For a business in Kota Kinabalu or Kuching, hosting in Singapore adds a round trip that no amount of optimisation removes. The speed of light over that distance is not a configuration parameter. Teams that treat latency as a tuning problem eventually discover it is a geography problem, and geography is decided long before the first line of application code is written.
We hear the same story from east customers in several forms. A point of sale terminal that pauses after every scan. A dashboard that refreshes slowly enough that operators stop trusting it. A nightly backup window that creeps past dawn because the link to a peninsula data centre is the slowest part of the job. Each one is a latency bill paid in staff time.
Where the milliseconds actually go
A round trip between Sabah or Sarawak and Singapore crosses submarine cable systems, lands at a cable station, and then travels through whichever terrestrial and peering path the network of the day selects. The optical distance is fixed, but the path is not. Routing changes, transit arrangements change, and congestion on a shared segment changes with the hour. That variability is what makes the delay feel unpredictable rather than merely long.
- Sabah or Sarawak to Singapore: a round trip that varies with the routing of the day
- Sabah or Sarawak to Klang Valley: a shorter domestic path
- Same-region: the round trip disappears from the budget entirely
- Same zone: what remains is switching, not distance
It helps to separate a round trip into its components, because they respond to different fixes. Propagation is the time light spends in fibre. Serialisation is the time the slowest link takes to place bits on the wire. Queuing is the time packets wait behind other traffic. Switching and routing add a small fixed cost at every hop. Only the first is physics; the rest are engineering, and engineering is where a provider can change the answer.
That is why we publish measured round trip figures rather than a single marketing number. A figure taken at three in the morning on an empty path is not the figure your staff meet at ten on a Monday. We report median and 95th percentile latency between our east Malaysian zones and our Klang Valley region, and we keep the measurement running so the trend is visible rather than a snapshot taken once for a sales deck.
The shape of a cross-sea path
Capacity between east and west Malaysia rides a small number of submarine systems, each with its own landing points and its own failure modes. The path a packet takes depends on the commercial and technical arrangements between the networks it crosses. When one segment is congested, or under repair, traffic is re-routed, and the alternative is usually longer. Where we can, we also interconnect directly with the networks that carry east Malaysian access traffic, so packets avoid a transit provider in the middle.
This is the practical reason a domestic path matters. Traffic from Kota Kinabalu to Klang Valley that stays inside Malaysia crosses fewer commercial boundaries, so there are fewer points where a policy set by someone else can send it the long way round. Where we can, we hold capacity on more than one system, and we engineer our own routing so that a single cable fault degrades performance rather than ending it.
What the delay costs in practice
Interactive applications feel latency first. Every round trip is a pause a user can perceive, and interfaces that make several round trips to draw a single screen multiply the delay. At a few tens of milliseconds the effect is invisible. At a few hundred, people change how they work. They batch tasks, they avoid the screen, or they keep a local spreadsheet that quietly becomes the real system of record.
Backup and replication windows are where latency costs the most money in the least visible way. A job that moves a large volume of data over a long path is limited by the round trip as much as by the bandwidth you bought, because reliable transfer protocols wait for acknowledgements. Teams respond by buying a bigger link, which helps less than they expect, or by shortening retention, which helps in the wrong direction.
What we run in the east
We run capacity in Sabah and Sarawak so that east Malaysian workloads can be served from inside east Malaysia. Where a workload must reach the peninsula, it takes a domestic path rather than leaving the country. Our eastern sites carry the same redundancy we apply elsewhere: dual power feeds, generator backup with fuel arrangements, cooling sized for a single unit failure, and diverse paths out of the building where the local market offers them.
Latency is not the only cost. For organisations that must evidence where data is processed, keeping east Malaysian data in east Malaysia simplifies the answer. Workloads stay under Malaysian law, including the Personal Data Protection Act 2010, and the reply to an auditor asking where a record lives is a location inside the country rather than a jurisdiction to be argued about.
Regional is not a synonym for near.
How to measure before you commit
Before you move a workload, measure the path you already have. Run round trip tests from your own premises to the candidate regions for a full week, not an afternoon. Look at the 95th percentile rather than the average, and look at how the figure moves through the working day. A provider that will not share measured figures before you buy is telling you what to expect afterwards.
- Measure from your own network, not from a public looking glass
- Run for at least a week to capture the working day
- Compare 95th percentile, not the best case
- Test the application, because round trips multiply
If the numbers support it, move the latency sensitive parts first: terminals, interactive systems, anything a person waits on. Batch workloads can follow at leisure, or stay where they are. We schedule migrations during MYT business hours, give at least seven days of notice before planned maintenance, acknowledge support requests within fifteen minutes, and publish a postmortem within five business days when something goes wrong.
Our availability commitment is 99.95%, our invoices are in Ringgit, we accept FPX and TnG eWallet, and customer data stays in Malaysia. Those commitments matter more on a long path than on a short one, because a long path has more places to fail and fewer people who can see it failing. That is the honest case for keeping east Malaysian traffic inside east Malaysia: not that distance disappears, but that the number of things between your users and their data goes down.
You cannot optimise a cable into being shorter, but you can stop paying for the detour.
