Checking for TCP/IP packet loss

Occasionally, a computer cannot back up files in full, due to the network connection. A poor network connection can result in corrupted data packets which are then automatically retransmitted, thereby slowing down the overall data transfer rate. Indications of this include; timeouts during backups, extremely long backup times for small amounts of data, and server disconnects.

This article explains two methods that may be used to determine if a user has large amounts of TCP/IP packet loss due to their network connection. These methods are, Performance Monitor and Netstat. It also includes advice for Reducing the effect of packet losses. However, if large TCP/IP packet losses are found then, a user may need to work with the IT group, computer vendor, network administrator, or Internet service provider (ISP) to resolve the issue.

Note:
The examples in this article use a desktop PC running Windows 8. The principle remains the same for other versions of Windows.

Performance Monitor

The Performance Monitor can be used to check for TCP/IP packet losses. To launch the Performance Monitor user interface, proceed as follows.

1. Right click on the Windows icon in the bottom left hand corner of your screen, and select Run from the pop-up menu.

2. In the Run dialogue box, type Perfmon.
3. Click OK.

The Performance Monitor window opens on the System Summary page.

4. In the left hand navigation pane, click Performance Monitor.

The real-time Performance Monitor window opens.

5. Click on the green cross icon  to open the Add Counters dialog box.

6. Make sure that, under Available counters, the <Local computer> option is selected. If not, then select it from the drop-down options.

7. In the Performance object box, locate TCPv4, then click the down-arrow next to it.

A list of TCPv4 counter options appears.

8. From the list of counter options,
   1. Select Segments Sent/sec, then click the Add button.
   2. Select Segments Retransmitted/sec, then click the Add button.

The two counters are moved into the Added counters pane.

9. Click OK.

Let the Performance Monitor run while the backup occurs and starts to indicate problems.

10. Check the Average value for Segments Retransmitted/sec. In the example below, this is 0.071 segments retransmitted per second.

Any persistent Average value of 0.1 or more indicates too many dropped packets, and points to a poor network connection.

Netstat

Use the Netstat command to view details of your network communication and packet loss during backups. 

Before backing up the system, capture the current Segments Sent and Segments Retransmitted as follows.

1. Right click on the Windows icon in the bottom left hand corner of your screen, and select Command Prompt from the pop-up menu.

The Command Prompt window opens.

2. In the Command Prompt window, type netstat –s –p tcp.

3. Press the Enter key.

The TCP Statistics for IPv4 are displayed.

4. Make a note of the current values (before backup) of Segments Sent and Segments Retransmitted.

5. Start a backup of the account.
6. With the backup in progress, repeat Step 2 to Step 4, to capture new values for Segments Sent and Segments Retransmitted.
7. Close the Command Prompt window.
8. Subtract the "before backup" Segments Sent and Segments Retransmitted values from the "after backup" values to get an idea of how many times packets had to be retransmitted during the backup. For example:

So, in the example shown, during backup there were 16415 segments sent and 49260 segments retransmitted. This represents a retransmission percentage of 300%.

This is a clear example of packet lost causing the local system to continue retransmitting segments to the datacentre. This will cause slow backups, timeouts, long backups and other connectivity related issues.  Due to packet loss like this, agents attempting to backup one gigabyte end up trying to send multiple gigabytes.

Reducing the effect of packet losses

The effect of corrupted data packets being automatically retransmitted, slows down the overall data transfer rate and adversely affects backup performance. If large TCP/IP packet losses are found then you may need to work with your IT group, computer vendor, network administrator, or Internet service provider (ISP) to resolve the issue.

Large Send Offload

One action that may be taken to reduce the impact of packet losses is to reduce the packet (or segment) size. Retransmitting a small packet takes less time than retransmitting a large packet. To reduce the number of large packets or segments, proceed as follows.

1. Right click on the Windows icon in the bottom left hand corner of your screen, and select Control Panel from the pop-up menu.

The Control Panel opens.

2. Click Device Manager.

The Device Manager opens.

3. In the list of devices, locate Network adapters, and click to expand it.

4. In the expanded list of Network adapters, locate your primary network adapter.

Note:
If you are unsure which is your primary network adapter then return to the Control Panel and select Network and Sharing Center. In the left hand pane of the Network and Sharing Center window, click Change adapter settings. A Network Connections window opens showing all your network adapters. The primary adapter is the one not marked by a red cross.

5. Right click on your primary network adapter and, from the pop-up menu, click Properties.

The network adapter Properties window opens on the General tab.

6. Click Advanced.

Note:
If there is no Advanced tab or, later, no Large Send Offload (IPv4) entry, then you have not selected your primary network adapter. Return to the list of network adapters and try another. If you still encounter problems, contact Cloud Direct Technical Services on 0800 368 1831, or email TechServices@clouddirect.net.

The Advanced tab opens.

7. In the Property pane, highlight Large Send Offload (IPv4).
8. In the Value pane, click the drop-down box, and select Disabled.

9. Click OK.

This disables the ability to send large size packets, and so reduces the impact of packet losses.

Note:
To learn more about Large Send Offload, (also known as Large Segment Offload) visit Wikipedia.com or better still, see a description from Peer Wisdom.