Cluster design

          my requirements:
          host1: webapp1, webapp2
          host2: webapp2, webapp3
          host3: webapp3, webapp1
          cluster1: host1, host2(for webapp2)
          cluster2: host2, host3(for webapp3)
          cluster3: host3, host1(for webapp1)
          my question:
          1. if no other host as a webserver(s), how to configure it? such as virtual
          hosts, httpclusterservlets, clusters.
          2. if i place a web server(located at another machine(host4)) before the 3 hosts,
          where to configure virtual host, on web server or management server behind it?
          it's urgent. detail description is welcome. thanks in advance.

Thanks aton Jaime Valencia & Aman Soi
Now I understood. Publisher and subscriber in a cluster should have same version.
Here is the hardware details
MCS7835I3-K9-CMD2
HW Platform      : 7835I3
Processors      : 1
Type                : Intel(R) Xeon(R) CPU       (E5504 @ 2.00GHz)
CPU Speed         : 2000
Memory           : 4096 MBytes

Similar Messages

Cluster Design Example

Does anyone have a link to an example of a cluster design document? For example, a visio diagram showing ESXi hosts, networking, storage and all the various details of a cluster design that you need to document?

There are sample case studies documents available at Prepare for the VCDX Certification.
You can through with these document, and get familiar with the process.
Thx

Updates on Cache, SEO, Mobile compatibility and Cluster Design for EM

Hi,
Was wondering if anyone has done the Experience Manager implementation and may have the answers to following questions:
1. Do we have any EM cache or if it can rely on Endeca cache for performance improvements?
2. Can EM help in any way for SEO?
3. How can we make sure that the content of our site is rendered properly in mobile devices also? Is it, whenever I create a page template, I have to create another replica for mobile devices?
4. Any insight of EM design in cluster environment? Will that be only a single instance on Endeca ITL with cartridge renderers on every App server or will there be an equivalent number of EM instance as that of App servers?
Any info on any of the topic would really be helpful. Thanks in advance for this.
Regards,
Vaibhav

1. Do we have any EM cache or if it can rely on Endeca cache for performance improvements?There is nothing like ExpMgr cache. Moreover these are not comparable. For performance improvements in data access time you can tune MDEX (assuming, that is what you mean by Endeca above), for performance improvements related to front end you will have to tune the front end application & the app server it is hosted on.
2. Can EM help in any way for SEO?In short no. But Endeca can help, they have SEO API's included into the assembler API, which will can take care of many aspects of SEO. Check out the [SEO Module|http://docs.oracle.com/cd/E35648_01/index.html] documentation.
3. How can we make sure that the content of our site is rendered properly in mobile devices also? Is it, whenever I create a page template, I have to create another replica for mobile devices?Sort of. Either you use 'Responsive Web Design' so that the same template works on different devices OR create separate templates for different category of devices..
4. Any insight of EM design in cluster environment? Will that be only a single instance on Endeca ITL with cartridge renderers on every App server or will there be an equivalent number of EM instance as that of App servers?You will have one WorkBench(Exp Manager) but your site will be deployed on each dgraph.
HTH, get back if you need more details.
Pankaj.

Cluster Design (11g)

Hi there!
I am working on the design of a cluster that will support more than 200 databases with smalls SGAs, my question is, which are the considerations to keep in mind knowing that many databases implies greater amount of interconnect traffic. If anyone has experience running over 100 databases on the Oracle Clusterware please share your experience, good or bad, unfotunately found no documents explaining which are the guidelines or best practices to implement a cluster with these characteristics. Workload design recommendations are welcome.
Thanks in advance.
J

hi,
this is not direct answer but gives some idea
http://www.dba-oracle.com/t_optimal_number_rac_nodes.htm
I think that tuning in this case will be reactive rather than proactive, having set some rough boundaries as there are many variables/Dbs.

