanced features are needed, an enterprise provider may be the best choice. If just plain DNS is needed, a self service provider like AWS Route 53 may be a better and more cost effective service. The most important factor is to make your selection based on relevant and objective criteria, not marketing spin.

View Full Report - August 2012

Is Joyent Really 14X Faster than EC2 and Azure the "Fastest Cloud"? Questions to Ask About Benchmark Studies

Many are skeptical of claims that involve benchmarks. Over the years benchmarks have been manipulated and misrepresented. Benchmarks aren't inherently bad or created in bad faith. To the contrary, when understood and applied correctly, benchmarks can often provide useful insight for performance analysis and capacity planning. The problem with benchmarks is they are often misunderstood or misrepresented, frequently resulting in bold assertions and questionable claims. Oftentimes there are also extraneous factors involved such as agenda-driven marketing organizations. In fact, the term "benchmarketing" was coined to describe questionable marketing-driven, benchmark-based claims. This post will discuss a few questions one might consider when reading benchmark-based claims. We'll then apply these questions to 2 recent cloud related, benchmark-based studies.

Questions to consider

The following are 7 questions one might ask when considering benchmark-based claims. Answering these questions will help to provide a clearer understanding on the validity and applicability of the claims.

1. What is the claim? Typically the bold-face, attention grabbing headline like Service Y is 10X faster than Service Z
2. What is the claimed measurement? Usually implied by the headline. For example the claim Service Y is 10X faster than Service Z implies a measurement of system performance
3. What is the actual measurement? To answer this question, look at the methodology and benchmark(s) used. This may require some digging, but can usually be found somewhere in the article body. Once found, do some research to determine what was actually measured. For example, if Geekbench was used, you would discover the actual measurement is processor and memory performance, but not disk or network IO
4. Is it an apples-to-apples comparison? The validity of a benchmark-based claim ultimately depends on the fairness of the testing methodology. Claims involving comparisons should compare similar things. For example, Ford could compare a Mustang Shelby GT500 (top speed 190 MPH) to a Chevy Aveo (top speed 100 MPH) and claim their cars are nearly twice as fast, but the Aveo is not a comparable vehicle and therefore the claim would be invalid. A more fair, apples-to-apples comparison would be a Mustang GT500 and a Chevy Camaro ZL1 (top speed 186).
5. Is the playing field level? Another important question to ask is whether or not there are any extraneous factors that provided an unfair advantage to one test subject over another. For example, using the top speed analogy, Ford could compare a Mustang with 92 octane fuel and a downhill course to a Camaro with 85 octane fuel and an uphill course. Because there are extraneous factors (fuel and angle of the course) which provided an unfair advantage to the Mustang, the claim would be invalid. To be fair, the top speeds of both vehicles should be measured on the same course, with the same fuel, fuel quantity, driver and weather conditions.
6. Was the data reported accurately? Benchmarking often results in large datasets. Summarizing the data concisely and accurately can be challenging. Things to watch out for include lack of good statistical analysis (i.e. reporting average only), math errors, and sloppy calculations. For example, if large, highly variable data is collected, it is generally a best practice to report the median value in place of mean (average) to mitigate the effects of outliers. Standard deviation is also a useful metric to include to identify data consistency.
7. Does it matter to you? The final question to ask is, assuming the results are valid, does it actually mean anything to you? For example, purchasing a vehicle based on a top speed comparison is not advisable if fuel economy is what really matters to you.

