Hosting Provider Performance Comparison Available

Buying a dedicated server or moving a site to a new network provider can be a stab in the dark in that it is often not easy to see the quality and reliability of the provider’s network performance until after the purchase has been made. Netcraft’s view is that lack of transparency on network performance and outages harms the whole industry, both consumers and providers. Just as the customer suffers from not being able to make an informed choice between suppliers taking into consideration network response times as well as price, vendors with fast and reliable networks have no easy way of empirically showing the prospect the relative quality of service of their network relative to other players in the market. Ignorance plays into the hands of the companies investing less in their networks, since they will be better able to discount, and their longer response times and network outages will be less obvious to the customer. More widespread knowledge helps the industry as a whole, because better informed customers are more willing to pay more for superior connectivity, and the extra revenue coming into the industry can be invested in further improving resilience, performance, and support creating a virtuous circle. Key metrics include;
  • fewer outages – no one wants to be on a network that suffers frequent loss of connectivity.
  • Shorter outages - customers will be more tolerant of short outages which may be operationally difficult to avoid;
  • faster response times - the shorter the response times, the better.
Netcraft is measuring and making available the response times of fifty leading hosting providers' sites to give an indication of the relative and absolute response times currently available in the industry. The performance measurements are made at fifteen minute intervals from four separate points around the internet, and averages are calculated over the immediately preceding 24 hour period.
Ranking by failed requests and connection time, 14:00 GMT, June 24th performance-graph-4.png
Using the performance of a hosting providers own site to determine the performance of the hosting companies network is only indicative. By default the sites are ranked in order of fewest failed requests, and shortest time to connect, in order to give the clearest indication of network capacity and congestion, with the least impact from the performance of the companies’ own web servers, though it is possible to sort by any column by clicking on the column heading. If you are using the table as a guide when choosing where to locate a dedicated or collocated server, remember that connection times fluctuate continually, and only hundredths of a second separate the top companies. Avoiding companies showing prolonged outages is likely to be a better strategy than necessarily going for the company with the fastest connection time. Factors other than network performance, including quality of support and price will also be important. If you are considering shared hosting then the load on the shared hosting system will likely be a greater constraint on the performance of your site than network connection time. If you represent a hosting company and would like to be included in the table, or if you are researching prospective hosting locations and would like more detailed performance information please mail us.

Speedera gains sites, reaches operating profitability

Speedera, which recently announced that it has reached operating profitability, has seen some impressive site gains in the last few weeks. Sites starting to use the Speedera caching service include NASA, Hewlett Packed, AMD and the India Times.

With Cable and Wireless announcing that its Global division, which includes the Digital Island business, is to be sold or closed, Speedera will be hopeful of making further gains within the next few weeks.

Where are the Windows Server 2003 sites hosted?

Some people expressed surprise that there could be thousands of sites running on Windows Server 2003 before it went on sale. The main reason for this is that some shared hosting providers who were part of the Windows Server 2003 beta program were prepared to run part of their sites on Windows Server 2003 prior to launch. The three providers with the largest number of active sites on Windows Server 2003 are all Microsoft Certified Partners while the fourth largest is Microsoft itself. No one else had more than 1000 active sites running on Windows Server 2003 at the start of May.

Top Microsoft Windows Server 2003
hosting companies by Active sites, May 2003
CompanyActive Sites
  Interland4,996
  Schlund3,125
  Fasthosts2,250
  Microsoft811

For comparison, Schlund and Interland [primarily due to its acquisition of Dialtone Internet] are also two of the largest Linux hosting providers, along with specialist Linux hosting providers Rackshack and Alabanza.

Top Linux hosting companies by Active sites, May 2003
CompanyActive Sites
  Schlund1,450,868
  Rackshack234,740
  Alabanza150,180
  Interland144,488
  About118,287
  Valueweb109,998
  Verio102,696

Who are the Fastest Growing Hosting Providers?

Over the last two years the most common reason for a Hosting Provider making the news was when the company went bust. Hoswever, outside of the mainstream newsflow, some companies have been growing strongly.

A table of the Hosting Providers who grew the fastest during 2002 is provided as an excerpt from our Hosting Provider Server Count. Companies are included in the filter if they started 2002 with more than 500 servers, finished with at least 1000, and grew at a rate of better than 33% year on year. This removes hosters which can show a significant percentage increase simply by virtue of starting small.

Fastest Growing Hosting Providers by % Increase in Web Visible Servers
December 2001 to December 2002
Parent CompanyDec 01Dec 02% ChangeMain Business Area
1&1 Internet AG 8127,203787%Mixed Hosting
Rackshack 1,2087,356 509%Dedicated
Colt 9992,049105%Telecoms & Colocation
Host Europe 1,6063,248102%Mixed Hosting
Tele Danmark Communications 7561,46994%Telecoms
Global Media Online Inc6421,19085%Mixed Hosting
Korea Telecom2,1783,67569%Telecoms
Rackspace2,8074,51861%Dedicated
Datapipe7461,13152%Dedicated
Deutsche Telekom1,8692,82451%Telecoms
Speakeasy1,1521,70648%Broadband
Advance Technology Works1,1611,68145%ISP
Interland1,9902,85143%Mixed Hosting
Sprint Communications2,2773,23842%Telecoms
Dialtone Internet2,2433,18242%Dedicated
Qwest Communications2,5203,54641%Telecoms
Telecom Italia1,9802,66234%Telecoms

Of the top 4 companies in the table 3 have recently issued their financial results for 2002 which confirm their continued growth. (1&1 Internet , Colt , HostEurope ).

The top of the table shows that fast growth in the hosting industry is closely linked to providing good value at a low price. 1&1 and Host Europe both offer extremely cheap shared hosting packages, while Rackshack has more or less defined the market for low-cost dedicated servers. Colt is often amongst the cheapest quotes for bandwidth in the cities in which it operates.

Several of the leading dedicated server companies have produced annual growth over 40%. Additionally, many of the larger telecoms companies are showing good rates of growth of web servers on their networks, in part from Hosting Resellers and DSL and Cable connections as well as their own hosting operations.

Limitations of the Hosting Provider Server Count include the following;

  • Only sites found by the Web Server Survey will be included. The number of hosts found running internet web sites by the Web Server Survey is large [over 40 million in April 2003], but not exhaustive.
  • Sites are attributed to companies by performing a reverse DNS lookup on each responding ip address in the Web Server Survey. If reverse DNS lookups have not been configured or otherwise fail, the count for the company will correspondingly reduced. To mitigate this we provide an additional view of the data compiled by Netblock registration. In practice the most successful hosting companies seem to set up reverse DNS correctly.
  • Backend machines such as database servers not running web sites will not be counted, as they are unseen from the Internet.
  • At most one server will be counted for each site. Round robin DNS, reverse web proxies, load balancing products like Cisco Local Director and BIG-IP and some connection level firewalls hide multiple web servers behind a single hostname.

Full details of the Hosting Provider Server Count are available.