IP communication to Azure SQL server suddenly 10 times slower than usual.

Question

IP communication from our office (data centre) to Azure SQL server suddenly 10 times slower than usual. psping usually take 10 msec or less but now 70 - 100msec resulting some apps very slow.

I can send you psping / tracert results over email.

Regards,

Answer 1

Hi I've been told that our provider found the route changed to Milan somehow then to Cardiff but they didn't understand why 70m (usually italy -> uk never take this long). The problem has solved a few hours later somehow, we can see clean tracert within UK. Thank you for your efforts, please close the ticket.

Answer 2

Hi @Takahiro Yoshida , Hope you are doing well.

a jump from ~10 ms to 70–100 ms that appeared suddenly almost always means the network path changed, not that Azure SQL Database itself degraded. The task is to find which segment of the path added the latency.

On the psping/tracert results

Q&A is a public forum, so we can't take files by email — but please paste the traceroute output directly in private messages.

Please confirm first: are you connecting over the public endpoint, a Site-to-Site VPN, or ExpressRoute? The diagnosis differs substantially, and it's the gating question.

Reading your own traceroute — the fastest path to an answer

Microsoft's network performance guidance recommends isolating to one of three routing domains: your corporate network, the WAN/Internet, and the Azure network. Work from the edges inward:

• Find the hop where RTT jumps by ~60 ms and stays elevated. Everything before that hop is fine.
• Jump at hop 1–3? The problem is inside your data centre — firewall, proxy, or a recently changed route.
• Jump in the middle, at ISP-owned hops? Your provider re-routed. Compare against an older traceroute if you have one; the hop list will differ.
• Elevated only on the last hops entering Microsoft? Then it's worth an Azure-side investigation.

A geographic clue is often decisive: reverse-DNS on the mid-path hops frequently reveals city codes. If traffic that used to stay regional is now transiting a distant city, that alone accounts for 60 ms.

Quick isolation tests

1. Is it SQL-specific? PSPing another Azure resource in the same region (a storage account or a VM). If that's also slow, it's the network path, not Azure SQL — this is the fastest way to split the problem.
2. Is it site-specific? Run the same psping from a different network — a mobile hotspot or another office. Normal latency from elsewhere points squarely at your corporate link or ISP.
3. Baseline the link properly**.** The Azure Connectivity Toolkit wraps PSPing and iPerf for exactly this:

(new-object Net.WebClient).DownloadString("https://aka.ms/AzureCT") | Invoke-Expression
Install-LinkPerformance
Get-LinkPerformance -RemoteHost <target> -TestSeconds 10

This gives you latency and bandwidth figures rather than a single ping number.

1. Rule out an Azure-side event. Check Azure Service Health in the portal for your subscription and region, filtered to the date the slowdown began.

What to check on your side

Since this started suddenly, something changed. Ask your network team about firewall/router firmware updates, routing or BGP changes, ISP maintenance or a failover to a backup circuit, new traffic inspection or DPI, and — if you're on ExpressRoute — whether the circuit failed over to a secondary path or to VPN. Also confirm DNS still resolves your server to the expected regional endpoint.

Meanwhile — reduce your exposure to latency

Erland's point is important: chatty applications are disproportionately hurt by latency. An app issuing 500 sequential round trips loses 30 seconds from a 60 ms increase, while one that fetches in a single batch barely notices. Batching queries, using stored procedures over multiple round trips, and enabling connection pooling won't fix the network, but they will materially reduce the pain while you chase it.

References

• Troubleshoot network performance — Azure ExpressRoute
• Troubleshoot connectivity issues and other errors — Azure SQL Database
• TCP/IP communication troubleshooting guidance

Post the traceroute and your connection type, and we can pinpoint the hop where the latency appears.

Kind Regards,
Microsoft Support Team.