Network Latency

I've been working with VMWare on an issue with our system, and after looking through the logs, they found that some of the heartbeat packets from the ESXi hosts to the vCenter VM are getting lost. They suggested that our network was not enough to handle the traffic. I don't believe this to be true. We have 8 UCS B200 M3 blades across 2 chassis, with 4 10gb uplinks. The FIs are hooked cross linked into two Nexus 5548UP switches, which are then crosslinked into our core Nexus 7009s.
The vCenter and ESXi hosts are on two different VLANs, but the core switches are far from overloaded. I've been told to use Wireshark and perform a packet trace, but I'm not sure which switch to grab. I'm also not sure what I'd even be looking for during that time. Any help on what to look for or even where to look would be appreciated.

Hello Renx_08 and welcome to community,
You may try this and check if there is any success.
I assume the network is on wireless mode.
Kindly go to Control Panel -> Power Options and click "Change plan settings" for your current power plan.
On the next screen click on "Change advanced power settings"
On the next screen expand "Wireless Adapter Settings" and then "Power Saving Mode"
Make sure it is set to Maximum Performance.
Best Regards,
Tanuj
Did someone help you today? Press the star on the left to thank them with a Kudo!
If you find a post helpful and it answers your question, please mark it as an "Accepted Solution".! This will help the rest of the Community with similar issues identify the verified solution and benefit from it.
Follow @LenovoForums on Twitter!

Similar Messages

Looking for clarification on network latency issue vs drive mapping

Hi,
I am seeing this as mystery and not getting crystal clear idea on the reason for the issue. Issue is related to the performance of the application interms of time it is taking in processing the input file.
I wrote a swing application, which is a client application. Which takes some parameters like server name and iphostaddress and connects to the Process Server which is, responsible for processing client application requests. Client application will communicate with process server through TCP/IP connection and process the input file and returns the decisions back to the user through the output file.
Below is the scenarios I am using for launching the application:
1. If both client application and server are running locally in my desktop the time it is taking to process the input file is 2 minutes.
2. If client is running my application and server is running remotely on wondows server, it's taking 13 minutes to process same input file.
3. To reduce the time in scenario2, I installed the client appliation also on the remote server ( so that both client and server application are running on the windows server). and mapped the server's share drive to my desktop. And launched the application from my desktop (from U drive, where application is mapped), now it's taking 10 minutes to process the same input file.
I am struggling in understanding why it's taking that long in scenario 3. Because application is installed locally on the server and input file and output files also copied onto the U drive. Sometimes thinking am I launching the application in the right way or not?
Can somebody explain me, if we launch the remote java application through drive mapping will there be network latency there eventhough everything is there on the server (U drive)? Here I need to tell one more scenario 4, If loginto the remote windows server and launch the client application time it's taking to process the same input file is about a minute.
Below are some more details on the issue: I am not encoding the file, I am using third party application, which provides an API to communicate with the process server. Just using the API methods and classes to pass the input file data to server. I have used the 'tracert' command for the remote server and I am seeing 8 hops between my desktop and remote server. I even installed network sniffer tool in my laptop and captured the files when application running.
The input file has 140000 records (text lines with comma delimited) of 6.271MB in size. I have posted to understand the time it is taking in scenario3, where evrything is on mapped drive (i.e, client application and input file are technically recides on the server right?) , but client application is launched from desktop. The reason I am doing this way is, instead of log-in into the remote server, user can easily launch the application from the desktop. So, when I try to launch the application this way, this doesn't count as if client application is running local to the server? Will it becomes remote (I have even captured the network traffic file in this scenario too, and I have seen the comminication between my desktop ip address to server ip address and server is taking abour 3.84 milliseconds for each item to respond to client, I think it's just travel time not the process time). I am assuming, even when application launched from drive that is mapped, it should take about 1 minute (the time taking when I launch the aplication after log-into the server,not through drive mapping) to prcess input file as everything is on the server.
Thanks in advance,
Jyothi

