Clustering over a distance

Hi Guys,
Forgive me if this has been addressed before but I could not find anything quickly.
When I did my cluster training, I was told that you could not satisfactorily run a cluster over a distance due to latency issues in the interconnect. So, if you are using long wave fibre, you can get about 10 kilometers and if you are really keen, you can use Dense Wave Division Multiplexing DWDM for up to about 400 kilometers.
I suppose someone has done that and I would like to know if it is as good as they say. I will have lots of data flowing between these sites and DWDM sounds worth investigating especially with something like true copy on a HDS running and lots of IP traffic that might be using the same link.
Comments welcome before I get Sun's sales pitch.
Regards
Stephen

Tim,
Thanks for the reply. I worked for Sun for some time and dealt with companies that obviously had demands for business continuity where campus clustering in particular was not an acceptable choice. Power grids and water supplies needed an appropriate distance factor to provide a service to meet their demands.
When I raised this just over two year ago, Sun kept harping on the ability of DWDM to make long distance clustering and replication of data achievable.
We currently have a requirement for a minimum of 400 million transactions per day (plus whatever we do now) and I am still waiting to see if this is possible even in our building. Even at 50 million transactions, logging is proving to be a bottleneck on our SAN.
I just wonder where we can go with this.
Speaking of blueprints. Someone really needs to ensure the highest level of quality control in blueprints as they are some people at Sun who just want to get their name in print. After reading some of the blueprints, there are a (thankfully) minority that write blueprints that don't have a clue in the real world.
Regards
Stephen

Similar Messages

Monitor vibration over long distance ( 300m)

Hi everyone,
Currently, I have a project to monitor vibration (16 IEPE channels) from a station that is more than 300m away from monitoring station. Looking at NI's products, I have a few options:
1. Using ENET-9234:
- Only 4 channel/module. I need to use 4 modules in total, and need to connect to a router.
- Ethernet distance is limited at 100m. To reach 300m, I need to use at least 2 repeaters, which are not suitable for my project (off-shore environment)
2. Using MXI-4 with Fiber Optic: PXI-8336
- Maximum allowed distance is 200m.NI recommend to add another segment (i.e. add another PXI chassis in between) to extend range to 400m. Or use optical converter to convert between multi-mode fiber to single-mode fiber to extend range
http://digital.ni.com/public.nsf/allkb/C86BF63CC73C6721862575F2005A246D
These recommendations are not suitable either due to harsh contion in the sea and difficult maintenance.
3. Last option is acquire and process data at the site. Then transfer only processed data back to monitoring station over GPIB or RS-485.
I would like to seek your advices on these options. Is there any way to avoid the repeaters in between?
If you have any other ideas, please kindly let me know.
Thank you very much!
Have a great week ahead!
PQH