Case Study #1: Joyent Cloud versus AWS EC2

In this case study, Joyent sponsored a third party benchmarking study to compare Joyent Cloud to AWS EC2. The study utilized our own (CloudHarmony) benchmarking methodology to compare 3 categories of performance: CPU, Disk IO and Memory. The end results of the study are published on the Joyent website available here. In the table below, we'll apply the questions listed above to this study. Answers will be color coded green where the study provided a positive response to the question, and red where the results are misleading or misrepresented.
Questions & Answers

What is the claim?	Joyent Cloud is 3x - 14x Faster than AWS EC2
The claims are broken down by measurement type (CPU, Disk IO, Memory), and OS type (SmartOS/Open Solaris, Linux). The resulting large, colorful icons on the Joyent website claim that Joyent Cloud is faster than EC2 by a margin of 3x - 14x
What is the claimed measurement?	CPU, Disk IO, Memory Performance
Our benchmarking methodology was used to measure these different categories of performance. This methodology consists of running multiple benchmarks per category and creating a composite measurement based on a summary of the results for all benchmarks in each category. The methodology is described in more detail on our blog here (CPU), here (Disk IO) and here (Memory).
What is the actual measurement?	CPU, Disk IO, Memory Performance
Is it an apples-to-apples comparison?	Dissimilar instance types were compared
In the Linux comparison, Joyent claims 5x faster CPU, 3x faster Disk IO, and 4x faster memory. Based on the report details, it appears those ratios originate from comparing a 1GB Joyent VM to an EC2 m1.small. This selection provided the largest performance differential and hence the biggest claim. While price-wise, these instance types are similar (disregarding m1.small spot and reserve pricing where it is 1/2 the cost), that is where the similarities stop. At the time of this report, m1.small was the slowest EC2 instance with a single core and older CPU, while Joyent's 1GB instance type has 2 burstable cores and a newer CPU. The m1.small is not intended for compute intensive tasks. For that type of workload EC2 offers other options with newer CPUs and more cores. To provide an apples-to-apples comparison on performance, the claim should be based on 2 instance types that are intended for such a purpose (e.g. an EC2 m2 or cc1).
Is the playing field level?	Operating system and storage type were different
The study compares Joyent Cloud VMs running SmartOS or Ubuntu 10.04 to AWS EC2 VMs running CentOS 5.4. Joyent's SmartOS is based on Open Solaris and highly optimized for the Joyent environment. Ubuntu 10.04 uses Linux Kernel 2.6.32 (release date: Dec 2009) which is over 3 years newer than the 2.6.18 kernel (release date: Sep 2006) in CentOS 5.4. Newer and more optimized operating systems will almost always perform better for similar tasks on identical hardware. This provided an advantage to the Joyent VMs from the offset. Additionally, the tests compared EC2 instances running on networked storage (EBS) to Joyent instances running on local storage, which also provided an advantage to the Joyent VMs for the disk IO benchmarks.
Was the data reported accurately?	Mistakes were made in calculations
