In May 2013, we have made a major deci­sion for the future of always­da­ta. After tak­ing near­ly nine months to research, com­pare, cal­cu­late, pur­chase, install, devel­op, debug, and doc­u­ment, it’s now time to announce it.

Our own infrastructure

Since its incep­tion in 2007, always­da­ta has relied on third-par­ty providers (main­ly OVH, Online, Hetzner and Amazon) for its infra­struc­ture (hard­ware, net­work). This has allowed us to con­cen­trate on the heart of our archi­tec­ture: the soft­ware. always­da­ta would not exist with­out these providers; we owe them a lot.

Not hav­ing our own infra­struc­ture has many advan­tages, but it also has draw­backs:

  • Part of our reli­a­bil­i­ty depends on our providers
  • We are lim­it­ed to equip­ment and fea­tures offered by our providers
  • We are sub­ject to changes in their ser­vice pol­i­cy

As our infra­struc­ture grows, these draw­backs become more impor­tant. After a long dis­cus­sion, we have decid­ed to take the plunge and become inde­pen­dent.

Details

We have two objec­tives:

  • Build a reli­able and durable archi­tec­ture
  • Be com­plete­ly inde­pen­dent

These objec­tives result in many choic­es that must be made. Here are the key ones:

  • We are host­ed in a neu­tral data cen­ter (oper­at­ed by Equinix), with­out inter­me­di­ary
  • We joined the RIPE (we are “LIR”) to ensure our inde­pen­dence
  • All our equip­ment is dou­bled in pow­er
  • Our inter­nal net­work is ful­ly redun­dant: each serv­er has two net­work con­nec­tions (of dif­fer­ent brands), which are them­selves con­nect­ed to two dif­fer­ent routers (of dif­fer­ent brands)
  • We have three dif­fer­ent net­work oper­a­tors (tran­sit)

Some addi­tion­al tech­ni­cal details:

In practice, what changes?

For our cus­tomers, noth­ing changes direct­ly: our ser­vice and our price remain the same.

Our cur­rent servers, locat­ed with our providers, will be moved to our infra­struc­ture pro­gres­sive­ly over time — the total migra­tion is like­ly to take up to two years. We will announce each migra­tion before­hand. Our ded­i­cat­ed cus­tomers will be con­tact­ed per­son­al­ly and we will orga­nize the whole serv­er migra­tion.

In a few weeks, new accounts will be host­ed to our own infra­struc­ture. For new ded­i­cat­ed servers, this is already the case.

The real changes will be vis­i­ble indi­rect­ly; some exam­ples:

  • We will be able to offer new ser­vices. Did you know that Equinix oper­ates in 14 coun­tries?
  • The equip­ment we use is more suit­able for our pur­pos­es. For exam­ple, the dri­ves and pow­er sup­plies of our servers can be hot swapped
  • The range of equip­ment that we can use is broad­er — for exam­ple, SSD or dri­ves with very high capac­i­ty
  • Our net­work, ful­ly redun­dant and sim­pli­fied, should be more reli­able

And after?

We spend much of 2013 in the con­struc­tion of our infra­struc­ture, which has some­what slowed the pace of new devel­op­ments. We are plan­ning sev­er­al things for 2014 — our team has also wel­comed a new devel­op­er, Martin, a few weeks ago. Our goal is to pro­vide more reli­a­bil­i­ty, more fea­tures, and to remain avail­able to quick­ly answer all ques­tions and con­cerns from cus­tomers.