Skip to chapter 00

COOLORANGE SOLUTIONS LTDCudworth, Barnsley

helpdesk@coolorangesolutions.digital A call is arranged by e-mail first, because the office is more often on a client site than at a desk.

00The notebook

The work is written down, and then the writing is left behind

COOLORANGE SOLUTIONS LTD is an independent information technology consultancy working with small businesses across South Yorkshire from an office in Cudworth, Barnsley. Four kinds of engagement account for nearly everything the company is asked to do, and each one is named here by the document it leaves in the client's own hands.

The page below is laid out as eight chapters, because that is how an engagement actually reads: a walk round, a drawing, a plan, a window, a shelf of copies, a vocabulary, and a pack of paper at the end. The index beside these chapters moves with you, so the document can be entered anywhere.

runbook.txt — page two of the pack

  1. POWER: switch, server, desks.
  2. BACKUP: 22:00 nightly.
  3. DRIVE LIGHT: green by 23:30.
  4. RESTORE: ask by e-mail.
  5. REGISTER: in the safe, two keys.
  6. NEW STARTER: account first.
  7. MAP: page one of the pack.

A runbook is the short sheet of operating instructions a business keeps beside the equipment, and the lines above show the shape of one rather than the contents of anybody's. A real page carries the client's own times, the client's own kit and the name of the person who holds the second key.

Chapter 00 of 07 · next, the first hour on site

01Systems review

The first hour on site

A review is not an inventory, and the difference matters: an inventory lists equipment, whereas a review explains what the business would lose if a given piece of it stopped. The order below is worked through in sequence, because each step narrows the next one, and the first hour settles most of it.

  • First What the business does on an ordinary Tuesday: which files are opened, which system takes the orders, and which task nobody is able to do while the network is down.
  • Second Where the data actually lives, meaning the physical or hosted location of every set of records the business depends on, including the sets that live on one laptop.
  • Third Identity and access: who holds an account, which accounts still belong to people who have left, and whether administrative rights are shared between several people.
  • Fourth The network path from a desk to that data, including the router, the switch, the wireless access points and the internet circuit, with the speed of that circuit measured and not assumed.
  • Fifth Backup and restore, which are two separate things, because a backup is a copy that was made and a restore is a copy that was proved to open.
  • Sixth Licensing and support contracts, so that renewal dates and the party responsible for each system are written in one place instead of four inboxes.

Nearly every unreviewed network contains one switch that nobody present is able to account for, and it is usually the one carrying the accounts machine.

Chapter 01 of 07 · the longer account sits on the systems reviews page

02The drawing

As found, as handed over

The two drawings below occupy one frame. Drag the handle, or use the slider beneath it, to wipe between a network as it is commonly found on a first visit and the same network once it has been documented. Both show the same six devices; only the knowledge of them differs.

??? ??? six devices, no map Circuitmeasured, not assumed Switch .18 ports, 3 in use Server .10records and backup source Desks .40 to .60named per user Access point .20office and yard six devices, every one named

As found As handed over

The right-hand drawing takes about an afternoon to produce and saves the same afternoon every time somebody new is asked where the accounts machine plugs in.

Chapter 02 of 07 · next, where the rollback is written down

03Migration planning

Where the rollback is written down

A migration is the move of a system from where it runs now to where it will run next, and the planning decides whether the move is dull. The plan names every move in order, the person who performs it, the check that confirms it worked, and the point at which the old arrangement is taken out of service.

The rehearsal comes before the cutover. A rehearsal copies the data into the new system while the old one keeps running, which means the copy time is measured on the real circuit and the awkward files are found a fortnight early rather than at midnight. The cutover is the short window in which the business stops using one system and starts using the other, and it is agreed in hours, because that number governs every other number in the plan.

The rollback position is the state the old system must be left in so that abandoning the move takes minutes, and it is written down before the cutover begins, together with the condition that would trigger it. In practice a written rollback is almost never used, which is the argument for writing it, since the version held only in somebody's head is the version that fails at two in the morning.

Engagements are shaped around that structure and not around a fixed package. Some clients want the review and the plan, and will move the system themselves; others want the move performed and the notes handed over afterwards.

Nothing in the plan sits on a week grid, because a small business moves in the evenings and on the two Saturdays when the yard is quiet.

Chapter 03 of 07 · the sequence in full sits on the migration planning page

04First panel

Working out the rate a cutover window demands

The window is agreed with the client first, and the rate has to follow from it. Enter the volume to be moved and the hours actually available, and the arithmetic below solves for the sustained transfer rate that window requires. Tick the read-back box when the data is to be read back and compared afterwards, because the same volume then crosses the link twice.

