The Traffic Matrix Test: Which Traffic Should Stay Local, Peer Regionally Or Use Transit?

Transit remains essential for full Internet reach. Peering serves a different purpose. It enables networks with mutual traffic to exchange selected routes more directly, often reducing reliance on intermediaries for those flows. The objective is not to replace transit. It is to build a more deliberate mix of local peering, regional peering and upstream connectivity.

The Internet Society describes peering through Internet exchange points as a way to route data more efficiently, keep appropriate traffic local and potentially reduce cost and latency. The result still depends on network topology, the participating peers and each operator’s routing policy.

Start With The Traffic Matrix, Not The Product Catalogue

The right peering decision begins with evidence from the network. A 30-day traffic matrix should combine flow telemetry, routing tables, performance measurements and commercial cost data. No single source is enough. A route with fewer network hand-offs is not necessarily the physically shortest route, and a lower average latency does not prove that the path is resilient at peak load.

For each major source and destination network, review:

Average, peak and 95th-percentile traffic volume, split by inbound and outbound direction.
The destination networks, prefixes and services responsible for the largest recurring flows.
Whether Malaysian traffic leaves the country before returning to a Malaysian destination.
Latency, packet loss, jitter, congestion windows and incident history on the current path.
Transit price, committed capacity, burst exposure and the cost of the next upgrade.
Port, transport, cross-connect, router and operational costs required for peering.
Peering-policy fit, reachable routes, route-security readiness and NOC ownership.

This analysis converts a general ambition such as ‘reduce transit’ into a shortlist of measurable flows. It also prevents low-volume or policy-incompatible traffic from being forced onto a peering path that does not create enough value.

Classify Traffic Into Three Operational Buckets

1. Local peering candidates

These are material, recurring flows between your network and networks that are reachable at a Malaysian exchange. They are strongest candidates when the current path takes an unnecessary international or long-distance detour, when both sides have compatible peering policies, and when there is enough sustained volume to justify the access cost.

2. Regional peering candidates

These flows serve networks elsewhere in Southeast Asia and may benefit from a regional peering platform. The commercial case should include regional transport, route availability, path quality, capacity headroom and operational complexity. Regional reach is valuable only when the advertised routes and actual traffic paths match the design intent.

3. Transit candidates

Transit remains the appropriate path for broad global reach, long-tail destinations, networks that are not available for peering, and any traffic where the peering economics or policy do not fit. It may also remain part of the fallback design. A strong interconnection strategy uses peering and transit as complements, with clear preference and failover rules.

A Neutral IX Changes The Cost Structure,
Not The Laws Of Economics

Peering is not free traffic. An operator still pays for ports, cross-connects or transport, router capacity, monitoring and engineering. The economic advantage comes from using shared interconnection infrastructure to reach multiple relevant networks, instead of carrying every eligible flow through paid transit or building a separate physical connection for each relationship.

The model should use observed 95th-percentile traffic and realistic growth, not total monthly bytes alone. It should also separate hard savings from softer benefits. Transit capacity that can be deferred is measurable. Better route diversity or faster fault isolation may be valuable, but should not be presented as guaranteed financial savings without an agreed valuation method.

Route Control Is The Second Return

Cost is only one part of the decision. A peering path gives the network another routing option. The operator can decide which routes should prefer peering, which should remain on transit, and how traffic should fail over. That control becomes more important as traffic volumes grow and a single congested or indirect route creates a larger operational impact.

The official DE-CIX ASEAN route server guide documents a redundant pair of route servers that facilitate the exchange of routing information between participating peers. It also describes routing tags that control where announcements are distributed, plus RPKI and Internet Routing Registry checks for route hygiene. In practical terms, the route server reduces session complexity while the operator retains routing-policy choices. The user traffic itself is exchanged over the IX fabric, not through the route server.

More route choice does not automatically produce the lowest latency or the best path. The NOC should verify the result through before-and-after measurements, route monitoring and a tested failover procedure.

Keep Suitable Malaysian Traffic Local

DE-CIX Malaysia operates carrier-neutral Internet Exchanges in Kuala Lumpur, Johor Bahru and Penang. For Malaysian networks, a distributed local presence creates more options to exchange suitable domestic traffic closer to the networks serving it. This can reduce avoidable distance and reduce the portion of eligible traffic carried on upstream transit.

The word ‘suitable’ matters. Local peering requires a reachable counterpart, compatible policy, adequate capacity and a path that is genuinely local in operational terms. The traffic matrix should prove these conditions before the business case is approved.

Expand ASEAN Reach Without Rebuilding From Zero

The official GlobePEER ASEAN service combines local peering with access to networks at other DE-CIX Internet Exchanges in Southeast Asia. This gives Malaysian operators a shared platform for regional peering, instead of treating every market as an entirely separate interconnection build.

That does not mean every regional route becomes available or optimal by default. Reach depends on participating networks, the prefixes they announce, peering policies, transport design and capacity.
The peering team should validate the available routes and the NOC should confirm the live path before shifting critical traffic.

A Five-step Operator Plan

Use A Scorecard That Network And Wholesale Teams Both Trust

A successful peering programme should report technical and commercial outcomes together:

Traffic moved from paid transit to approved peering paths, by network and direction.
Change in upstream 95th-percentile utilisation and the timing of the next capacity upgrade.
Median and high-percentile latency, packet loss and jitter to priority networks.
IX port utilisation, peak headroom and growth trend.
Route availability, failover performance and incident impact.
Total recurring cost, one-off implementation cost and operational workload.

This scorecard prevents a single attractive metric from defining success. Lower latency without enough capacity is not a win. Transit offload without resilience is not a win. More reachable routes without commercial relevance are not a win.

Review The Matrix Before Ordering The Next Block Of Transit

Scaling capacity without reviewing the path can scale the same inefficiency. Before the next transit upgrade, identify which networks dominate traffic, which routes leave Malaysia unnecessarily, which peers are reachable locally or regionally, and how much headroom the new design creates.

Bring that traffic matrix to a DE-CIX Malaysia discussion. Review candidate routes, access locations, peering options and capacity requirements, then request an offer based on measured demand rather than assumptions.

Interconnect in Malaysia. Scale Across ASEAN. Connect Globally.

Working Hours: Monday – Friday, 9am – 6pm

Call Us: +603 9212 5961

Email Us: enquiry@de-cix.my