How to get your website up and running?

V

Vladimir Sokolovsky2010-11-11 16:23:28

Web development

Vladimir Sokolovsky, 2010-11-11 16:23:28

There is a website that is very important for the client. The main task after development is uptime, that is, 100% uptime.
What are some ways to keep your site up and running? The maximum time that the site can be unavailable is 10 minutes.

Do not offer “good hosting”, since not a single hoster will give a 100% guarantee of uninterrupted operation, and if it does, it will be far from the truth.

Reply

Answer the question

In order to leave comments, you need to log in

7 answer(s)

P

pentarh, 2010-11-11
@pentarh

Cloud hosting from Amazon. You create instances in different parts of the world and go ahead.

C

casey, 2010-11-11
@casey

Write down the reasons for the falls.

S

Sergey, 2010-11-11
@bondbig

GSLB cluster.

B

bdmalex, 2010-11-11
@bdmalex

And how can you answer the question “... in case of problems with the current hosting”, if the name of the current hosting is NOT
NAMED ?

S

Sufiks Sufiksovich, 2010-11-11
@Solovej

If we are talking about very big money, then I would recommend using the services of www.akamai.com expensive, but effective.
If you are looking for something much cheaper, then you need to rent servers in different data centers and “connect” them.
What countries is the site aimed at?
If Europe, then these are data centers in Germany, Holland, Sweden, Norway (Teliasonera and The bunker).
If we are talking about America, then it is most likely Amazon and Verizon - of course, they are also suitable for Europe.

A

ArtemS, 2010-11-11
@ArtemS

Reliability can be increased indefinitely, and even then it will not protect against all cases. Both Yandex and Facebook crashed.
To give advice, you need to understand the architecture of the site. There are also many reasons for falls, and each one can be defended. Everything depends on money. You can suffer both from force majeure (the DC burned down), and from various attacks.
Means of uptime'a
RAID
DB replication, DB sharding
Balancer (either via nginx or DNS IP) load on N frontends.
Splitting front-ends and back-ends across
BGP data centers
There may not be enough money for everything :)

V

Vlad Zhivotnev, 2010-11-14
@inkvizitor68sl

scatter vdski, buy instances in different parts of the world ... who will think about synchronization? And with DNS, who will shaman if something falls? do not forget that they only work as round-robin.
2 servers in m9 and m9-vdnh, heartbeat, a smart gateway that will respond and transfer the IP address to the second server if the first one falls. The cost is 9-15k + rent / colo servers + traffic.