Updated: 6/3/2003; 9:38:07 AM.
John Finnerty's Radio Weblog
        

Saturday, May 17, 2003

This is a multi-part message in MIME format.

------_=_NextPart_001_01C31C73.FE8CE850 Content-Type: multipart/alternative; boundary="----_=_NextPart_002_01C31C73.FE8CE850"

------_=_NextPart_002_01C31C73.FE8CE850 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable

<> <> FW: Info on Network Load Balancing for Magic

-----Original Message----- From: Ken Stainsby To: Finnerty, John Cc: Helly, Dan Sent: 5/9/03 7:24 AM Subject: Info on Network Load Balancing for Magic

Network Load Balancing Alternatives for Magic Load balancing is the process of dividing the amount of work that a computer performs between two or more computers so that more work gets done in less time and all users get served faster. In general, load balancing can be implemented with hardware, software, or a combination of both. There are 3 (three) primary methods of load balancing. Only the third type, Microsoft Network Load Balancing, has been tested with Magic and is therefore recommended for use with Magic. 1. Round Robin DNS is a simple and low-cost scalability and performance solution for enabling a limited form of load balancing for Internet server farms. * Advantage: usually comes free of charge as a standard feature of most popular operating systems, such as Microsoft Windows NT 4.0 with Service Pack 4 and Windows 2000 * Disadvantage: is not a high-availability solution. 2. Load-balancing switches such as Cisco, F5, and Alteon products, redirect TCP/IP requests to multiple servers in a server farm. * Advantage: highly scalable, interoperable solution that is also very reliable. Much better and more scalable than using Round Robin DNS. * Disadvantages: 1-can be quite expensive 2-requires multiple switches to avoid making the switch the single point of failure for the entire Web application. 3. Microsoft Windows 2000 Network Load Balancing (NLB) Advantages: (there are no significant Disadvantages)

is less expensive than a load-balancing switch

avoids having a single point of failure.

is available in:

Windows 2000 Advanced Server

Windows 2000 Datacenter Server

Windows 2000 Server [rather than Advanced Server] as part of Microsoft Application Center 2k

provides high availability and high scalability

utilizes a common virtual IP address

transparently partitions client requests across multiple servers Adding Web servers can provide near-linear scalability (meaning that throughput increases linearly with additional servers) when guidelines for good design are followed. For more information see the Microsoft article Building a Highly Available and Scalable Web Farm, available at the Microsoft website: The near-linear scalability in NLB means that although one machine is capable of handling 300 concurrent users, adding a second machine does not double this number. With additional machines the sum of all concurrent users is reduced by a 25% overhead cost across all machines. Therefore, the total number of concurrent users for two machines is 450 (instead of 600). Although an NLB cluster can include up to 32 machines, Microsoft recommends no more than eight (8). If further capacity is required, configure additional NLB clusters on more subnets. Using NLB clustering necessitates a larger database server than a single-server environment. In addition to the common external IP address for the cluster, each server in the cluster responds to a dedicated network address as well. So, each machine responds to two network addresses: * a cluster network address, and * a dedicated network address. Network Load Balancing is implemented using a network driver that is logically placed between the higher-level protocol TCP/IP and the network adapter of the host. All the cluster hosts receive the incoming traffic. The NLB network driver acts as a filter and allows the host to process only a part of the incoming traffic. The incoming requests are accepted according to the NLB settings for the host. The benefits of deploying NLB include: * Fault tolerance. Network Load Balancing automatically detects a nonfunctional host and can recover from it. In the case of an offline host, the incoming traffic is distributed across all the servers. It requires no additional hardware to avoid single points of failure. * Higher scalability. Network Load Balancing supports clusters of up to 32 computers, although smaller clusters of 16-24 computers usually result in better resource utilization. * Higher performance. Unlike most load balancing switches, the performance of MS Network Load Balancing automatically improves with the speed of the cluster hosts and local area network (LAN), ensuring that the load balancing will never become a bottleneck as LAN and host speeds increase. * Manageability. You can specify load balancing for a single TCP port to customize processing load for a host. You can also block access to certain ports. * Ease of use. MS Network Load Balancing installs a standard networking driver component and does not need any special hardware to run. - NLB distributes incoming TCP traffic using a statistical algorithm based on server-load percentage settings. - NLB provides dynamic scaling by automatically accommodating the addition and removal of cluster members without affecting clients. For more information see the article Load Balancing COM+ Components, available on the Microsoft website: http://www.microsoft.com/technet/treeview/default.asp?url=/technet/prodt echnol/acs/evaluate/LB-Cmpnt.asp Windows Load Balancing Service (WLBS) on Windows NT 4 Server Enterprise Edition has not been tested with Magic, and therefore is not supported. Ken Stainsby Nashco Consulting Ltd. Office: (866) 590-0846 Cell: (403) 804-7828 Fax: (403) 206-7693 Email: ken@nashco.ca Web: www.nashco.ca