reading and writing the data shouldn't be the problem, its what you are doing with the data which will be taking all the time.
Try this
public class WriteFile {
    public static void main(String... args) throws IOException {
        String filename = "record.csv";
        int records = 140 * 1000;
        int values = 6;
            long start = System.nanoTime();
            PrintWriter pw = new PrintWriter(new BufferedWriter(new FileWriter(filename)));
            for (int r = 0; r < records; r++) {
                for (int i = 0; i < values; i++) {
                    if (i > 0)
                        pw.print(',');
                    pw.print(r * 10 + i);
                pw.println();
            pw.close();
            long time = System.nanoTime() - start;
            System.out.printf("Time to write %,d records was %.3f sec. file size=%.3fMB%n",
                    records, time / 1e9, new File(filename).length() / 1e6);
            long start = System.nanoTime();
            BufferedReader br = new BufferedReader(new FileReader(filename));
            String line;
            while ((line = br.readLine()) != null) {
                // do some work.
                String[] parts = line.split(",");
                int[] nums = new int[parts.length];
                for (int i = 0; i < parts.length; i++)
                    nums[i] = Integer.parseInt(parts);
br.close();
long time = System.nanoTime() - start;
System.out.printf("Time to read %,d records was %.3f sec%n",
records, time / 1e9);
PrintsTime to write 140,000 records was 0.462 sec. file size=6.193MB
Time to read 140,000 records was 0.792 sec

Global UCCX deployment - maximum network latency permitted?

We are looking at a global deployment of UCCX. The number of agents will be small, around 25 in total, but globally dispersed.
What is the maximum network latency UCCX can tolerate?
We would be looking to have the UCCX server, PSTN connection and CUCM in a datacentre in London and agents in London, Toronto, Singapore and Santiago.
There would be QoS set on the WAN but latency between London and Santiago could be an issue as it may be 200ms or worse.
Any feedback would be appreciated.
Thanks
Andy

Above 150ms, or esspecially 200ms your voice quality will begin to degrade. This is not a CCX limitation.
What will really get you is the amount of CAD signaling. It is a "chatty" application; delay will slow it down. Here is an excerpt from the CCX SRND:
Impact of delays inherent in the applications themselves. 25 seconds is the initial Cisco Unified
CCX agent login setup time with no WAN delay. The overall time to log in agents and base delay
adds approximately 30 seconds of delay per 30 milliseconds of WAN delay.
From page 108 of the Solution Reference Network Design for Cisco Unified CCX and Cisco Unified IP IVR, Release 7.0.
I do not recall there being a documented maximum in the SRND because the BU likely expected the voice RTT delay limit of 150ms to become a problem first.
Also remember that all reporting in CCX is recorded and displayed in server time. Global deployments could be somewhat annoying because all reporting for an agent in another timezone would have to be mentally offset by the person viewing the report. There are similar issues with time-of-day/day-of-week code within scripts but these can be overcome with some work at least.

Impact of network latency

We have developed an application based on j2EE architecture and are running it on iWS and iAS. I would like to know the impact network latency would have on the application.
What should be the threshold of the network latency for our application's good response ?
Regards,
Vibha

Network latency between where and where?
Between the user and iWS? Between the iWS and the iAS? Between the various iAS servers?
Latency between the user and iWS is pretty unavoidable. Except for technologies like Akamai, there isn't much you can do about it.
Latency between the iWS and the iAS is somewhat negative. The impact on the user is pretty much linear. i.e. an increase of 100ms latency between iWS and iAS is likely to cause an increase of 200ms in the perceived response time. (100 ms inbound and 100 ms outbound) The problem is that the application also has to keep the processing thread around another 200 ms because of the network latency. This means that are going to see some degradation in the performance of the server.
Large latencies between the iWS and iAS are relatively easy to resolve, however. Just increase the size of the network pipe and/or increase the capacity of the firewall between iWS and iAS.
Latency between various iAS instances is VERY bad. Large latencies between iAS server can interfere with clustering both degrading performance and data consistency. You should definately resolve any network latency between iAS servers.
David
http://www.amazon.com/exec/obidos/ASIN/076454909X/

Monitoring network Latency

I want to write a script using "fping" to monitor my network latency and packet loss. since on the server I am running the script don't have root permission I can't use ICMP the other option I have TCP or UDP depend on available open port on router/switch . my question is there any draw back using TCP or UDP ping in terms of result quality and traffic? I am really appreciate fyour feedback. thanks. paul

TCP and UDP "ping" typically rely on port 7 being open on the target devices. This is almost never the case these days. You can enable the UDP and TCP echo services on IOS devices with the command "service {udp|tcp}-small-services" but many consider them a security hole. What would be best is if you can convince your admin to setuid the fping executable to root (i.e. chmod 4555 fping). This way fping will be able to open raw sockets to produce ICMP traffic. You are certainly more likely to have this work than relying on the echo service being available.
Another alternative is to procure a small IOS device and run IP SLA on it. With IP SLA you can test simple ICMP echo, or use more advanced probes such as TCP, UDP, jitter, HTTP, DHCP, and DNS.

