Glossary
HLD vs LLD
A high-level design (HLD) describes the overall architecture of a solution: its building blocks, zones and principles. A low-level design (LLD) takes the same solution down to the detail needed to build it: host names, addressing, VLANs, settings and firewall rules.
Updated
What is the difference between HLD and LLD?
HLD stands for high-level design and LLD for low-level design. They are two levels of the same design effort, and in infrastructure and network projects they are usually two separate documents delivered at two different milestones.
The high-level design sets out the overall architecture: building blocks, zones and principles, with no IP address and no server name. It can be read in half an hour and presented to a review board. The low-level design goes down to the level where someone can install: host names, addressing, VLAN numbers, cluster settings, firewall rules.
Both cover the same subject. The difference between a high-level design and a low-level design lies in the reader and the decision at stake. The HLD answers "is this the right architecture?" The LLD answers "what exactly do we configure?" No standard dictates the outline of either one. Every integrator and vendor has a template, and the line between the two moves from one organization to the next.
What does an HLD document contain?
An HLD document opens with the need and the requirements: expected availability, volumes, security constraints, sites in scope. It then presents the chosen architecture as blocks and conceptual diagrams, and above all the choices: why two active data centers and not active-passive, why this cloud provider, why four network zones.
Take the replacement of a network core across two sites. The HLD states that each site gets a redundant pair of core switches, that the sites are joined by two links on separate paths, that routing between zones goes through the firewalls, and which applications are expected to fail over from one site to the other. It gives no port number and no VLAN ID.
A good HLD also records what was rejected and why. That part ages the least, and it is the part people come back to three years later when someone proposes to undo a decision.
What does an LLD document contain?
The LLD takes each block of the HLD and details it until an engineer can build it without asking a question. It holds more tables than prose.
For the same network core, the LLD gives the name of each device, its model and software version, the addressing plan, the list of VLANs with their IDs and subnets, the port-by-port cabling, the routing protocol settings and the link aggregation groups. On the security side, it carries the firewall rules to open, taken from the flow matrix.
For a virtualization platform: host and cluster names, management addresses, virtual switches and port groups, datastores, affinity rules, service accounts, backup settings. In the cloud, the LLD lists accounts or subscriptions, virtual networks and their address ranges, security groups and instance sizes.
An LLD document template is mostly a set of empty tables in that order: naming, addressing, VLANs, cabling, routing, security rules, platform settings. A good LLD passes one test: two engineers following it separately end up with the same configuration. Many need a second pass after installation to record the as-built state, because the work forced deviations. If that pass is skipped, the LLD describes an intention and not what runs.
Who writes the HLD and the LLD, and when is the split useful?
The HLD is written by the solution architect, on the customer side or at the integrator. Its readers are the people who decide: the architecture review board, the security officer, the head of operations, sometimes the business project lead. It has to stay readable for someone who will never configure a device.
The LLD is written by the engineers who will do the build, or by the integrator delivering it. It is read by the deployment team, then by operations, then by the auditor who wants to check that a rule or an address range matches what was approved.
The split is useful when those two audiences differ and approval happens in two steps: the architecture is settled before the days of detailed design are spent. That is the case for projects handed to an integrator, where the approved HLD opens the next phase, and for projects that involve several teams.
It is less useful on a small scope. To add two servers to an existing application, a single ten-page document with a design chapter and a detail table avoids maintaining two files that will end up contradicting each other.
How do HLD and LLD relate to an architecture document and a runbook?
Many organizations do not use the words HLD and LLD at all and write an architecture document, sometimes called a solution design. French IT departments call it a DAT (dossier d'architecture technique). Depending on the organization, that document covers the HLD, the LLD or both. Some keep a general and a detailed version, which amounts to the HLD and LLD pair. Others have a single document, and installation detail ends up in a build document or stays in tickets.
The runbook, or operations manual (DEX, dossier d'exploitation, in French projects), answers another question. The design documents say how the solution is built. The runbook says how to run it: start and stop sequences, backups, monitoring, what to do during an incident. It is written for the operations team.
These documents should reference each other without copying each other. The HLD points to the LLD for detail, the LLD points to the runbook for procedures.
How do an HLD and an LLD go stale, and where should they be kept?
The two do not age at the same speed. An HLD holds for years: principles change slowly, and when they do, a project produces a new document. Its risk is to remain valid on paper after an exception contradicted it, such as a direct flow opened between two zones the HLD declares isolated.
An LLD goes stale within months. A virtual machine changes host, a VLAN is added, a version is upgraded, a rule is opened through an urgent ticket. None of these actions goes through the document.
Several places can hold them, each with its strengths. A word processor suits the HLD, which gets approved and signed. For the LLD, a spreadsheet or a wiki is easier to correct, provided an owner is named. A CMDB or an IP address management tool can hold part of the detail (hosts, addresses, VLANs), and the LLD points to it instead of copying it. Infrastructure as code stands in for the LLD for whatever it deploys, since the configuration file holds the detail.
None of them removes the need to check against the sources: compare the LLD tables with a hypervisor export, the cloud inventory and the firewall rules. The article four checks to know whether an architecture document still tells the truth gives the method.
What Uncia brings, and what it does not do
Uncia reads the sources of your IT estate in read-only mode (connectors to clouds, hypervisors, CMDB, firewalls and scanners; dropped-in files such as an RVTools export, a flow matrix or an architecture document) and reconciles them in a knowledge base. You drop in a design document, whether it is called an HLD, an LLD or an architecture document: its diagrams and tables are read as they are, its prose by a language model that can run on your premises. Each element of the document is compared with the sources, and the gaps are shown with the source of every value: a host that changed, an operating system version past its date, an application missing from the exports, a diagram and a text that disagree (see architecture documents checked against the sources).
Those gaps concern concrete elements, so mostly what an LLD contains. An architecture principle or the reason behind a choice appears in no export, and Uncia does not rule on it: whatever no source confirms is reported as undetermined. Uncia changes nothing in your systems or in your documents. The language model proposes, a person decides, and the comparison is updated at each collection.
Frequently asked questions
What does HLD stand for?
HLD stands for high-level design. The document describes the building blocks, zones and principles of a solution, with the reasons behind the choices, and stops short of IP addresses and server names.
What does LLD stand for?
LLD stands for low-level design. The document gives what is needed to install and configure the solution: host names, addressing plan, VLANs, settings and firewall rules.
Do I always need a separate HLD and LLD?
No. The split helps when the architecture has to be approved before detailed design starts, or when the two levels have different readers. On a small scope, one document with a design chapter and detail tables is enough.
Which one is written first?
The HLD. It settles the architecture choices that the LLD then details. Writing the LLD first means approving an architecture through its settings, and fundamental disagreements surface too late.
Is there a standard HLD or LLD template?
No standard imposes one. Templates come from integrators, vendors and internal practice. Most HLDs follow requirements, architecture, choices and rejected options; most LLDs follow naming, addressing, network, security rules and platform settings.