------_=_NextPart_002_01C31C73.FE8CE850 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable

xupdate

<< image001.gif >> << image002.gif >>


 FW: Info on Network Load Balancing for Magic

-----Original Message-----


From: Ken Stainsby


To: Finnerty, John


Cc: Helly, Dan


Sent: 5/9/03 7:24 AM


Subject: Info on Network Load Balancing for Magic

Network Load Balancing Alternatives for Magic


Load balancing is the process of dividing the amount of work that a


computer performs between two or more computers so that more work gets


done in less time and all users get served faster. In general, load


balancing can be implemented with hardware, software, or a combination


of both.


There are 3 (three) primary methods of load balancing. Only the third


type, Microsoft Network Load Balancing, has been tested with Magic and


is therefore recommended for use with Magic.


1. Round Robin DNS is a simple and low-cost scalability and performance


solution for enabling a limited form of load balancing for Internet


server farms.


• Advantage: usually comes free of charge as a standard feature of most


popular operating systems, such as Microsoft Windows NT 4.0 with Service


Pack 4 and Windows 2000


• Disadvantage: is not a high-availability solution.


2. Load-balancing switches such as Cisco, F5, and Alteon products,


redirect TCP/IP requests to multiple servers in a server farm.


• Advantage: highly scalable, interoperable solution that is also very


reliable. Much better and more scalable than using Round Robin DNS.


• Disadvantages:


1-can be quite expensive


2-requires multiple switches to avoid making the switch the single point


of failure for the entire Web application.


3. Microsoft Windows 2000 Network Load Balancing (NLB)


Advantages: (there are no significant Disadvantages)

 


is less expensive than a load-balancing switch


avoids having a single point of failure.


 


is available in:

 


Windows 2000 Advanced Server


Windows 2000 Datacenter Server


Windows 2000 Server [rather than Advanced Server] as part of Microsoft


Application Center 2k


 


provides high availability and high scalability


utilizes a common virtual IP address


transparently partitions client requests across multiple servers


Adding Web servers can provide near-linear scalability (meaning that


throughput increases linearly with additional servers) when guidelines


for good design are followed. For more information see the Microsoft


article Building a Highly Available and Scalable Web Farm, available at


the Microsoft website:


<http://msdn.microsoft.com/library/default.asp?url=/library/en-us/dnduwo


n/html/d5nlb.asp>


The near-linear scalability in NLB means that although one machine is


capable of handling 300 concurrent users, adding a second machine does


not double this number. With additional machines the sum of all


concurrent users is reduced by a 25% overhead cost across all machines.


Therefore, the total number of concurrent users for two machines is 450


(instead of 600). Although an NLB cluster can include up to 32 machines,


Microsoft recommends no more than eight (8). If further capacity is


required, configure additional NLB clusters on more subnets. Using NLB


clustering necessitates a larger database server than a single-server


environment.


In addition to the common external IP address for the cluster, each


server in the cluster responds to a dedicated network address as well.


So, each machine responds to two network addresses:


• a cluster network address, and


• a dedicated network address.


Network Load Balancing is implemented using a network driver that is


logically placed between the higher-level protocol TCP/IP and the


network adapter of the host. All the cluster hosts receive the incoming


traffic. The NLB network driver acts as a filter and allows the host to


process only a part of the incoming traffic. The incoming requests are


accepted according to the NLB settings for the host.


The benefits of deploying NLB include:


• Fault tolerance. Network Load Balancing automatically detects a


nonfunctional host and can recover from it. In the case of an offline


host, the incoming traffic is distributed across all the servers. It


requires no additional hardware to avoid single points of failure.


• Higher scalability. Network Load Balancing supports clusters of up to


32 computers, although smaller clusters of 16-24 computers usually


result in better resource utilization.