I also like the cRIO approach. There are copper to fiber media converters on the market that offer single mode fiber which is what you would need for the longer distance.
Check out the following tools to help you get up and running quickly with cRIO:
This listing is an attempt to list and briefly describe core tools that speed the development of waveform acquisition and analysis applications on the CompactRIO and RT platforms.
NI CompactRIO Developers Guide: This is a core developer's guide for a wide range of waveform acquisition and analysis applications. It provides useful background on CompactRIO and control applications in particular. It has chapters on waveform acquisition as well as deployment and management of applications on CompactRIO.
http://www.ni.com/compactriodevguide/
Reference Applications for cRIO Waveform Acquisition: This is the core template for acquiring waveforms for analysis, including vibration, electrical power, and structural monitoring. It makes a great starting point for waveform applications on cRIO.
http://zone.ni.com/devzone/cda/epd/p/id/6206
cRIO Reference Designs for Structural Health Monitoring: This reference builds on the waveform acquisition reference and adds scan engine for slowly changing signals as well as TDMS data logging on event.
http://zone.ni.com/devzone/cda/tut/p/id/9851
Simple Messaging Reference Library (STM): This tool provides a TCP/IP communications protocol which is used in the benchmarks shown on the waveform acquisition reference page. It is an excellent tool for efficient communications between cRIO and a host, easily supporting time waveform live streaming.
http://zone.ni.com/devzone/cda/epd/p/id/2739
Reference Library for Converting Between LabVIEW and XML Data (GXML): This reference provides a remote XML configuration tool, to be used for remote configuration of the cRIO LabVIEW application.
http://zone.ni.com/devzone/cda/epd/p/id/6330
LabVIEW Syslog Protocol Reference Library: This reference offers a UDP cRIO status monitor to indicate to the host the operational status of the remote cRIO
http://forums.ni.com/ni/board/message?board.id=Components&thread.id=51
Reference Applications for cRIO Order Analysis Using a High Speed Tach: This reference offers tachometer or encoder interpretation in FPGA when the I/O device is in a separate and higher speed loop.
http://zone.ni.com/devzone/cda/epd/p/id/6301
DSA Module Synchronization Reference Design for Multiple cRIO Chassis: This reference is useful when the timing of I/O needs to be tightly synchronized between two cRIO chassis. It works with a signal line, as well as with GPS.
http://zone.ni.com/devzone/cda/epd/p/id/6146
Additional NI Systems Engineering Components are listed here: Additional tools can be found on the Systems Engineering Components discussion forum.
http://forums.ni.com/ni/board?board.id=Components
Hope this helps you get started with cRIO, which is a great choice for industrial oil platform applications for temperature, and Class 1 Div 2 certifications.
Preston Johnson
Principal Sales Engineer
Condition Monitoring Systems
Vibration Analyst III - www.vibinst.org, www.mobiusinstitute.com
National Instruments
[email protected]
www.ni.com/mcm
www.ni.com/soundandvibration
www.ni.com/biganalogdata
512-683-5444

Network Management over Long Distances

I've got a question for you experienced OS X Server Admins out there, as well as for more experienced Network Professionals than myself.
As my family increasingly moves to Mac, my tech support requests from them have dropped sharply (Hallelujah!), though the complexity of them when they do come in has increased to the point that I can't really troubleshoot them over the phone anymore. So as I look into getting a Mac Mini, the idea occured that I could use Mac OS X Server and Apple Remote Desktop to manage these remotely in the event of something going wrong, and lock down accounts so they aren't installing apps outside the App Store or Gatekeeper rules. My problem is the physical distance between the Macs involved, and bandwidth usage:
1) The Macs I'd manage are located in several states with different family members, all over different connections of varying quality. A constant VPN option isn't viable, though I need a way to make sure the Mac retains its Security Settings in the event of network connectivity being lost.
2) For when the Macs can/do connect to the Management Server, I'd need a way to make sure they aren't tunneling all traffic over VPN or some other secure connection to the server. In other words, their Mac would need to be able to send traffic from Netflix or Hulu over their local net connection, without the packets being piped over VPN to travel over my line; they need to do this while being reachable by said server.
See my dilemma? I want to manage them as if they're all internal devices, yet also make sure they make use of their own network connections and can remain independent rather than depend on VPN/my network connection.
Is there a feasible solution to this problem, or am I living in a dreamland here?

Having used TeamViewer at home for my Windows box, I've become unimpressed with how it refuses to let me easily disable the service, eats random amounts of CPU and Memory when not in use, and randomly spins up my hard disks to page data. I'm also not thrilled with solutions like LogMeIn or Join.Me, nor do they solve the problem at hand.
Ultimately, I want Domain Admin-like control over these machines at all times, so I can do the routine maintenance, software upgrades, etc that are needed, that way the users (family) can go about their business without trouble. In essence, I want to be the proverbial man behind the curtain.