Cluster design with subscriber 10.X on VM to exisiting MCS 7835I running with v8.6

Hi
At present CUCM is running with single node MCS 7835I (v8.6) in cluster. Now planning to implement cluster by building subscriber with version 10.X on VM to existing MCS 7835I which remains as publisher.
Does cluster support dissimilar hardware and software mismatch setup ?
Also having plans to upgrade MCS H7835I to 9.1 later, is it possible ?
Thanks in advance

Thanks aton Jaime Valencia & Aman Soi
Now I understood. Publisher and subscriber in a cluster should have same version.
Here is the hardware details
MCS7835I3-K9-CMD2
HW Platform      : 7835I3
Processors      : 1
Type                : Intel(R) Xeon(R) CPU       (E5504 @ 2.00GHz)
CPU Speed         : 2000
Memory           : 4096 MBytes

Error while generating Data provider

We are facing a weird issue while using Design Studio 1.2. Locally the tool works fine but when we use it from BI-Launchpad or through BI Enterprise it gives intermittent error "Error while generating Data Provider"
Same setup works fine in our Validation Environment.
We have tried running repair for Design Studio, redeploy the war files, creating new view but issue still exist.
We have BI 4.0 SP7 installed on Windows 2008 R2 system.
4 CMS running in cluster, Design Studio 1.2 is installed on 3 Nodes out of 4.
For now we are using 1 tomcat to access the system to narrow down the issue.
Experts please help us to resolve this issue.
Attach is the error message we are receiving

I am not clear what your data source is, but I recommend checking the PAM https://websmp102.sap-ag.de/~sapidb/011000358700001013822013E because there are some limitations with Design Studio and BI4.0

Storage spaces + Hyper-V with multiple 1GBe nics for storage?

Hi guys!
So I just got my private cloud hardware. I actually put in the order before summer, but due to firmware and certification issues on my desired SuperMicro JBODs delivery was seriously delayed. So much that I've completely forgotten my networking ideas. I
need help/verification. Or at least a URL - most described setups are 10 GBe nowadays... Or even a "not gonna work" :-)
My setup is supposed to be a 3 JBOD, 2 head node storage spaces/sfos cluster providing storage to a 4 node Hyper-V cluster. I didn't have a budget for a 10 GBe setup, but got a great price on a lot of 1 GBe nics. After allocating management, Hyper-V, etc
I have 3x 1 GBe ports left on all Hyper-V and Storage servers.
I think my original plan was to create three subnets and add one nic from each server. And then I guess I've imagined some kind of SMB3 magic discovering these paths between Hyper-V and storage and just aggregating bandwitdh and providing fault tolerance
by sprinkling fairy dust. Must have been the heat...
So now I'm "replanning" and I realize that I'm going to create a failover cluster at the storage level providing a cluster name and IP. I'm thinking the management subnet where domain info resides is appropriate, but then what about the other three
subnets? I don't want to flood my management subnet with storage traffic, but do want bandwidth and resilience. Did I make a design error, and how do I make the best of the situation?
Disclaimer: My previous experience on virtualization clusters is ISCSI SAN and 2008 R2 Hyper-V clusters. Storage Spaces is completely new to me :-)
And due to overlapping technologies I struggled a bit on placing this thread. Hope I got it right

Hello,
i did not understand how many NICs you have in each Host. Hyper-V Cluster with 1 GBe NICs work as long as you know that it is not 10 GBe.
In this article is the complete Hyper-V Cluster design in checklist form. I think you should work with this list for some further ideas:
http://blogs.technet.com/b/askpfeplat/archive/2013/03/10/windows-server-2012-hyper-v-best-practices-in-easy-checklist-form.aspx
Sorry that i cant give a better answer, but i lack information about about your environment.
Regards,
Thomas
Thomas Hanrath [MCT | Regional Lead Germany] |
http://www.hanrath.de
Microsoft Learning Blog |
http://blog.microsoftlearning.de
MCSE | Private Cloud

Device Weight: Route Pattern Vs IP Phone

Hi, we are starting the planning for removing two CO class GTD5 switches that route 28,000 DID numbers for the County. The switches are +16 years old and maintenance is expensive. Our Callmanager cluster is designed to replace the existing phone infrastructure (2 CO switches, +60 PBXs, and +20 key systems). Here's a quick overview of our Cluster:
- Publisher only running database services (MCS7845H2)
- Two TFTP servers only running TFTP service and MOH streaming (multicast) (MCS7835H)
- Two subscribers configured as 1-to-1 backup (MCS7845H2)
- CCM version 4.1(3)sr3b
We are planning to add the additional six MCS7845H2 subscribers as we need the capacity.
The first step in our migration is to move all the DID's to new T1/PRI's on a set of six Communication Media Modules (CMM) spread out over four Cat6513 switches.
Essentially our CCM cluster will be acting as a Tandem switch until we get all the trunks moved off the GTD5 CO switches to the CMMs.
Unfornutately all of the 28,000 DIDs are pretty much shot gunned all over the County. So, we will have +20,000 route patterns in the beginning. Over time we will also be converting sites to IP Tel and removing the route patterns as we migrate.
My question: Does a route pattern for one directory number carry the same device weight as an IP phone with the directory number assigned to a line?
We are thinking it does and if it does, then we need to scale up our CCM cluster for 30,000 devices before we start the migration.
Thanks in advance for any advice.
Tom.