• Higher performance. Unlike most load balancing switches, the


performance of MS Network Load Balancing automatically improves with the


speed of the cluster hosts and local area network (LAN), ensuring that


the load balancing will never become a bottleneck as LAN and host speeds


increase.


• Manageability. You can specify load balancing for a single TCP port to


customize processing load for a host. You can also block access to


certain ports.


• Ease of use. MS Network Load Balancing installs a standard networking


driver component and does not need any special hardware to run.


- NLB distributes incoming TCP traffic using a statistical algorithm


based on server-load percentage settings.


- NLB provides dynamic scaling by automatically accommodating the


addition and removal of cluster members without affecting clients.


For more information see the article Load Balancing COM+ Components,


available on the Microsoft website:


http://www.microsoft.com/technet/treeview/default.asp?url=/technet/prodt


echnol/acs/evaluate/LB-Cmpnt.asp


<http://www.microsoft.com/technet/treeview/default.asp?url=/technet/prod


technol/acs/evaluate/LB-Cmpnt.asp> 


Windows Load Balancing Service (WLBS) on Windows NT 4 Server Enterprise


Edition has not been tested with Magic, and therefore is not supported.


 


 


Ken Stainsby


Nashco Consulting Ltd.


Office: (866) 590-0846


Cell:    (403) 804-7828


Fax:    (403) 206-7693


Email:  ken@nashco.ca <mailto:ken@nashco.ca>


Web:   www.nashco.ca


<file:///C:Documents%20and%20SettingsKenApplication%20DataMicrosoft


Signatureswww.nashco.ca>


 

------_=_NextPart_002_01C31C73.FE8CE850--

------_=_NextPart_001_01C31C73.FE8CE850 Content-Type: image/gif; name="image001.gif" Content-Transfer-Encoding: base64 Content-ID: Content-Description: image001.gif Content-Location: 1_multipart/2_image001.gif

R0lGODlhDwAPAHcAACH/C01TT0ZGSUNFOS4wDQAAAAFzUkdCAK7OHOkAIf8LTVNPRkZJQ0U5LjAY AAAADG1zT1BNU09GRklDRTkuMBri9SzAACH/C01TT0ZGSUNFOS4wGAAAAAxjbVBQSkNtcDA3MTIC AAAGkmNPuAAh+QQBAAABACwAAAAADwAPAITe3t7AwMClpb21tcbOztbW1tbOzs7W1t69vcbGxs6M jK2UlK2trb3GxtaEhKV7e6WMhKWEe6Wtrca9tc6cnLWUlLWtpb29vc6MhK2UjK0BAgMBAgMBAgMB AgMBAgMBAgMFZGAgjmRpnichWYI0oYITOcrCNGb8yEplGQcSwOF4QDIUwYBQAIyGkNlCuTw4RUOM A6JgDBLNawCwWNQEjIsh/KRUFhQGIgEEiAGFC2OPMDDtQgUDc35sJQUJc0xNKHiFBSiRIiEAOw= ------_=_NextPart_001_01C31C73.FE8CE850 Content-Type: image/gif; name="image002.gif" Content-Transfer-Encoding: base64 Content-ID: Content-Description: image002.gif Content-Location: 1_multipart/3_image002.gif

R0lGODlhDAAMALP/AMDAwMzMmcyZzMyZmcxmmcxmZpmZmZlmmZlmZgAAAAAAAAAAAAAAAAAAAAAA AAAAACH5BAEAAAAALAAAAAAMAAwAQAQ+cJhx5BgCACGwxwEHEMiBFEeFbdRpVkIISCRhqNmmiztg EKkDcJK7AU22ge5XQhpiGYoFp8ltAlWNbpvVRQAAOw= ------_=_NextPart_001_01C31C73.FE8CE850--
8:56:56 AM    comment []


© Copyright 2003 John Finnerty.
 
May 2003
Sun Mon Tue Wed Thu Fri Sat
        1 2 3
4 5 6 7 8 9 10
11 12 13 14 15 16 17
18 19 20 21 22 23 24
25 26 27 28 29 30 31
Apr   Jun


Links

JF Links


Click here to visit the Radio UserLand website.

Subscribe to "John Finnerty's Radio Weblog" in Radio UserLand.

Click to see the XML version of this web page.

Click here to send an email to the editor of this weblog.