TelePresence Conductor Clustering over WAN

Hi Cisco,
We want to deploy Clustering of 3 Conductor servers (2 local) and 1 over the WAN in a different DC.
Based on clustering guide 2.x and 3.x it states we need 30 ms RTT between Conductor servers, currently we have 50 ms RTT. Does Conductor 3.0 supports higher RTT? What are the implications of deploying Conductor with higher RTT?
http://www.cisco.com/c/dam/en/us/td/docs/telepresence/infrastructure/conductor/config_guide/xc3-0_docs/TelePresence-Conductor-Clustering-Unified-CM-Deployment-Guide-XC3-0.pdf
Thanks

Thanks Patrick,
I'm surprised this RTT is too low since Conductor just handle mostly signaling. Probably some under the hood DB requirement. I'm thinking of having 2 separate clusters, each in a different Data Center monitoring the same TPS. (Active/Standby), but CUCM will be configured to only select the "standby cluster" when the "closest" Conductor cluster is unreachable.
Do you know if TPS 4.X in remote manage mode supports this scenario?
Thanks

Clustering over Layer 3

Hi,
I have an end-to-end Layer 3 routed access layer network using OSPF point-to-point connections. The design is Cisco 3750s for Access, 6500s Distribution and 7600s Core. There is a likely possibility that some servers will need clustering at the access layer, but reside on different access switches, hence different subnets. The servers aren't traffic intensive and I've been looking at a few options where I can span a VLAN over several access switches for the cluster to work. Cross-site clustering may also be an option and to date i've looked at VPLS, IRB.. If anyone has some options to consider it will be appreciated.
Thanks, WRG

WRG
You may also want to consider L2TPv3 which will allow you to span a vlan across a L3 routed network -
http://www.cisco.com/en/US/docs/ios/12_3t/12_3t2/feature/guide/gtl2tpv3.html
Jon

How do you determine the distance between the QRS interval, as well as the time over that distance, in order to calculate a heart rate

Hi!
I have a continuous sample data acquistion VI created. I collected an ECG signal from electrodes connected to an ELVIS unit. Once I viewed my signal, I am curious to determine the heart rate. I need to know how I can find the R-R interval, as well as the time over than span, so I can roughly determine the heart rate, continuously. Each time a new ECG wave occurs, the heart rate should be modified automatically. Some algorithm is required.
Thanks!

Hello bmeecg,
Welcome to the NI forums! If I understand your question correctly, you wish to calculate the instantaneous heart rate from your ECG (the inverse of your R-R interval). I would recommend using the Peak Detection vi (Functions Palette»Signal Processing»Signal Operation) to determine the location of the R wave (since it is the largest magnitude, you could set a threshold above your P and T waves so that it only picks up the R wave). The calculation for time is in the detailed help for the Peak Detection vi (Ctrl+H»Hover over the VI»Detailed Help).
I also found this forum which has a VI that reads an ECG and then does a lot of analysis. Maybe this would help you out as well.
And just FYI, the Counter/Timer forum is oriented toward our counter/timer boards (e.g. the PCI-6602), the Multifunction DAQ forum probably would have more of the type of people that could help with this issue.
However, please post back here if you have more questions about this.
Neal M.Applications Engineering National Instruments www.ni.com/support

Chaining Thunderbolt Displays over wide Distance

I have one MacBook Pro and two Thunderbolt displays. One display sits right in front of me alongside my Mac, and the other I'd like to place 15' away in my office. Chaining the two displays when they are side by side is no problem. But how do I power/light up, i.e., chain the 2nd Thunderbolt when the distance is much greater than the cord length on the device? If one answer is to place a 2nd Mac in front of the 2nd Thunderbolt well that can happen but I'd like to see if I can drive this from my main Mac first.
Thank you for your suggestions!

Now my next question will be about getting my cable tv to display on one of the Thunderbolts.
I have a Comcast cable box with all sorts of output ports, including optical and HDMI. I also have the latest Apple TV. What do I need to do to: 1) display my Comcast cable on my Thunderbolt as if it were my TV.
I think the general answer here is forget about HD out from your comcast box. HDCP will prevent that. But you can use component, composite, or s-video. Look at Elgato products for TV support for input to your computer.
You won't be able to attach the TB displays to anything other than TB source so I don't know where that leaves you with respect to using it on anything other than the TB computer. Maybe appletv can handle it but I know nothing about that.