Thanks Greg for taking the time to reply. Our cluster design is following the "Cisco Callmanager Best Practices" book. The Publisher and TFTP servers are not running the Callmanager service. This allows 8 subscribers running the Callmanager service in the Cluster.
Our research is trying to understand the cost of a route pattern in terms of the dialing forest and the impact on the subscriber's memory and CPU.
We own two complete prefixes plus another 8,000 DNs from a third prefix. We fortunately do not have an overlapping dialing plan. Each directory number will either be assigned to an IP phone or it will belong to phone on a PBX.
All directory numbers for IP phones belong to the same partition and we also followed the Best Practices book for our dialing plan and use the line/device CSS design.
All +80 remote sites connect back to our GTD5 and the GTD5 routes all the numbers to the remote sites.
We are first migrating all the DID services to our new PRI's handled by the Callmanager cluster. We when start the migration it will be a simple process of a route pattern such as [4-5]XXXX to route the 874-xxxx and 875-xxxx numbers to the GTD5. Then as we disconnect the tie-lines from each remote PBX and re-connect it to the CMM for Callmanager to route, we will need to add all the specific route patterns to route the numbers for the site.
It would be ideal if we did not have to retire the GTD5 switch. We would follow our 5 to 7 year plan to migrate the entire County to IP Tel and leave the GTD5 in place routing the numbers to the remote PBXs. However, we have been directed by management to decommission the GTD5 switch within 12 months.
So we are trying to understand the impact to the subscribers when we begin adding 1,000's of route patterns. We are planning to consolidate as many of the route patterns as possible to reduce the number of route patterns. However, we inherited a design that we refer to as "Number-lose-ability", where individual numbers are routed and not blocks of numbers. Over the years of adds, moves, and changes the numbers have been scattered to all the sites. We have very few sites with consecutive numbers.
Another question that we are trying to answer: what is the cost of a route pattern such as 5555x compared to 10 individual route patterns for the same number range. Again, in terms of memory and CPU on the subscriber doing the digit analysis. We are asking this question because we may have 6 of the 10 numbers going to the same PBX, but the other 4 numbers each going to a different site. To consolidate route patterns we would add the 5555X pattern and the four individual route patterns. What we do not know is how the 5555x is added to the dialing forest. Is it expanded to 10 patterns or just one expression.
Thanks again for any help,

OAS: Building a HA environment with OAS

Hi all,
we have tio scale up our OFM capacity asap. Because migration to WLS in a short time is not possible, we have to prepare OAS Cluster + SOA 10g environment. We are working at this moment in a single node environment.
I'd like to know about best practice for OAS-Cluster design and realization.
Best Reagrds,
Moh

we will use active/active 3 named instance FCI in primary site
name of instance are instance_a, instance_b and instance_c
how i will accomplish this in primary?
how i configure only one SQL in DR site for 3 named instane?
how i will configure AOHA on both primary and DR site?
I am not expert please help me step by step if possible with snap shot.
Hi ABDUL,
As Edwin’s post, build Windows Server Failover Cluster (WSFC) firstly, for more details, please review this
blog.
Secondly, build SQL Server Failover Cluster, you can follow the steps in this similar blog:
SQL 2012 Failover Cluster Build (Step by Step).
Thirdly, add SQL Server AlwaysOn Availability Groups to the existing SQL Server Failover Cluster Instance (FCI), please review this similar blog:
Adding SQL Server AlwaysOn Availability Groups to existing Failover Clusters.
Thanks,
Lydia Zhang

Dedicated switches needed for RAC interconnect or not?

