DplyMini currently offers 5 nodes in Singapore, Japan (Tokyo), South Korea (Seoul), Hong Kong, and the US West. Each node is a dedicated physical machine with compute resources reserved for you—not a virtual machine.
Both available configurations are offered at every node: DeployMini M4 16 with M4, 16GB RAM, and 256GB SSD; and DeployMini M4 24 with M4, 24GB RAM, and 512GB SSD. Every catalog combination can be ordered; availability is confirmed in real time by the portal.
The map shows the geographic distribution of nodes available to order. It does not split the US West into fictional cities or mistake test-origin cities for available regions. All five nodes have equal weight in the catalog, so choose based on your workflow—not card order.
The connections highlight the need to consider your team’s access point, code repository, signing services, artifact storage, and end users together. The graphic does not represent a private network topology or guarantee a fixed route.
Each regional run sheet lists the traffic it is suited to cover, both available configurations, and an ordering link. “Suited” is a planning recommendation, not a substitute for testing from your actual office network.
Service region: Southeast Asia
A strong fit for business workflows in Singapore and Southeast Asia, and a candidate node for Australian teams accessing Asian repositories and artifact services.
Service region: East Asia
A strong fit for teams in Japan, Japan-focused testing, and continuous builds that need to stay close to repositories, signing services, or artifact storage in Tokyo.
Service region: Northeast Asia
A strong fit for teams in South Korea, interactive desktop work across Northeast Asia, and projects that build, test, and return artifacts over networks around Seoul.
Service region: South China & Southeast Asia
A strong fit for workflows between South China and Southeast Asia. If your team, repository, and artifact services span multiple regions, check both uplink stability and cross-region routing.
Service region: North American West Coast
A strong fit for teams on the North American West Coast, North America-focused artifact workflows, and projects whose repositories, automation services, and final test users are mainly in North America.
Median latency shows typical round-trip time, while P95 highlights slower samples. Each cell shows “median / P95” in milliseconds. Seoul, Shanghai, Sydney, Frankfurt, and Los Angeles are test-origin cities only—not additional nodes available to order.
| Test-origin city | Singapore Median / P95 |
Japan (Tokyo) Median / P95 |
South Korea (Seoul) Median / P95 |
Hong Kong Median / P95 |
US West Median / P95 |
|---|---|---|---|---|---|
| Seoul | 68 / 92 ms | 34 / 49 ms | 8 / 14 ms | 42 / 61 ms | 136 / 168 ms |
| Shanghai | 62 / 91 ms | 48 / 76 ms | 51 / 79 ms | 31 / 55 ms | 149 / 188 ms |
| Sydney | 95 / 127 ms | 121 / 151 ms | 134 / 169 ms | 112 / 145 ms | 146 / 178 ms |
| Frankfurt | 164 / 198 ms | 236 / 279 ms | 221 / 261 ms | 194 / 232 ms | 151 / 182 ms |
| Los Angeles | 178 / 216 ms | 112 / 142 ms | 127 / 158 ms | 151 / 185 ms | 26 / 39 ms |
Interactive graphical work is more sensitive to jitter and P95. Background builds typically depend more on sustained throughput for code pulls, dependency downloads, and artifact uploads.
A fast connection from your team to the node does not mean the node will be equally fast to your repository or artifact storage. Test the access path and external dependencies separately before deployment.
Access networks, congestion, egress routing, and test time can all change the result. Verify again from your own office network.
Testing was completed on August 22, 2026. For each “test-origin city → available node” pair, we sent 100 ICMP requests of 64 bytes each, recorded the round-trip time of successful responses, and calculated the median and P95.
Each origin used a local commercial broadband or fixed-line IPv4 public egress, with no VPN, proxy, or traffic acceleration. Paths involving different legal regions were treated as cross-border tests; a local city-to-local node path was not.
The results represent the test window and the corresponding origin network only; they are not a guarantee of fixed latency. Real projects are also affected by DNS, TLS connection setup, repository response times, object-storage throughput, file size, and concurrency.
A single ping answers only a very small question. Before ordering, map your team access, code, signing, artifacts, and end users as one chain, then look for the node that is most stable overall.
For frequent graphical desktop work, prioritize developer-to-node P95, jitter, and evening peak performance. Background-only builds can place less weight on interactive latency.
Compare full clones, incremental pulls, and submodule access. Large repositories are often more affected by sustained throughput and small-file round trips than by a single ping.
Archiving and signing may involve multiple network requests. Verify certificate tools, time synchronization, and connectivity to required services from the target node.
IPAs, symbol files, videos, and model files depend more on upload throughput. Test with sanitized files close to real sizes instead of drawing conclusions from tiny samples.
If your workflow includes checking pages, APIs, or media for a specific market, include the end-user region in your node choice—not just the engineers’ location.
Compare team-to-node P95 first, then verify the node’s connection to repositories and signing services.
Compare sustained throughput from the node to repositories, dependency sources, and artifact storage first.
Estimate data-transfer volume first, then evaluate the node, rental period, and additional storage together.
Each link opens the same configuration flow with the corresponding region preselected. You’ll still need to confirm the model, rental period, and additional storage; live availability is returned by the portal.
Both dedicated physical configurations are available across all five nodes, with daily, weekly, monthly, and quarterly rental options. All charges are settled in US dollars (USD).