-
Why is the proposed service believed to apply to the case at hand (what is the purpose for wanting to add the service, what are the user issues, what deficiency is thought to exist)?
The Classroom and Lab Computing (CLC) group within Teaching and Learning with Technology (TLT) has a requirement to periodically re-image all of the classroom and lab computers. So that this does not interfere with the operation of the classrooms and labs, CLC performs this operation while the students are away during session break. As the university has moved to increase the utilization of its on campus facilities, it has continued to shorten the time when there are no students on campus. Currently, the time is short enough that the gigabit link provided by the Integrated Backbone (IB) to the CLC central servers does not have the capacity to transfer the amount of data needed in the time available.
While we do provide a 10 Gigabit Ethernet service that provides more capacity, the technology involved would require a prohibitively expensive wholesale infrastructure replacement for CLC in order to solve a problem that only occurs for about a week a few times a year.
By providing a backbone service that uses IEEE 802.3ad Link Aggregation Control Protocol (LACP) we can bind together either two or three standard gigabit interfaces into a single logical channel providing additional capacity using the existing CLC infrastructure.
-
Would this service apply or be beneficial to others at UP?
This service would be beneficial to any organization within Penn State with bandwidth requirements that exceed 1 Gbps but do not merit the cost of a ten-fold increase.
-
Would this service apply to all University locations?
This service would apply to all University locations where we currently provide gigabit Ethernet backbone service. We will not provide this service for connections to the redundant router.
-
What are the benefits to this service over existing offerings, or does this service provide something completely new (from a global perspective, not a single case perspective)?
This service provides an incremental bandwidth improvement over the existing choices of 1 Gbps or 10 Gbps. The technology required to support 10 Gigabit Ethernet is relatively new and expensive. The technology required to support link aggregation is well established and is likely already available in the currently deployed infrastructure.
-
From a strategic standpoint, what are the long-term benefits/drawbacks of providing the service?
The strategic benefits of providing this service are mainly in the area of improved resource utilization. At first it might seem that this proposal would consume additional router interfaces. On closer inspection we would find that any customer in this situation probably already has multiple backbone interfaces in an effort to distribute the load of their application, typically with reduced efficiency. Binding those interfaces using LACP allows us to get greater application efficiency using the same physical resources.
The situation is reversed when moving the customer to 10 Gigabit Ethernet. Currently, router blades supporting 10 Gigabit Ethernet are expensive and low density. Each one consumes a chassis slot that could be used for other, higher density applications. While the 10 Gigabit Ethernet service is valuable for those that have a genuine need, we should preserve the scarce resource for them, and not use it for applications that only occasionally need more than one gigabit.
-
Can the service be provided by another University group or by an outside third party?
Only TNS can provide this service since it involves the method for connecting to the Integrated Backbone.
-
What options does the user/customer have if TNS does not provide the service?
They can impact the use of their service, or they can undertake to re-architect their infrastructure and find corresponding budget resources.
-
What are the proposed one time, recurring and usage rates for the service, and how do they compare to other available options (assuming there are any)?
A customer that currently has the Gigabit Ethernet Integrated Backbone Service pays a one-time fee of $1,000 and no monthly fee.
If their bandwidth requirements exceed one gigabit, they can upgrade to our 5 Gigabit Ethernet Integrated Backbone Service — using a 10-gigabit physical Ethernet that we throttle to 5 gigabits. In this case, the customer pays TNS a one-time fee of $6,000 and a monthly fee of $375.
There are also infrastructure costs involved, as equipment to support 10-gigabit Ethernet is relatively new. In the case of CLC, they currently use HP 2848 switches to terminate the backbones in question. The equivalent switch with 10-gigabit capability is the HP 3500yl-48. The CDW cost for this switch, equipped to accept 10-gigabit Ethernet is $8,718.98.
So, in this example, the cost difference for CLC to move from an existing gigabit to a new 5-gigabit service is on the order of $15,000 one-time and $375 per month.
On the other hand, using LACP, the existing switches can already accept additional gigabit Ethernet connections bound to form a single backbone connection. No more hardware is necessary than if the customer had simply purchased a second backbone connection — either for the customer or TNS.
IPv4 space is actually conserved over simply buying a second backbone connection, since in the LACP case we only assign one subnet and the protocol takes care of distributing the traffic between the links.
There may be an argument that the Network Operations Center may need training in order to respond to customer issues regarding LACP backbone connections, simply because they are different. However, the LACP protocol is well established and provides a level of resiliency not available with a single connection.
We would propose $1,000 per backbone router interface one-time, and $75 for equipment certification (performed by the customer with TNS supervision), plus $50 per month to cover the need for ongoing TNS training for the NOC and test engineers (see 10, below).
-
What are the operational support issues — do spares need to be carried and how many are suggested, does it use proprietary equipment, what are the delivery (lead) time issues?
No spares are required as no new, unique, or proprietary equipment forms a part of this service.
-
What are the training issues for operations, ITS user training staff and user training
The NOC will need training in order to properly diagnose customer issues. Within MRTG, each physical interface appears, as it normally would. In addition, a new virtual interface representing all of the traffic on the bound interfaces appears.
Also, since this service requires more customer involvement in terms of configuration required beyond the service demark, we would suggest a “certification lab” be established where a customer could bring their device in order to ensure interoperability and proper configuration of their device before installation. This would consist of a few interfaces on a test router with canned interface configurations and equivalent test data and expected results on the network data generator/analyzer (SmartBits) along with a written procedure to follow. The customer would be expected to perform the certification under the supervision of TNS personnel. Training for those supervising would be necessary.
-
What are the start-up costs (including operational support and training costs) for TNS
NSG and NOC training as well as the creation of the certification lab and training of its support personnel are necessary at start-up.
-
What are the technical issues — will technology changes make the service unnecessary in the future, is it a mature technology or prototype, has it been proven (tested) to work in the applications suggested (and what was the process for testing), what are the expectations regarding reliability, can it be applied to the existing infrastructure, etc.?
This service requires additional configuration capability on the part of the customer. The requirement for the customer to pass equipment certification addresses this.
Eventually, gigabit Ethernet interfaces will fall out of fashion, replaced by ubiquitous 10 Gigabit Ethernet and TNS will end support for the service. Before we reach that point, we will encounter customers who need an incremental improvement over the 10 Gigabit service and we will extend this one to use the new technology.
There will be a need for this service as long as our customers are using Ethernet and some of them require incremental increases in bandwidth rather than exponential ones.
TNS already has undertaken to convert from SONET to WAN Ethernet from our inter-campus network provider in part because it offers smaller bandwidth increments that offer more attractive pricing.
LACP is a mature technology. It was defined in the IEEE 802.3 Ethernet standard as IEEE 802.3ad in March 2000. TNS has experience with similar technology that provided 2 Gigabit trunks for the core network at Penn State and for combining multiple T1 circuits for off campus facilities.
Once properly configured, this service should be more resilient than alternative methods, since the remaining link continues to carry all traffic should the other link fail.
This service will work well with the existing infrastructure.
-
Will new equipment be required?
No new equipment is required.
-
If applicable, what exactly is the proposed equipment (manufacturer and model numbers)?
N/A.
-
If applicable, what is the expected life cycle of the equipment (is it stable or will it require frequent change-outs until the technology stabilizes)?
N/A.
-
What is the projected start date for this service?
In order to satisfy the needs of the trial customer, we would like to have the trial in place by August 2, 2007.