Currently working on an Extended RAC cluster design implementation, I asked the network engineer for dedicated switches for the RAC interconnects.
Here is a little background:
There are 28 RAC clusters over 2X13 physical RAC nodes with separate Oracle_Home for each instance with atleast 2+ instances on each RAC node. So 13 RAC nodes will be in each site(Data-Center). This is basically an Extended RAC solution for SAP databases on RHEL 6 using ASM and Clusterware for Oracle 11gR2. The RAC nodes are Blades in a c7000 enclosure (in each site). The distance between the sites is 55+ kms.
Oracle recommends to have Infiniband(20GBps) as the network backbone, but here DWDM will be used with 2X10 Gbps (each at 10 GBps) links for the RAC interconnect between the sites. There will be separate 2x1GBps redundant link for the Production network and 2x2 GBps FC(Fiber-Channel) redundant links for the SAN/Storage(ASM traffic will go here) network. There will be switches for the Public-production network and the SAN network each.
Oracle recommends dedicated switches(which will give acceptable latency/bandwith) with switch redundancy to route the dedicated/non-routable VLANs for the RAC interconnect (private/heartbeat/global cache transfer) network. Since the DWDM interlinks is 2x10Gbps - do I still need the dedicated switches?
If yes, then how many?
Your inputs will be greatly appreciated.. and help me take a decision.
Many Thanks in advance..
Abhijit

Absolutely agree.. the chances of overload in a HA(RAC) solution and ultmate RAC node eviction are very high(with very high latency) and for exactly this reason I even suggested inexpensive switches to route the VLANs for the RAC interconnect through these switches. The ASM traffic will get routed through the 2x2GB FC links through SAN-Directors (1 in each site).
Suggested the network folks to use Up-links from the c7000 enclosure and route the RAC VLAN through these inexpensive switches for the interconnect traffic. We have another challenge here: HP has certified using VirtualConnect/Flex-Fabric architecture for Blades in c7000 to allocate VLANs for RAC interconnect. But this is only for one site, and does not span Production/DR sites separated over a distance.
Btw, do you have any standard switch model to select from.. and how many to go for a RAC configuration of 13 Extended RAC clusters with each cluster hosting 2+ RAC instances to host total of 28 SAP instances.
Many Thanks again!
Abhijit

"#" key as the first digit of site prefix code

The plan at present is to implement the centralized CCM cluster design with a main office and 10 remote sites in Asia where the first digit of site prefix code is "#" follows by two digits of country code and 4 digit internal extension. So that a remote site in Tokyo would be #816200. Can "#" key be used as first digit of site prefix code? Thanks.

Yes, you can use the # to route calls.

Design Problem : How to design/code a java client which is deployed to all the machines in a cluster and make sure that only one of the java client is active

          hi ,
          I have to design a java client (which is basically a JMS message listener)which
          is deplloyed to all the servers in the cluster. But as these are message listeners,
          i want only one of the instance to be active at a time.
          If the server on which the client is active goes down , I want the second server
          to start listening to messages.
          How do i design this ? Also is there a public api for multicasting that we can
          use ?
          Anybody has an idea on how to go about this..
          Thanks
          nisha


Hi Nisha,
          Failover message listeners? Sounds like you want MDBs, which are deployed on all nodes in a
          cluster. If your JMS destination is a queue, then only one MDB will pick up the message. And just
          like any other ejb service, MDBs failover.
          Gene
          "Nisha" <[email protected]> wrote in message news:[email protected]..
          hi ,
          I have to design a java client (which is basically a JMS message listener)which
          is deplloyed to all the servers in the cluster. But as these are message listeners,
          i want only one of the instance to be active at a time.
          If the server on which the client is active goes down , I want the second server
          to start listening to messages.
          How do i design this ? Also is there a public api for multicasting that we can
          use ?
          Anybody has an idea on how to go about this..
          Thanks
          nisha

Can I design a SOFS-cluster using non-clsutered storage?

Hi.
I've been trying to figure out if I can build a SOFS-cluster with 4 nodes and 4 JBOD cabinets per node for a total of 16 cabinets, like this:
I haven't seen this design though so I'm not sure if it's even possible and if it is, what features do I lose (enclosure awareness, etc)?
Thanks.