Windows Server 2012 Hyper-V network latencies

Hi All,
I have an issue with our Windows Server 2012 Hyper-V hosts that I can't seem to figure out. Situation:
2 x Dell PowerEdge R815 servers with AMD opteron 6376 16 core CPU's and 128 GB RAM running Windows Server 2012 with Hyper-V.
2 virtual machines running on the same physical host and connected to the same virtual switch show high TCP connection latencies. One virtual machines runs a SQL Server 2012 database instance and a Dynamics AX 2012 R2 instance. The other machine a
SharePoint 2013 instance and the AX client. We see latencies of 20ms and higher on most of the TCP connections that are made from the sharepoint machine to the sql server machine.
At first I thought it might have something to do with the physical NIC's. It turned out that VMQ wasn't correctly supported by the firmware of the Broadcom BCM5709c cards. By default this setting is enabled. Turning off the VMQ setting somewhat improved
the situation but the latencies are still at 8ms and higher.
What I don't understand is what influence enabling/disabling VMQ should have on network performance. As I understand it now virtual machines connected to the same virtual switch bypass the physical altogether. Another point is that VMQ should actually improve
performance, not decrease it.
Another question I have is about the various tcp offloading settings on the physical NIC's. After installing the new firmware and drivers from Dell most of these settings are set to disabled. The documents I have been able to find talk about Windows Server
2008, any thought how these settings relate to Windows Server 2012 and whether they should be enabled?
Thanks in advance for your time and thoughts
Kind regards,
Dennes Schuitema

Hi Denes,
Please try to update your BroadCom NIC driver version ,the newest version should be 7.8.51
For details please refer to following link :
http://www.broadcom.com/support/ethernet_nic/netxtremeii.php
Best Regards
Elton Ji
If it is not the answer , you can unmark it to continue .
We
are trying to better understand customer views on social support experience, so your participation in this
interview project would be greatly appreciated if you have time.
Thanks for helping make community forums a great place.

How to confirm the Network latency in welclogic server

Hi,
Please help me on below query.
If a production server is having a network bottle neck or network throttling. How can i confirm? and What would you do?
Request you to guide me on this.
Thanks,
Raj

Hi,
N/W Issues
If your WebLogic Servers are part of Cluster and if you see any Multicast related errors or warnings like ... ManagedServer is getting Kicked of of Cluster and then again Joining after some time...or if you see any Multicast Lost Messages more than 2-3 times...it means there may be some N/W Issues ...to test this you can use MulticastTest ot MulticastMonitor Utilities.... Like this u will get some idea abt n/w issues.
N/W latency
WebLogic Servers logs/Access Logs ... will never be useful to identify the N/W Latency.
Thanks
Jay SenSharma

DataGuard question - overcoming network latency problem

Hi
Dataguard for 10.2.0.4
Assuming we have a latency problem with the network between the primary and secondary databases in a dataguard configuration.
That means that my bottleneck is currently the network that cannot accommodate the amount of redo that is generated per second by the primary, and archive logs are accumulating on the primary site
Is there a way to configure the redo transport on the primary to transfer the archive logs in parallel – in case there is an accumulation of archive logs waiting to be shipped
I am talking parallel transfer to the same site , not to different sites – just to overcome the network bottleneck
Thanks in advance
Orna

Hi Sybrand,
Quoting from the doc
'Oracle Database 10g Release 2 introduces the ability to have multiple ARCn processes sending redo for a single archived-redo log over the network, reducing the time required to ship a single archive log to a remote destination. In previous Oracle Database releases, only one ARCn process at a time could archive redo from a given log file. The maximum number of network connections that will be used in parallel to perform archival to a destination is controlled using the MAX_CONNECTIONS attribute of the LOG_ARCHIVE_DEST_n initialization parameter. The LOG_ARCHIVE_MAX_PROCESSES initialization parameter can be set as high as 30, (the previous maximum was 10)'
My 'back of a cigarette packet' calculation was not meant to be exact just meant to illustrate a point :-)
I still think your confusing bandwidth with latency but I don't want to poke an angry response :-)
Cheers,
Harry
Edited by: Harry on Jul 10, 2009 3:33 PM - seems the language filter doesn;t like my British slang. Changed to cigarette......