Megabytes to move200 × 1024 = 204,800 MB

Seconds in the window6 × 3600 = 21,600 s

Volume crossing the link204,800 MB, one pass

Sustained rate required204,800 ÷ 21,600 = 9.48 MB/s

The same rate in circuit units9.48 × 8 = 75.85 Mbit/s

Largest volume that fits the same window at three link rates a client is able to check on their own circuit

100 Mbit/s, that is 12.5 MB/s12.5 × 21,600 ÷ 1024 = 263.7 GB

300 Mbit/s, that is 37.5 MB/s37.5 × 21,600 ÷ 1024 = 791.0 GB

1 Gbit/s, that is 125 MB/s125 × 21,600 ÷ 1024 = 2,636.7 GB

Every figure on this panel is a unit conversion of the two numbers entered above: one gigabyte is treated as 1024 megabytes, one hour as 3600 seconds and one byte as 8 bits. The result states what the agreed window demands of a link, so it is arithmetic and a guide only rather than a measurement, and what a circuit actually sustains is established during the rehearsal.

Chapter 04 of 07 · next, sizing the shelf the copies sit on

05Second panel

Sizing the shelf the copies sit on

Retention is the number of copies kept before the oldest one is overwritten, and it is set separately for each tier: a daily, a weekly and a monthly. Enter the size of one full copy and the three counts, and the panel totals the shelf those counts ask for. The bar draws each tier at its share of the whole, which is usually where a client discovers that the dailies, not the monthlies, are the bulk of the bill.

Daily tier14 × 400 = 5,600 GB

Weekly tier8 × 400 = 3,200 GB

Monthly tier12 × 400 = 4,800 GB

Copies held in total14 + 8 + 12 = 34 copies held

Shelf the counts ask for34 × 400 = 13,600 GB

The same figure in terabytes13,600 ÷ 1024 = 13.28 TB

What one further daily adds400 GB, which is 2.94 per cent of the shelf above

The shelf drawn at each tier’s share of it

  • Daily tier
  • Weekly tier
  • Monthly tier

Each tier is priced here as a full copy, so a weekly costs the same as a daily and no compression, deduplication or incremental saving is assumed anywhere on the panel. That makes the total the upper figure rather than the likely one, and it is a guide only: the shelf a client buys is settled once a real backup has run for a fortnight and its own sizes are on record.

Chapter 05 of 07 · topology and restore testing sit on the network and backup page

06Vocabulary

Two ways of saying the same thing

Consultancy vocabulary is mostly shorthand for an ordinary sentence, and the shorthand is only useful once the ordinary sentence has been said out loud. The table gives both, and names the document in which each term is actually settled, so a client reading a proposal knows which page to turn to.

The shorthand, the plain sentence, and the document that decides it
The phrase used in the trade What it means on the morning it matters Where it is settled
Recovery point objective The amount of work the business agrees to redo after a failure, counted from the last copy that is known to open. Backup schedule
Recovery time objective How long the business is willing to stand still while somebody rebuilds the system and puts the data back. Network and backup design
Cutover window The agreed hours in which one system stops taking work and the next system starts taking it. Migration plan
Rollback position The state the old system is left in so that abandoning the move is a five-minute decision rather than a night. Migration plan
Restore test A copy taken from the backup, opened on other equipment, and checked against something the client recognises. Restore-test notes
Credential register The list of accounts that matter and the record of where each password is kept and who is able to reach it. Handover pack
Single point of failure The one device whose loss stops everybody, which in a small office is usually a switch on a shelf. Systems review
Addressing scheme The rule deciding which number each device is given, so that a stranger can read the network from the numbers alone. Network design

The same figure, counted in the only hours that cost anything

Recovery point is normally quoted in elapsed hours, and elapsed hours flatter nobody. A nightly copy is often described as putting twenty-four hours of work at risk, which is true of the clock and false of the ledger, because most of those hours are hours in which the office is shut and nothing was typed into anything.

A copy finishes at 22:00 on Tuesday. The server fails at 16:30 on Wednesday, which is 18 hours 30 minutes later. The office works 09:00 to 17:00, so 11 hours of that gap are hours with nobody at a desk: 18.5 − 11 = 7 hours 30 minutes of work to redo.

Seven and a half hours is a bad Wednesday and a survivable one, and it is a different conversation from eighteen and a half. It is also the arithmetic that decides whether a second copy during the working day is worth paying for, which is the question most clients are really asking.

Chapter 06 of 07 · next, what is left with you

07Close of engagement

What is left with you

An engagement closes when the following four artefacts exist, are readable by somebody who was not present while the work was done, and are held by the client rather than by the consultancy.

  • The network map. The right-hand drawing from chapter 02, produced for the client's own site: every device named, every address recorded and every link drawn as it physically runs.
  • The location of the credential register. The register itself stays under the client's control; the notes state where it is, who holds it and which accounts it needs to contain.
  • The backup schedule. What is copied, how often, where each copy is kept, and how many copies are retained before the oldest is overwritten.
  • The restore-test notes. The date of the test, the file that was opened, how long the restore took, and what the outcome was.

Those four documents are the point of the work. A client who holds them is able to instruct any competent engineer, including one who is not this company, which is a reasonable definition of a handover.

A first e-mail to helpdesk@coolorangesolutions.digital is most useful when it says what the business does, roughly how many people sit at desks, and which single thing is currently causing the trouble. That is enough to say whether a review is the right first step or whether the answer is shorter than one.

Chapter 07 of 07 · the full pack is described on the documented handover page