Moving files and migrating workloads into Azure

Copying files: the window, and the command line

Three tools do the same job at different scales. The portal is fine for one file. Azure Storage Explorer is the desktop application for browsing accounts and dragging things about, which is the right answer when somebody needs to look at what is there. AzCopy is the command-line utility, and it is what you reach for the moment the task has to run twice: it scripts, it parallelises large transfers, it resumes after an interruption, and it can copy between two storage accounts without pulling the data down to your machine and pushing it back up.

When the network is the constraint

Before choosing a tool, do the arithmetic: total volume divided by the bandwidth you can actually use. When the answer is measured in weeks or months, no copy utility fixes it, because parallelism helps you saturate a link and cannot make the link wider. That is the point at which the correct answer stops being software and becomes a lorry.

Azure Data Box is the offline option: Microsoft ships a ruggedised, encrypted appliance, you copy the data onto it on site, and you ship it back. The family scales with the job, from a disk-sized device up to appliances holding hundreds of terabytes, and it works in the other direction too when bulk data has to come out of Azure. The requirement that gives it away is a stated volume against a stated link speed — and the distractor is always a faster or more parallel copy tool.

Keeping an on-premises server in step

Azure File Sync turns a Windows Server into a cache of an Azure file share. The authoritative copy lives in Azure; the branch server keeps serving users at local speed; and cloud tiering decides what actually occupies local disk, keeping frequently used files there while everything else lives only in Azure until somebody asks for it. It also does multi-site sync, so several servers can present the same share and converge on the same content.

A nightly copy job is the wrong shape for that requirement and is offered as an answer anyway. It is a backup rather than a synchronisation: changes made in Azure never come back, conflicting edits are nobody's problem until they are, and nothing ever frees space on the server that was filling up. Mounting the Azure share directly and switching the local server off is the other tempting wrong answer — one authoritative copy, certainly, but every read now crosses the wide area link the office depends on not using.

Migrating servers rather than files

Azure Migrate is the hub for moving an estate rather than a dataset. It discovers what is running, analyses which servers depend on which, produces assessments that size the Azure equivalents and project a monthly cost, and then carries out the migration of servers, databases and web applications. When a scenario asks for an inventory, a dependency map, a sizing and a cost before anybody commits, that arc of work is what it is describing.

Keep the four hybrid tools apart by the verb in the requirement. Migrate assesses and moves. Arc governs what stays where it is. Data Box ships bulk data. Azure Advisor advises on what already runs in Azure — which is why it cannot help with four hundred servers in somebody's datacentre, because it cannot see them.