This study was based on a cloud performance comparison methodology we (CloudHarmony) developed for a series of blog posts in 2010. For CPU performance, we developed an algorithm that combined the results of 19 different CPU benchmarks to provide a single performance metric that attempts to approximate the AW ECU (Elastic Compute Unit). To do so, we utilized EC2 instances and their associated ECU value as a baseline. We called this metric CCU and the algorithm for producing it was described in this blog post. Part of the algorithm involved calculating CCU when performance exceeded the largest baseline EC2 instance type, the 26 ECU m2.4xlarge. In our algorithm we used the performance differential ratio between an m1.small (1 ECU) and m2.4xlarge (26 ECUs). The third party, however, used the ratio between an m2.2xlarge (13 ECUs) and m2.4xlarge (26 ECUs). Because m2s run on the same hardware type, the performance difference between an m2.2xlarge and an m2.4xlarge is not very great, but the difference in ECUs is very high. The end results was their calculations producing a very high CCU value for the Joyent instances (in the range of 58-67 CCUs). Had the correct algorithm been used, the reported CCUs would have been much lower.
Does it matter to you?	Probably not
There isn't much value or validity to the data provided in these reports. The bold headlines which state Joyent Cloud is 3X - 14X faster than EC2 are based on very shaky grounds. In fact, with Joyent's approval, we recently ran our benchmarks in their environment resulting in the following CPU, disk IO and memory performance metrics: CloudHarmony Generated Joyent/EC2 Performance Comparison CPU Performance: AWS EC2 vs Joyent View Full Report Provider Instance Type Memory Cost CCU EC2 cc1.4xlarge 23 GB $1.30/hr 33.5 Joyent XXXL 48GB (8 CPU) 48 GB $1.68/hr 28.44 EC2 m2.4xlarge 68.4 GB $2.00/hr 26 EC2 m2.2xlarge 34.2 GB $1.00/hr 13 Joyent XL 16GB (3 CPU) 16 GB $0.64/hr 10.94 Joyent XXL 32GB (4 CPU) 32 GB $1.12/hr 6.82 EC2 m2.xlarge 17.1 GB $0.50/hr 6.5 Joyent Large 8GB (2 CPU) 8 GB $0.36/hr 6.19 Joyent Medium 4GB (1 CPU) 4 GB $0.24/hr 5.53 Joyent Medium 2GB (1 CPU) 2 GB $0.17/hr 5.45 Joyent Small 1GB (1 CPU) 1 GB $0.085/hr 4.66 EC2 m1.large 7.5 GB $0.34/hr 4 EC2 m1.small 1.7 GB $0.085/hr 1 Disk IO Performance: AWS EC2 vs Joyent View Full Report - Note: the EC2 instances labeled EBS utilized a single networked storage volume - better performance may be possible using local storage or multiple EBS volumes. All Joyent instances utilized local storage (networked storage is not available). Provider Instance Type Memory Cost IOP EC2 cc1.4xlarge (local storage - raid 0) 23 GB $1.30/hr 212.06 EC2 cc1.4xlarge (local storage) 23 GB $1.30/hr 194.29 Joyent XXXL 48GB (8 CPU) 48 GB $1.68/hr 187.38 Joyent XL 16GB (3 CPU) 16 GB $0.64/hr 144.71 Joyent XXL 32GB (4 CPU) 32 GB $1.12/hr 142.19 Joyent Large 8GB (2 CPU) 8 GB $0.36/hr 130.84 Joyent Medium 4GB (1 CPU) 4 GB $0.24/hr 110.78 Joyent Medium 2GB (1 CPU) 2 GB $0.17/hr 109.2 EC2 m2.2xlarge (EBS) 34.2 GB $1.00/hr 87.58 EC2 m2.xlarge (EBS) 17.1 GB $0.50/hr 83.62 EC2 m2.4xlarge (EBS) 68.4 GB $2.00/hr 82.79 EC2 m1.large (EBS) 7.5 GB $0.34/hr 56.82 Joyent Small 1GB (1 CPU) 1 GB $0.085/hr 56.08 EC2 m1.small (EBS) 1.7 GB $0.085/hr 27.08 Memory Performance: AWS EC2 vs Joyent View Full Report Provider Instance Type Memory Cost CCU EC2 cc1.4xlarge 23 GB $1.30/hr 137.2 EC2 m2.2xlarge 34.2 GB $1.00/hr 109.41 EC2 m2.4xlarge 68.4 GB $2.00/hr 109.14 EC2 m2.xlarge 17.1 GB $0.50/hr 103.35 Joyent XL 16GB (3 CPU) 16 GB $0.64/hr 100.87 Joyent XXXL 48GB (8 CPU) 48 GB $1.68/hr 92.5 Joyent XXL 32GB (4 CPU) 32 GB $1.12/hr 90.79 Joyent Large 8GB (2 CPU) 8 GB $0.36/hr 90.37 Joyent Medium 2GB (1 CPU) 2 GB $0.17/hr 84.2 Joyent Small 1GB (1 CPU) 1 GB $0.085/hr 78.51 Joyent Medium 4GB (1 CPU) 4 GB $0.24/hr 76.04 EC2 m1.large 7.5 GB $0.34/hr 61.8 EC2 m1.small 1.7 GB $0.085/hr 22.24

