Услуга · System administration

Virtualisation

A virtual environment gives you what bare servers never will: rollback to the state before an update, moving a machine to other hardware in minutes, and sensible use of capacity instead of five boxes idling.

Snapshot
rollback in a minute
Transfer
between nodes on the fly
Density
less hardware
Redundancy
a second node

What the work includes

We build from scratch and maintain environments already running, including moving physical servers inside.

Discuss the scope

Sizing

We size processors, memory and disks for the planned set of machines with headroom for peaks.

Deployment

Installing the environment, configuring the network and the storage subsystem.

Migration

Moving physical servers into virtual ones with a short outage window.

Limits

Limits and priorities, so that one greedy machine does not leave its neighbours without resources.

Snapshots and archives

The rule: a snapshot before any change, while data safety is provided by separate backups.

Redundancy

A second node that automatically brings machines up when the first fails.

How it goes

A small environment comes up in one or two weeks; we move the servers one by one in agreed windows.

01

Estimate

Which servers exist, what the load is, what to move and what is better left as it is.

02

Plan

The node specification, the network and storage diagram, the migration schedule.

03

Migration

First we raise the environment, then move the machines one at a time, checking each.

04

Routine

Snapshots, copies, resource monitoring, a written plan for a node failure.

Virtualisation puts all the eggs in one basket, and something has to be done about that. While the servers stood apart, one failure took out one service. Now a node failure stops everything at once. A second node, or at the very least fast recovery onto spare hardware, ceases to be a luxury.

Questions and answers

The overhead of modern environments is small and unnoticeable in ordinary work. Slowness appears when more machines are put on a node than it can serve. That is a sizing mistake, not a property of the technology.

You can and should; that is how most run it. What matters is allocating enough resources and not putting it on the same node as machines that hammer the disks. The drive requirement is unchanged: solid state only.

A snapshot lives on the same array and disappears with it, and if forgotten for a month it starts slowing the machine down as well. It is a rollback tool for after an update, not a means of preserving data.

We will build a virtual environment

Write how many servers you have and what they do. We will size the node and lay out a migration with no outage.