Yeah, I was in a hurry when I posted my initial question and didn't explain my thought process clearly enough.
Are you saying that you can't build a unified CSV namespace on top of multiple SOFS-clusters, despite MSFT proposing this exact design at multiple occasions?
As for building one-node clusters; it's certainly possible, albeit a bit pointless I suppose unless you want to cheat a bit like I did. :)
The reason I'm asking about this particular design is that the hardware vendor that the customer wants to use for their Storage Spaces design only supports cascading up to 4 JBOD-cabinets in one SAS-chain.
As their cabinets only support at the most 48 TB per cabinet and the customer wants roughly 220 TB usable space in a mirror config that gives us 10 cabinets. On top of this though we want to use tiering to SSD and with all those limitations taken into consideration
we end up with 16 cabinets.
This results in 8 server nodes (2 per 4 cabinets) which is quite a lot of servers for 220 TB of usable disk space and hard to motivate when compared to a traditional FC-based storage solution.
Perhaps not the cost, pizza boxes are quite cheap, but the rack space for 8 1U servers and 16 2U cabinets is quite a lot.
I'll put together a design based on these numbers and see what the cost is though, perhaps it's cheap enough for the customer to consider. :)
Thanks for the feedback.
1) I'm saying we did not ever manage to have unified namespace from multiple SoFS with no shared block storage between all of them, we did not find any references from MSFT how to do this, we did not find any people who had done this either. If you'd search
this particular forum you'll see this question asked many times but not answered (we also did ask). If you'd manage to do this and share some information how I'd appreciate this as we;re still interested. See:
SoFS Scaling
http://social.technet.microsoft.com/Forums/windowsserver/en-US/20e0e320-ee90-4edf-a6df-4f91b1ef8531/scaling-the-cluster-2012-r2
SoFS Architecture
http://social.technet.microsoft.com/Forums/windowsserver/en-US/dc0213da-5ba1-4fad-a5e4-091e047f06b9/clustered-storage-spaces-architecture
Answer from the guy who was presenting this "picture" to public on TechEd:
In this specific example, I started by building up an at-scale Spaces deployment - comprising of "units" of 2-4 servers, attached to 4 SAS JBODs, for a total of 240 disks. As scaling beyond those 240 disks with the 2-4 existing servers would become impractical
due to either port connectivity limitations of the JBOD units themselves, or limitations of the servers due to PCI-E or HBA limitations, further scale is achieved by adding more units to the cluster.
These additional units would further comprise of servers and JBODs, but the underlying storage connectivity (Shared SAS) exists only between servers and JBODs within individual units. This means that each unit would have it's own storage pool,
and it's own collection of provisioned virtual disks. Resiliency of data and creation of virtual disks occurs within each unit.
As there can be multiple units with no physical SAS connectivity between them, Ethernet connectivity between all the cluster nodes and cluster shared volumes (CSV) presents the means to unify the data access namespace between all the cluster nodes regardless
of physical connectivity at the underlying storage level - making the logical storage architecture from a client and cluster point of view completely flat, regardless of how it's actually physically organized. Further, as you are using scale-out file server
atop of CSV (with R2 and SMB 3.0) client connections to the file server cluster will automatically connect to the correct cluster nodes which are attached to the clients’ data.
Data availability and resiliency occurs at the unit level, and these can be extended across units through a workload replication mechanism such as Hyper-V replica, or data replication mechanisms such as DFSR.
I hope this helps clear up any confusion on the above, and let me know if you have any further questions!
Bryan"
2) Sure there's no point as single node cluster is not fault tolerant which compromises a bit whole idea of having a cluster :) Consensus!
3) Idea is nice except I don't know how to implement it w/o third-party software *or* SAS switches limiting bandwidth, and increasing cost and complexity :(
StarWind VSAN [Virtual SAN] clusters Hyper-V without SAS, Fibre Channel, SMB 3.0 or iSCSI, uses Ethernet to mirror internally mounted SATA disks between hosts.

Cluster Version control Design question

Hi All,
I have an application where I store the data in a cluster. Suppose this is my ver. 1 of the application and the cluster consists of 3 I32. and I ship this application to the customer.
Now for the ver 2 I want to store a string to the cluster so I updated my cluster to 3 I32 and 1 String. But this ver 2 should also be able to read ver 1. just like MS Word 2007 is able to read .doc files
In my design, I am stroing the version number and data(as variant) in a cluster and convert that cluster to variant and do a 'variant to flattened string' before saving it to a file.
I am attaching my desing protoype here so you get a better idea of what I am talking about and suggest me if you think it can be done in a better way.
Thanks,
Ritesh
Attachments:
ClusterVersionControl_Design.vi ‏17 KB

Any Ideas?

Use of 7024 D switch for designing cluster

There is a request to design a cluster for general purpose
applications. I would like to know if SFS
7024D switches can be used. If so, I would really appreciate it if
you give us a sample design with necessary and/or
optional equipment . I mean exactly
for a 1024-node cluster but this time for HPC cluster.

This URL shoul help you:
http://www.cisco.com/en/US/products/ps6421/products_qanda_item0900aecd804c227d.shtml
http://www.cisco.com/en/US/products/ps6421/products_data_sheet0900aecd804c2272.html

Cluster design

Similar Messages

Maybe you are looking for