# Onboarding a computing provider

This page connects a **computing provider**, a provider that hosts other firms' servers (a bare-metal host or an on-premises cloud), and the firms it hosts to TheHUB. It is for the provider's infrastructure or network engineer, after commercial terms are signed. For how TheHUB works, see [What is TheHUB](<https://docs.jointhehub.com/what-is-thehub/index.md>).

TheHUB provisions each of your **downstream customers**, the firms whose servers you host, as a [tenant](<https://docs.jointhehub.com/what-is-thehub/index.md#how-thehub-works>).

> [!NOTE]
>
> **Connecting your own equipment as a venue or trading firm?**
>
> Follow [Onboarding a venue or trading firm](<https://docs.jointhehub.com/onboarding/venue-or-trading-firm/index.md>) instead.

> [!NOTE]
>
> **Each tenant is its own delivery**
>
> The provider connection being up says nothing about whether any given tenant behind it is up: each tenant needs its own role, addressing, and checks.

## Before you start

You need:

- Your **placement**, one of two: a cabinet in your own space in the same data center as TheHUB's cage, connected over a **cross-connect**, a fiber link ordered from the data-center operator; or a cabinet in TheHUB's cage. Only equipment in TheHUB's cage receives in-hub market data.
- Your technical contacts for the **kickoff** (the technical meeting after signature where target dates are agreed and your TheHUB contact is named).
- A **role for each tenant** you bring (see [Choose a role for each tenant](<#choose-a-role-for-each-tenant>)). TheHUB cannot issue addressing without one.
- A router that can run external Border Gateway Protocol (eBGP) sessions on whichever attachment you choose for each downstream customer (see [Step 2](<#step-2-interface>)): a routed port per tenant; a routed Institute of Electrical and Electronics Engineers (IEEE) 802.1Q sub-interface per tenant on a trunk, each bound to one virtual local area network (VLAN) tag; or both.

## Choose a role for each tenant

The role decides which other tenants a tenant can reach. TheHUB recognizes two:

| Role | For | Reaches | Does not reach |
| --- | --- | --- | --- |
| Exchange role | A trading venue | Every tenant in the trader role | Any other tenant in the exchange role |
| Trader role | A trading firm | Every tenant in the exchange role | Any other tenant in the trader role |

- **A firm that is both a venue and a trading firm** is provisioned as two tenants.
- **Changing a tenant's role later means renumbering it.**
- **Other participant types,** such as a validator (a blockchain node with a consensus role) or an oracle publisher (a participant that publishes price data onto a blockchain), also take one of the two roles, chosen by what the tenant must reach. The role is agreed at the kickoff.

## Step 1: Physical connection and equalization

Your equipment connects to a **leaf**, a switch where participants plug in; the leaves connect through TheHUB's **spines**, the core switches every flow crosses. Market data and order entry run on [dedicated physical and data paths](<https://docs.jointhehub.com/what-is-thehub/index.md#dedicated-paths>), with failover in the core.

### Where your equipment sits

|  | Your own space in the same building | A cabinet in TheHUB's cage |
| --- | --- | --- |
| How you connect | A cross-connect to the border leaf | A patch inside the cage to a cage leaf, with no Equinix order |
| Your hardware | Stays in your own cabinet, with your existing rack, power, and remote-hands arrangements | Installed in your cabinet in TheHUB's cage (see [Equipment in TheHUB's cage](<#equipment-in-thehubs-cage>)) |
| In-hub market data for your downstream customers | Not available: the border leaf carries no in-hub market data | Available where your first-hop switch runs the multicast control plane (see [Receive in-hub market data](<#receive-in-hub-market-data>)) |
| Resilience | One border leaf, with no second leaf to fail over to | Two cage leaves |

DoubleZero, a private fiber network, is open to downstream customers in either placement (see [Using DoubleZero](<#using-doublezero>)). The routing contract is the same for either placement: the policy, the addressing, and the BGP session do not depend on which leaf you land on.

### Ordering cross-connects

Cross-connects are for equipment in your own space; in TheHUB's cage a patch replaces them, as the last row shows. The number depends on the attachment you choose for each downstream customer in [Step 2](<#step-2-interface>):

| Attachment | Cross-connects |
| --- | --- |
| Dedicated port | One per tenant port, each built to the common fiber length with its own **compensating spool**, fiber cut to make up the length |
| Trunk | One for the provider, with one compensating spool, carrying every tenant on the trunk, each on its own sub-interface |
| Either attachment, in TheHUB's cage | None. Each pre-patched port runs from a patch panel at the top of your cabinet to a cage leaf, which takes dedicated or trunk ports |

Because each dedicated port needs its own cross-connect, the order grows with the number of dedicated ports you take.

To order them, see [Cross-connects from your own space](<https://docs.jointhehub.com/operations/cross-connects-and-site-access/index.md#from-your-own-space>). Your request to TheHUB must come from, or be confirmed by, one of your [Authorizing Officers](<https://docs.jointhehub.com/onboarding/index.md#requests-to-thehub>).