Network Round Trips when loading a Form - reducing Network Latency

I am working on a Forms migration from Client-Server to WebForms running on Application Server 10.1.2.0.2
Does anyone know the minimum number of network round trips that can be achieved when loading a Form?
For most of the Forms in the migration I am seeing 3 network round-trips before the Form is displayed and on Forms that use Webutil I am seeing 7 round-trips before the Form is displayed. Is this normal or are there ways of optimizing the round-trips down?
The Forms code has already been optimized by removing SYNCHRONIZE commands and TIMER commands and I have already refered to metalink note 363285.1 which gives lots of good advice however there is no specific information on the minimum number of round trips that you can expect for loading a Form.
Thanks for any help you can give.
Philippe

Ok, a bit of an update is needed.
There were a number of posts below that have been lost due to an accidental deletion but I shall try and summarize them here:
Tony Garabedian - Posted a number of code optimizations that he'd used for speeding up code when querying within a Form. Tony if you are able to re-post those that would be great :-)
Grant Ronald - posted some more information in which he mentioned that the extra pings are used to synchronize between the App.Server and the client. He also commented that a network that uses satellite was inherently slower and would therefore require a different front end to handle the increase in latency.
For my part I said nothing new other than ask loads more questions. I did mistakenly comment that we are using satellite for our network however upon yet more discussion with our network guy it turns out that we're using optical cable for everything. Out guaranteed round trip time is about 320 milli-seconds to Australia from the UK.
Currently I am left with the understanding that we are not likely to go below 3 round trips for most of our Forms and in the case of Webutil we are likely to be limited to 7 round trips.
Of course if anyone out there has any specific experience with this issue and can comment further please feel free.
Many thanks,
Philippe

How to calculate network latency ?

Hello fellow programmers,
I need to calculate the latency between a client and a server using both TCP and UDP. I have built a TCP client and a server which accepts and sends a String using the writeBytes(String) and readLine() function.
The problem am facing is to send specific number of bytes like 500, 1000, 2000 to calculate latency since the client and server accept Strings.
Any suggestions ?

I just realized what you are really asking. You are asking how you can create a String that will write out to 500 bytes right? I think as long as you are using ascii encoding the nubmer of byte and number of characters in the String should be the same. It's easy to check. Use the getBytes method to see what the length the default encoding produces given a String 500 chars long.

What is acceptable level of Private Interconnect Latency for RAC

We have build 3 Node RAC on RHEL5.4 on VMware.
There is node eviction problem due to loss of Network Heartbeat.
ocssd.log:[ CSSD]2010-03-05 17:48:21.908 [84704144] >TRACE: clssnmReadDskHeartbeat: node 3, vm-lnx-rds1173, has a disk HB, but no network HB, DHB has rcfg 0, wrtcnt, 2, LATS 1185024, lastSeqNo 2, timestamp 1267791501/1961474
Ping statistics from Node2 to Node1 are as below
--- rds1171-priv ping statistics ---
443 packets transmitted, 443 received, 0% packet loss, time 538119ms
rtt min/avg/max/mdev = 0.150/2.030/630.212/29.929 ms
[root@vm-lnx-rds1172 oracle]#
Can this be reason for Node eviction? What is acceptable level of of private interconnect latency for RAC ?

What is acceptable level of of private interconnect latency for RAC ?Normal local network latency should be enough. By the way latency settings are very generous.
Can you check if your to-be-evicted node runs and is reachable when seeing the node eviction messages?
In addition to that: Can you check the log files of the eviced node. Check for time stamps around "2010-03-05 17:48:21.908". Make sure all systems are NTP synchronized.
Ronny Egner
My Blog: http://blog.ronnyegner-consulting.de

Thunderbolt latency?

Hello everyone!
I was going to buy a Mac Pro but would instead get a new MacBook Pro if there were any low latency network adapters available. It doesn't have to be available now but I would need it in the near future. Since Thunderbolt is a PCI Express interface nanosecond scale network latency (networking / interconnects, however you want to look at it) should be possible. Has anyone here heard if this type of peripheral is in the works for Thunderbolt?
Thanks!

Adapters
Promise SAN Link Fibre Channel adapter
Sonnet Allegro FireWire 800 adapter
Sonnet Presto Gigabit Ethernet adapter
http://www.apple.com/thunderbolt/

High latency causing problems... and frustration w...