Case Study #2: Microsoft Azure Named Fastest Cloud Service

In October 2011, Compuware published a blog post related to cloud performance. This post was picked up by various media outlets resulting in the following headlines:

• Azure Tops Cloud Provider Performance Index (ReadWrite Cloud)
• Windows Azure beats Amazon EC2, Google App Engine in cloud speed test (ars technica)
• Microsoft Azure Named Fastest Cloud Service (Information Week)
• Windows Azure cloud fastest according to tests (Ewandoo)

Here's how the test worked in a nutshell:

• Two sample e-commerce web pages were created. The first with items description and 40 thumbnails (product list page), and the second with a single 1.75 MB image (product details page)
• These pages were made accessible using a Java application server (Tomcat 6) running in each cloud environment. The exception to this is Microsoft Azure and Google AppEngine (platform-as-a-service/PaaS environments) which required the pages to be bundled and deployed using their specific technology stack
• 30 monitoring servers/nodes were instructed to request these 2 pages in succession every 15 minutes and record the amount of time it took to render both in their entirety (including the embedded images)
• The 30 monitoring nodes are located in data centers in North America (19), Europe (5), Asia (3), Australia (1) and South America (2) - they are part of the Gomez Performance Network (GPN) monitoring service
• After 1 year an average response time was calculated for each service (response times above 10 seconds were discarded)

Now lets dig a little deeper...
Questions & Answers

What is the claim?	Microsoft Azure is the "fastest cloud"
What is the claimed measurement?	Overall performance (it's fastest)
What is the actual measurement?	Network Latency & Throughput
Rendering 2 html pages and some images is not CPU intensive and as such is not a measure of system performance. The main bottleneck is network latency and throughput, particularly to distant monitoring nodes (e.g. Australia to US)
Is it an apples-to-apples comparison?	Types of services tested are different (IaaS vs PaaS) and the instance types are dissimilar
Microsoft Azure and Google AppEngine are platform-as-a-service (PaaS) environments, very different from infrastructure-as-a-service (IaaS) environments like EC2 and GoGrid. With PaaS, users must package and deploy applications using custom tools and more limited capabilities. Applications are deployed to large clustered, multi-tenant environments. Because of the greater structure and more limited capabilities of PaaS, providers are able to better optimize and scale those applications, often resulting in better performance and availability when compared to a single server IaaS deployment. Not much information is disclosed regarding the sizes of instances used for the IaaS services. With some IaaS providers, network performance can vary depending on instance size. For example, with Rackspace Cloud, a 256MB cloud server is capped with a 10 Mbps uplink. With EC2, bandwidth is shared across all instances deployed to a physical host. Smaller instance sizes generally have less, and more variable bandwidth. This test was conducted using the nearly smallest EC2 instance size, an m1.small.
Is the playing field level?	Services may have unfair advantage due network proximity and uplink performance
Because network latency is the main bottleneck for this test, and only a handful of monitoring nodes were used, the results are highly dependent on network proximity and latency between the services tested and the monitoring nodes. For example, the Chicago monitoring node might be sitting in the same building as the Azure US Central servers giving Azure and unfair advantage in the test. Additionally, the IaaS services where uplinks are capped on smaller instance types would be at a disadvantage to uncapped PaaS and IaaS environments.
Was the data reported accurately?	Simple average was reported - no median, standard deviation or regional breakouts were provided
The CloudSleuth post provided a single metric only… the average response time for each service across all monitoring nodes. A better way to report this data would involve breaking the data down by region. For example, average response time for eastern US monitoring nodes. Reporting median, standard deviation and 90th percentile statistical calculations would also be very helpful in evaluating the data.
Does it matter to you?	Probably not
Unless your users are sitting in the same 30 data centers as the GPN monitoring nodes, this study probably means very little. It does not represent a real world scenario where static content like images would be deployed to a distributed content delivery network like CloudFront or Edgecast. It attempts to compare two different types of cloud services, PaaS and IaaS. It may use IaaS instance types like the EC2 m1.small that represent the worst case performance scenario. The 30 node test population is also very small and not indicative of a real end user population (end users don't sit in data centers). Finally, reporting only a single average value ignores most statistical best practices.