### Equalizing your path

TheHUB builds each path, from your **demarcation** (where your network hands off to TheHUB's cabling) to a given spine, to the common fiber length (see [Equal path lengths](<https://docs.jointhehub.com/what-is-thehub/index.md#equal-path-lengths>)).

- **From your own space,** TheHUB measures each cross-connect run and fits a compensating spool at your demarcation. TheHUB schedules the fitting with you. Fitting it is a patch change at your demarcation, not a re-cable.
- **In TheHUB's cage,** TheHUB runs equal-length fiber to every cabinet, so your cabinet's position does not change your path length.
- Your top-of-rack switches, your downstream customer handoffs, and any oversubscription are on your side of the demarcation, outside the equalized path.
- If you add a cross-connect later, TheHUB builds it to the same fiber length as the first.

### The handoff for dedicated ports

From your own space, each downstream customer on bare metal that takes a dedicated port has a path of its own to each spine.

*Each downstream customer has its own port and demarcation on your side, its own compensating spool, cross-connect, and dedicated border-leaf port, and every path (heavy line) is cut to the common fiber length.*

In TheHUB's cage, each dedicated port is instead pre-patched to a cage leaf over TheHUB's [equal-length in-cage fiber](<https://docs.jointhehub.com/what-is-thehub/index.md#equal-path-lengths>).

### Equipment in TheHUB's cage

If your cabinet is in TheHUB's cage, follow [Site visits and work in the cage](<https://docs.jointhehub.com/operations/cross-connects-and-site-access/index.md#site-visits-and-work-in-the-cage>) before your equipment ships. Your downstream customers get no cage access: TheHUB refers their site-access requests to you.

## Step 2: Interface

You and TheHUB agree autonegotiation and the layer-2 security settings for your port at the kickoff.

### Choose an attachment for each downstream customer

Port speeds are in gigabits per second (G).

|  | Dedicated port | Trunk |
| --- | --- | --- |
| For | Downstream customers on your bare-metal servers | Downstream customers on your cloud (virtualized) servers |
| Shape | One **routed physical port** per tenant | One port, with a **routed 802.1Q sub-interface** per tenant, bound to one VLAN tag and one routing table |
| Separation between tenants | At layer 1: the port is the tenant's alone | By VLAN tag and routing table |
| Speed | 10 G standard, 25 G on request | 100 G |
| Maximum Transmission Unit (MTU) | 1500 bytes, the standard MTU, end to end through TheHUB | The same |

> [!WARNING]
>
> **Do not configure jumbo frames toward TheHUB**
>
> An MTU mismatch fails silently: the link stays up, small packets pass, and large packets are dropped.

- **You can mix both:** a dedicated port for each downstream customer on bare metal, and the trunk for downstream customers on cloud servers.
- **Tenants on a trunk share the port,** so one tenant's burst can delay another's traffic. A tenant that needs capacity of its own takes a dedicated port.
- **A dedicated port at trunk speed for a single downstream customer is not offered.**
- **The routing contract is the same on both.** Policy, addressing, and the BGP session do not change with the attachment.
- **On the trunk, a VLAN tag is significant only to that port,** so it cannot collide with another provider's tags.
- **Each tenant's eBGP session runs from your router, not from the tenant's host,** with a separate routing instance on your side for each downstream customer.

> [!TIP]
>
> **Verify**
>
> Check that TheHUB's allocation table lists the attachment you chose for each tenant and, for each tenant on the trunk, its VLAN tag.

### Keep your tenants apart

Traffic between two of your tenants must cross TheHUB's fabric, as between any two tenants. Isolate your tenants from each other inside your network, so that no path between them bypasses TheHUB. TheHUB asks you to document that isolation.

## Step 3: Network architecture and addressing

For each tenant:

1. Configure its attachment with the values in the allocation table TheHUB sends you.
2. Advertise the tenant's assigned block, and nothing else.
3. Bring up the tenant's eBGP session to TheHUB's Autonomous System Number (ASN).

### What TheHUB supplies

TheHUB allocates every value below and sends you an allocation table generated from its registry, the record its switch configuration is rendered from, so what you are told matches what is configured. **Do not derive any of these values yourself.**

| Value | Source |
| --- | --- |
| Each tenant's address block | TheHUB's public Internet Protocol version 4 (IPv4) space, inside the aggregate for the tenant's role |
| Each tenant's peering link | A /31 from Request for Comments (RFC) 6598 shared address space (100.64.0.0/10), from a block reserved for you. Never advertised |
| Each tenant's ASN | A private ASN per tenant, from a block TheHUB delegates to you |
| TheHUB's ASN | One ASN for every tenant on every leaf, whichever attachment the tenant uses |
| Each tenant's VLAN tag on your trunk port | Allocated per tenant. Trunk only |
| For a tenant that uses DoubleZero: the device code of TheHUB's DoubleZero device (DZD), the value of the DoubleZero client's `--device` flag; the DZD's two user-tunnel endpoint addresses, where tenants' tunnels terminate; and the tenant's **DoubleZero client address** (the address from its block that its host sends DoubleZero traffic from) | In your allocation table (see [Using DoubleZero](<#using-doublezero>)) |

### Advertise exactly your assigned block

TheHUB's inbound prefix-list for each tenant accepts only the block registered for that tenant, and each session has a maximum-routes ceiling, a cap on the number of routes TheHUB accepts on it.