Saving files over a distance

At home, 20-30 miles away from where I work, I've got a wireless router connected to an iMac G3 via an ethernet cable ( the iMac doesn't have an airport card ) I also have a 350gig network hard drive connected to the router via an ethernet cable. When I take my MacBook home, I can connect wirelessly to my network hard drive to save my work. My question is: is it possible to configure things so that I can save my work wirelessly to my network hard drive at home from any distance, and if so, how is it done? Thank you

Depends what distance your talking about, but essentially, you need better antannae(s), or repeaters/access points...
One Mile...
http://www.tuaw.com/2006/01/23/extend-your-wireless-range-to-1-mile/
Easy things...
http://guides.radified.com/magoo/guides/wireless/wirelessnetworking07.htm
http://www.networkworld.com/columnists/2005/121205nutter.html
Extending WLAN Range with Repeaters ...
http://www.wi-fiplanet.com/tutorials/article.php/1571601

Distributed deployment migration to Clustering over the WAN

Hello All,
The existing deployment is distributed with different CM versions running on each of them i.e 6.1.3(2 Clusters), 7.1.5(1 Cluster) and 8.6(1Cluster). It has been requested to get these clusters migrated to version 10.x in either centralized or CoW deployment. However, some of the clusters have around 1700 and 2600 phones in the cluster and because of SRST gateway sizing limitation, we have to approach CoW deployment. We have CUCM, UCxn(new dep) and MediaSense(new dep) applications. There are currently around 8000 users registered to the CM but considering future expansion, it must be proposed for about 20K users. My questions are :
1) With UCSS in place, how does the Migration work for different versions? I am not aware of the existing used and unused DLU's in 6.1.3 version and methinks it has to be taken into account to convert into type of Users for UCL.
2) Without UCSS in place, how tricky does the migration become?
2) There are 8 sites, with 5 sites having at least 2000 users. Considering future expansion for 20K users, how do we position the number of subscribers in each site for CoW deployment? I understand there are BHCA and bandwidth calculations involved, 1PuB, 1 SuB, 1 TFTP/MOH servr in DC, 1 SUB each in other sites and an additional TFTP server in one of the sites?
Appreciate the help!
Q

Q,
The first place to start is to look at the cucm 10.X SRND. Look in detail at capacity planning and call processing.
http://www.cisco.com/c/en/us/td/docs/voice_ip_comm/cucm/srnd/collab10/collab10/callpros.html
Each cucm 10.X node can support a total of 10,000 users/devices using the 10,000 OVA template. The cluster can support a total of 40,000 users/devices. I am sure this will meet your present and future needs.
Next you need to look at sizing. The cisco virtualization wiki details what to do in this case..
http://docwiki.cisco.com/wiki/Unified_Communications_in_a_Virtualized_Environment#How_to_Design
Download the sizing tool from here
http://tools.cisco.com/cucst/faces/landing.jsp
Then use the placement tool to determine how you should place your VMs..
You may want to send your final output to Cisco to verify your design. Virtualization is not so straight forward as bare metal server deployments, so you need to design carefully.
From what you have said, looking at about 20,000 users I would suggest the following
1. 4 cucm subscribers--each deployed with 5,000 users/devices. This setup will allow for one to one backup
2. 2 dedicated tftp/moh servers
3. 1 publisher
That gives you a total of 7 virtual machines. This means that you will need two UCS servers-C series servers
Although the sizing tool suggested 12 VMs with six subscribers.. (this means that you will need three UCS servers-C series)
For SRST, depending on how critical the phone system is you either do the ff
1. Use the voice gateway at the site to support SRST. This obviously is limited to the capacity of the voice gateway at the site
2. Re-use the existing cucm servers and deploy them at the local site just for SRST. This will ensure you can support the whole site in SRST
It is advisable to place your UCS servers in your DC. If you have a single DC, then you place one there and then the other at another site (Hq)

Server clustering over WAN

Hi
i am running two Microsoft servers. One is live and the other is in DR site, they are currently on differnt subnets, and for them to cluster, they send broadcast. what can i do for this two servers to talk to each other.
changing ip's on the router is out of the question.

