Member Avatar for shealtiel

I'm building a web application for a small office, to run their daily business. The employees will be logged in and working in the system thorugh the whole day.

My client is located in Israel, and I'm shopping for a hosting cloud provider for him to host the application that I'll build

My options is a server in the US, in Europe, and a local one - in Israel. A local provider is far more expensive.

I need to estimsate how much will the location of the server influence the reponsiveness of the system. And need to do it before I actually order the hosting, install the application and can measure it directly.

Is it a problem at all? Is the latency for a server located over the Atlantic generally felt by users and uncomfortable? How about from one side of Europe to another? (3000 Kilometers by air, from Germany to Israel)

Are there common numbers (like: average ping time between these two countries, or geographical regions, is such and such)

  • 2 Contributors
  • 3 Replies
  • 274 Views
  • 3 Days Discussion Span
  • Latest Post Latest Post by CimmerianX

Dani AI

Generated

For a small-office CRM where staff are signed in all day, server proximity matters because many UI actions translate into small, frequent network round-trips. As noted, each round-trip adds latency to the user’s workflow; design targets from usability research show why this matters: humans perceive ~0.1s as instant, ~1s breaks the flow, and ~10s loses attention. (useit.com)

Typical network numbers vary by provider and path, but region choice usually gives predictable differences: a datacenter inside Israel gives the lowest RTT, a Western‑Europe datacenter is next-best and often a good tradeoff, and US locations generally add substantially more RTT. Provider-region latency tables (example: Azure’s region P50 matrix) show Israel-related pairs spanning tens to a few hundred milliseconds depending on endpoints — so multi-RTT workflows will be noticeably slower when the server is across the Atlantic. (learn.microsoft.com)

A short pre-purchase checklist that avoids guessing: ask each host for a pingable test IP or use their looking‑glass; run ping -c 10 and traceroute from an Israeli location; or schedule RIPE Atlas pings/traceroutes from Israeli probes for real-world vantage points. To reproduce client-side effects in the dev lab, simulate latency with Linux netem (example below). Collect P50 RTT, jitter, and packet-loss numbers for each candidate site. (ui.prod.atlas.ripe.net)

Architectural mitigations reduce the pain if a distant host is needed: collocate app and DB in the same region; batch writes and avoid blocking synchronous calls; use HTTP/2 or HTTP/3 (multiplexing reduces penalty from many small requests); keep connections persistent; offload static assets to a CDN; and push heavy processing into background jobs. These changes often give a much larger UX improvement than small differences in raw bandwidth. (developer.mozilla.org)

Example netem (dev VM):

# simulate ~150ms one-way delay on eth0 (remove with the last command)
sudo tc qdisc add dev eth0 root netem delay 150ms
sudo tc qdisc del dev eth0 root netem

Summary: if budget allows, local hosting is best for an always-on CRM; otherwise Western Europe is a pragmatic compromise. Measure candidate providers before committing and prioritize reducing round-trips in the app.

Recommended Answers

All 3 Replies

Member Avatar for CimmerianX

Latency will be the big issue. And how it affects your program depends on the program.

Does the program need 1 call, 20 calls, 100 calls with heavy DB inserts?

If this is the lifeblood of the system, then it should be located as close to the majority of users as possible. The Bill for the hosting services can be passed onto the customer anyway. The customer may gripe at a slightly larger hosting bill, but they will gripe more if it takes 10 times as long to complete any work process due to high latency issues.

Member Avatar for shealtiel

CimmerianX, it's a simple CRM application with a standard MVC architecture.

Can you explian more your comment?

What do you mean when you ask about the ammount of the calls? Do you mean per hour or something like that?

Member Avatar for CimmerianX

Think of it this way.

If a single write takes 100ms at the close DataCenter 10 non-parallel writes takes 1000 ms.

If the more distant center needs 500 ms for a single write, then 10 non-parallel items take 5000 ms.

If you app needs hundreds of inserts to function, you can see how quickly the lag time would increase.

Be a part of the DaniWeb community

We're a friendly, industry-focused community of developers, IT pros, digital marketers, and technology enthusiasts meeting, networking, learning, and sharing knowledge.