This soft copy for use by IBM employees only.

Next, we can use microsegmentation to reduce the number of workstations that have to share the token-rings 16 Mbps of bandwidth. The segment size is limited only by the granularity of the token-ring concentrators you are using. If you are using a port-switching hub, such as the IBM 8260 Nways Multiprotocol Switching Hub, you can make your segments any size from two devices up to the architectural limit of the token-ring. And for those users who need the most performance that token-rings can provide, simply connect them directly to a port on the LAN switch module. This will provide them with up to 32 Mbps of dedicated bandwidth. In this case, since we have already used up 3 of our 24 available token-ring ports for the servers, we can have as many as 21 segments switched by the LAN Switching module.

Finally, to further reduce the number of workstations sharing token-ring bandwidth, you can begin to move some of your power users directly to the ATM network by installing ATM25 adapter cards, such as the IBM Turboways 25 adapter, in their workstations, running the simple migration utility, and connecting them to an ATM 25 Mbps port on the IBM 8285 Base Unit. They now have dedicated bandwidth of 25 Mbps to their desktop and access to all of the resources of the ATM network, while still retaining access to the shared media resources, such as the full-duplex token-ring servers. We do that with 12 of our workstations, reducing the number of workstations on shared token ring segments to 68.

Table 10 illustrates the before and after effects of our changes on the available bandwidth.

Table 10. Bandwidth Improvement with Token-Ring LAN Switch Module

Device

Maximum Bandwidth Available

Bandwidth

 

 

After...(Kbps)

 

Improvement

 

 

 

 

 

Ratio￿2￿

 

Server

ATM

Micro-

ATM25

 

 

 

Offload

Server

segment

Offload

 

 

 

 

-ation￿1￿

 

 

 

 

 

 

 

 

Primary Server

32000

155000

155000

155000

811x

 

 

 

 

 

 

Secondary Server

32000

32000

32000

32000

167x

 

 

 

 

 

 

Token-Ring

200

200

4000

4942￿3￿

25x

Desktop

 

 

 

 

 

 

 

 

 

 

 

ATM Desktop

200

200

4000

25000

130x

 

 

 

 

 

 

Note:

￿1￿In this case assuming an even distribution of four workstations per segment on each of 21 available token-ring segments.

￿2￿Calculated by dividing the total bandwidth available after the change by the original bandwidth available per device, which was calculated by dividing the total bandwidth available by the number of devices sharing it. In this case, the original bandwidth calculation is: 16,000 Kbps/segment * 1 segment / (80 users + 4 servers) = 191 Kbps/device.

￿3.￿This is an average value. The 12 segments with three devices/segment would actually receive 5,333 Kbps while the remaining nine segments would have four devices apiece, each receiving 4,000 Kbps.

78ATM Workgroup Solutions: Implementing the 8285 ATM Switch

Page 94
Image 94
IBM SG24-4817-00 manual Device Maximum Bandwidth Available After...Kbps, Server, Micro, Segment Offload