I have a femtocell that requires the latency on any single hop on the internet to be less than 240ms, any longer than this and it drops my phone call. On contacting the supplier of the femtocell, they asked me to use Traceroute to see how long the path to a server was (I used Google.com) and many of the hops were over 240ms, some as long as 4922ms.
On scanning this forum, there are other users out there (mostly gamers) who are suffering from latency problems. Has anyone had any success resolving the problem? I can send copies of the Traceroute outputs that clearly show high latency between routers with addresses such as interconnect2-gig1-0.manchester.fixed.bt.net, core2-pos0-6-1-0.ealing.ukcore.bt.net and core4te-0-7-0-0.telehouse.ukcore.bt.net, to anyone that may be able to help.
The best I've so far had from BT is that:
- It's a problem with the femtocell supplier: no, the supplier requested from ISPs their longest expected latency prior to releasing the femtocell and set the timeout to be equal to this
- It's a problem with my Mac mail: the fact that I use a Mac has absolutely nothing to do with network latency
- There's interference between my mobile phone and my wireless: this isn't the case (I'm a wireless communications engineer by background) and, if it was, almost no-one would be able to use mobile phones given the number of wireless access points in homes and offices.
Every time I get a call from the call centre it's someone different, who doesn't understand the problem, doesn't know what a femtocell is and hasn't read the notes which clearly state that I want to be called on my mobile, not on my home phone. It's looking like my best option may be to try to find another ISP who doesn't have this problem.

I think you have the result of what is maybe a 2 fold issue.
Femtocell technology is still a relatively new technology as you;ll know from your job.
If you think about the security issues of running a femtocell, that;s to say what it actually has to fulfil,
Security for femtocell networks spans several distinct requirements. The service provider must authenticate users as they arrive on the network. The RF link between the handset and the femtocell must be secured for both user and control plane traffic. And lastly, the mobile network traffic must be placed into a virtual private network as it traverses the wired ISP network to ensure that the traffic is protected while transiting this public network and only authorized users can forward traffic to the mobile operator's network.
There is however one very important element of femtocell security which makes the implementation significantly more complex. This relates to latency as you say, which must be carefully managed especially for applications such as VoIP/SIP. Compounding this challenge is the unknown nature of the latency across the ISP network, which has resulted in service providers requiring latency in the femtocell to be minimized, as you say.
If you add to that the fact that the BT broadband system is based on an algorithm packet handled system to ensure data quality across maybe noisy telephone lines, it all compounds to add to the problems of the running of the femtocell.
In other words if you have a highly error corrected and interleaved broadband line, it's not going to help the femtocell out by adding additional latency to what is already a fairly complicated situation where data handling is concerned....

How to determine network bandwidth for EP 6.0 implementation

Hi
can you please guide me how to determine the network bandwith requirment for SAP EP6.0 implementation for smoother operation
Thanks & Regards
Ashutosh

Hi Ashutosh,
luckily network bandwidth is mostly not a problem as long as it is not very small, e.g. a modem dial-up 56kbps connection on client side. To make sure that the server side outgoing line is strong enough, you'd need to do some math and calculate from the expected request volume how much bandwidth you'd need. Some guidance might be taken out of the how to paper under this link:
https://www.sdn.sap.com/irj/servlet/prt/portal/prtroot/com.sap.km.cm.docs/documents/a1-8-4/how to perform sap enterprise portal load testing.zip.
Check chapter 2.4 in the pdf file.
For "smooth operation" you'd need network wise not only worry about bandwidth but also about network latency times and sometimes network quality in regard to % package losses or similar. If your end users are more then 1000 miles away you usualy start seeing response time increases from latency times...
The impacts of networking on performance are many, but so are the tuning possibilities. If you have a lot of end-users accessing EP over a WAN, perhaps even over the Internet, and if they do so over long distance you might want to make EP landscape WAN tuning a dedicated task in your project.
In addition the network requirements of Portal are discussed in this paper : http://service.sap.com/~form/sapnet?_SHORTKEY=01100035870000757275&_SCENARIO=01100035870000000112&_OBJECT=011000358700003943521999E , chapter 3.5.2 Enterprise Portal.
Best Regards, Joerg

Gathering network statistics on specific tcp ports

I have an application on Solaris 10 with one local zone and it listens on a few tcp ports.
I want to write a script to find out if a network latency occures on one of these ports.
I dont want to use "time telnet ..." because in that way I'll need to sample many servers from one place.
netstat -sP tcp also does not fit my needs, because I want to distinguish between tcp ports.
Thanks a lot

I have an application on Solaris 10 with one local zone and it listens on a few tcp ports.
I want to write a script to find out if a network latency occures on one of these ports.
I dont want to use "time telnet ..." because in that way I'll need to sample many servers from one place.
netstat -sP tcp also does not fit my needs, because I want to distinguish between tcp ports.
Thanks a lot

Network Latency

Similar Messages

Maybe you are looking for