What port number and protocol do they use. If you can assign them on the same subnet you can use IRB (integrated routing and bridging) over WAN.
http://www.cisco.com/en/US/products/hw/switches/ps5304/products_configuration_guide_chapter09186a00800f0a86.html
-amit singh

Can you share your itunes with someone over long distance

can you share your library with someone that isn't on the same wifi network as you?

No. Same wifi and same Apple ID is needed.

Clustering over two separate machines(CPU)

Hi all Oracle Experts out there,
I'm currently working on my a project which requires me to Cluster two weblogic servers on two separate CPU.
When a client accesses an application on one weblogic server and it goes down, the session will be passed to the other weblogic server which will back it up.
Does anyone know how should i go about doing it?
I've been playing around with Oracle Fusion Middleware 11.1.1.1.0 to try to cluster but have been unsuccessful.
Thanks
Gavin
Edited by: user12326632 on Dec 8, 2009 6:55 PM

Hello,
I am trying to do the same thing; that is: setting up a cluster with two machines running each one an instance of WL 10.3.
I wrote a simple Shopping Cart application using servlet and EJB.
I have created a new domain with one cluster, two managed server, one administration server and two machines.
I have then packed the domain and unpacked it in the other machine,
I have configured the replication groups in the first server like this:
primary replication group: myReplicationGroup1
secondary replication group: myReplicationGroup2
And in the second server instead:
primary replication group: myReplicationGroup2
secondary replication group: myReplicationGroup1
I use Apache web server and I have configured it to use the mod_wl_22.so plugin.
I have started the administration server and the two managed servers and deployed my shopping cart application in all the servers in the cluster.
Everything seems to work fine but... when I close (ctrl+C) the managed server running my current shopping cart instance, Apache send the following requests to the remaining server and I get:
"Error 500--Internal Server Error
javax.ejb.EJBException: Could not establish a connection with 3535692784398933269S:192.168.1.10:[7001,7001,-1,-1,-1,-1,-1]:clusteredDomain:server1, java.rmi.ConnectException: Destination unreachable; nested exception is:
(the 192.168.1.10 ip address is the one of the closed server)
So I think that the Http session state is not replicated... and I don't know why...
Am I missing something? Am I doing something wrong?
EDIT: If I deploy only the Web part of my application (the shopping cart servlet) (modified to works without the ejb) it works correctly,
the session replication works pretty good. But when I include my servlet in an ear file with the ejb... It doesn't work...
I have specified the persistent-store-type as replicated in the weblogic.xml descriptor in the servlet war file...
Do I have to set that attribute also in the weblogic-application.xml in the ear archive?
Thanks again
Edited by: user10874329 on 10-dic-2009 7.47

How to speed up SQL*Plus over long distances

We are using Oracle 10g on Solaris 10.
We have a customer that is located several thousand miles and is accessing our database using SQL*Plus as well as a custom java app we developed. Any query our customer runs against our database is as much as 500 times slower than run locally. I ran some test queries to see just how much slower it was.
If I run this query:
SELECT id FROM test_data WHERE ROWNUM < 300000;
(The test_data table contains about 1,000,000 rows of sample data and 31 columns)
It takes about 8 seconds on our intranet. It takes about 1 – 1.3 hours for our customer.
I chose this query because I wanted a query that took at least a few seconds locally. Most of our customer’s queries return less rows, and are more complicated, but this query is easily to work with and shows what I think is the problem.
I also made a java test app the runs the same query and it also takes the same amount of time.
If I spool the results of this query to a file, it creates a 4MB file. If I ftp this file to the customer, it takes only 30 seconds. So it seems like just moving the data is not the slow part; although I am not sure this is a fair comparison.
When I run a ping between the customer and me, the average latency is 250ms. It seems like this high latency may be slowing down the transfer of query results.
I need to make the results return faster in both SQL*Plus and Java. It is not necessary to see any of the results until they have all been transferred.
Is there a way to speed this up, like sending the results all at once instead of a piece at a time? Would incressing the size of the buffers or the size of the Session Data Unit have any effect?
I am not greatly experianced with Oracle, am I misunderstanding the way it transfers data to the client?
Any suggestions would be greatly appreciated.
Tom