> [!WARNING]
>
> **Two ways to lose reachability**
>
> - **Advertise the block, not host routes.** TheHUB's allocation table states the size of each block it assigns, and TheHUB's policy accepts prefixes of that size. A host route (a /32) does not match and is silently rejected.
> - **Coordinate any renumber with TheHUB first.** If you move hosts to new addresses before TheHUB updates its side, the prefix-list rejects every unregistered prefix and tenant reachability is down until TheHUB catches up.

You may translate addresses inside your own network, as long as you advertise the block you were assigned. Address translation must not change a DoubleZero host's DoubleZero client address: the access pass is bound to it. If your hosts use a default route to the internet, pin test traffic to the tenant block: source any test from an address in the tenant's block, or the traffic never crosses TheHUB at all.

### Session

Each tenant runs eBGP from its own ASN to TheHUB's single presented ASN, set out in your allocation table. The value is the same on every leaf.

You do not send anything to be identified: routes arriving from each tenant are tagged inbound at the leaf with a BGP community that identifies the tenant.

**The AS path you receive carries only TheHUB's ASN.** TheHUB's internal structure, and the origin ASN of other tenants' routes, are hidden. Do not use the AS path to tell tenants apart.

*The bring-up sequence: identity, the prefix-list, and the maximum-routes ceiling are applied at the leaf before any route propagates, and a spine returns the opposite role's aggregate and the DZD endpoints, and nothing else.*

## Receive in-hub market data

This section applies if your servers are in a cabinet in TheHUB's cage. TheHUB distributes in-hub market data as Internet Protocol (IP) multicast (see [Market data](<https://docs.jointhehub.com/what-is-thehub/index.md#market-data>)).

For a downstream customer on your servers to receive it, the **first-hop switch**, the switch that customer's server connects to, must run the multicast control plane:

- It takes the server's Internet Group Management Protocol (IGMP) membership reports.
- It peers with TheHUB in Protocol Independent Multicast (PIM) to carry the join toward the source.

Without that, the join never reaches TheHUB. Agree with your TheHUB contact the multicast groups for each feed and where TheHUB's end of the PIM peering sits. The multicast control plane is needed for in-hub multicast only; DoubleZero needs no multicast support on your network (see [Using DoubleZero](<#using-doublezero>)).

Everything inside your network, including the path from your first-hop switch to a downstream customer's server, is yours.

## Using DoubleZero

Using DoubleZero is optional (see [The DoubleZero device](<https://docs.jointhehub.com/what-is-thehub/index.md#the-doublezero-device>)). DoubleZero carries IP multicast as well as unicast traffic, and your network carries each downstream customer's DoubleZero tunnel as ordinary unicast traffic.

| Task | Who does it |
| --- | --- |
| Request the access pass and hand over the host's details (the steps below) | You, the provider |
| [Set up the DoubleZero client](<https://docs.jointhehub.com/market-data/doublezero-client/index.md>) on the host | The downstream customer's engineer |

For each downstream customer that uses DoubleZero:

1. **Request its [access pass](<https://docs.jointhehub.com/market-data/doublezero-client/index.md#before-you-start>),** DoubleZero's entitlement record, bound to one address, through your TheHUB contact. Ask for the pass to be bound to that downstream customer's DoubleZero client address and to name the group codes, DoubleZero's identifiers for multicast groups, for the feeds it publishes or subscribes to, as agreed with your TheHUB contact.

   > **Verify**
   >
   > Check that the pass you receive is bound to that downstream customer's DoubleZero client address.
2. **Give the downstream customer what its host needs:** its access pass and, from your allocation table, its DoubleZero client address, the DZD's device code, and the two user-tunnel endpoint addresses.

**Next:** [Verify and go live](<https://docs.jointhehub.com/onboarding/verify-and-go-live/index.md>), for each tenant.

## See also

- [Cross-connects and site access](<https://docs.jointhehub.com/operations/cross-connects-and-site-access/index.md>): ordering cross-connects, and working in TheHUB's cage.
- [Onboarding a venue or trading firm](<https://docs.jointhehub.com/onboarding/venue-or-trading-firm/index.md>): the same handoff for a participant connecting its own equipment.

Last reviewed 2026-11-20.
