1. How can you tell if the server
is fast?
You must test it by requesting files from the server and measuring how much time elapses while the file travels to your computer. The rate (kilobytes divided by the time in seconds) is a factor of the server speed and the quality of the network between the two terminals. In effect, the rate is an indication of the slowest part between the server and your computer. A bottleneck could take place at the server, the server's network, the downstream network through the ISP or even the modem on your own computer.
A dialup modem will likely be much slower than any server. Therefore when testing a server, use the fastest connection to which you have access--at minimum, cable or fast DSL is necessary.
Be sure to run the test on different days and
at different times of the day.
If two servers consistently send files at different rates, you can deduce which one is more powerful (or on the faster network). If they both serve files at the same rate, the speed could be limited by the downstream connection; you will need to use a faster connection to adequately compare the servers.
At Interstellar Data, we pride ourselves on managing some of the most powerful servers and the fattest pipes to major networks in the business. At any given time, our servers are processing files over 500 KB/sec keeping up with even the fastest connections.
2. What does ping time have to do
with it?
Ping time refers to the amount of
time (in milliseconds) elapsed between sending a query and receiving a server's response. This is a rough indication of how
far away the server is. Under 20 milliseconds - same
city; under 50 ms - within 600 miles; 100 ms - across
the U.S.; 500 ms - across oceans; 2000 ms - U.S. to
China and Australia. Ping time is a byproduct of a server's power but more often reflects bottlenecks encountered by data travelling over the internet.
For instance, a server sends a small file in half
a second (500 ms) and is located 20 ms away. A faster
server sends the same file in 150 ms and resides on
the opposite U.S. coast 110 ms away. The latter appears twice as fast to the user.
Therefore, ping time is less important than
server processing time and network quality. Of course, if the server is
located on the other side of the globe, even a fast
server cannot mitigate the delay.
You can use Symantec's traceroute utility to test the time it takes for a server to reach the Silicon Valley. Type in Interstellardata.com and see how fast our data moves up the West Coast.
3. Why can't I ping some servers?
Some servers don't respond to ping requests
for security reasons. You can, however, use the "tracert"
function which pings each node along the path to the server. The final hops may be unreachable if the server ignores the request. The results you get, however,
are probablyly in the same building.
4. Does running fast servers compromise
stability?
On the contrary, systems with adequate
processor and hard drive resources tend to operate with less
problems.
5. How do your servers connect to
the internet backbones?.
Our data center's internal switched network utilizes the Border Gateway Protocol Version 4 (BGP4) over Cisco 12000 series GSR routers. Our peering arrangements include the following exchanges:
MAE-West in San Jose (100Mbps FDDI,155Mbps OC3 ATM)
MAE-East (100Mbps FDDI,155Mbps OC3 ATM)
AADS in Chicago
PAIX(Palo Alto Internet Exchange) in Palo Alto (1000Mbps Gigabit Ethernet)
NYIIX (New York Internet Exchange) in New York (100Mbps Fast Ethernet)
MAE-East in Vienna
Level3 Communications
Sprint
MCI WorldCom
Frontier
Pac Bell (155Mbps OC3 ATM)
AADS (Ameritech) (155Mbps OC3 ATM)
|