I have tried running a few qurys from the remote computer both with FLUSH ON and FLUSH OFF. The querys execute in the same amount of time either way.

Measuring temperature with RTDs over a long distance

I have created an application using the FP-RTD-122 to take RTD mearurements from multiple locations in a plant. The measurments are correct when the RTD leads are connected directly to the module (about 6ft). However when we wire the RTDs in the plant to the module, the measuremtns are incorrect. For instance the temerature is 79C at one location, but shows as between -193C and 19C in Measurment and Automation Explorer. The distance is about 150 ft. I have found a reference in the internet that suggests using a temperature transmitter that converts the RTD signal to a 4 to 20 mA signal. Does anyone have any experience with measuring RTD over long distances? Do I really need a temperature transitter? If so, do I need to use the FP-AI-111 module instead?

Nate,
RTDs allow temperature measurement by reading the resistance of a sensor. This is normally done by exciting the sensor with a known current and measuring the resulting voltage. There are three general types of RTDs; two-wire, three-wire, and four-wire. Two wire RTDs are less accurate because the wire resistance is indistinguishable from the sensor resistance. Three wire RTDs use two wires on one leg (one to carry the known current and one to measure the voltage) allowing one portion of the wire resistance to be nullified. 4 Wire RTDs use two wires on each leg (one for the excitation current) to nullify the wire resistance. As a result, 4 Wire RTDs are the best ones to use for high precision or long distance measurements. To measure 4 Wire RTDs with FieldPoint, you will need to use the [c]FP-RTD-124 module instead of the [c]FP-RTD-122. With that said, I do not know how well 4 wire RTDs will do at 150 feet. 4-20 mA transmitters will easily handle 150 feet so they will probably be your best option. To measure the 4-20 mA signal, you can use the [c]FP-AI-100, [c]FP-AI-110, [c]FP-AI-111, [c]FP-AIO-600, or [c]FP-AIO-610 modules.
Regards,
Aaron

Pros and Cons of Clustering switches

Can someone tell me the pros and cons of clustering switches? It sound like its just for managing multiple switches or is it more trouble that its worth?
TIA

Hi,
well it has advantages as well as some flaws imho.
Advantage:
- Very scalable at low cost. You can add a lot of models onto a stack, ranging from 12port to 48 port, from 100Base-X Uplink to 10Gig-Uplink, with or without PoE with just the buying of a new switch. In modular switches, you have to buy the chassis and then add modules on it which might not come as divers as standalone-switches.
- A somewhat higher redundancy. Spread the uplinks onto different switches in the stack and you'll have nearly no single point of failure (enviromental errors such as power not considered), whereas modular switches always have the backplane as (very rare) spe... and, if your not rockefeller, is the single supervisor engine.
- You can spread a stack (with 3750) over some distance, at least more than on a modular switch. That gives you enough room to implement some cable-routing facilities inbetween the switches.
Disadvantages:
- surely the need for more power outlets, one for each switch
- management, as stated on previous post (well, it's quite similar to modular switches but not exactly, just have a look at snmp, it's a mess on stacks :-( )
- heat dissipation (not really checked on that). X powersupplies generate more heat than Y, if Y < X...
Mathias

Clustering over a distance

Similar Messages

Maybe you are looking for