Want to stay in the loop...?
News updates:
REANNZ will never disclose your email address.
Technical FAQs
What will it cost?
How fast will it go?
What can I use KAREN for?
Who can provide access?
Is VOIP supported?
What is a LFN?
What is path MTU discovery?
What is the difference between a POP and an AAP?
Why do I need an ASN?
Can I access the internet?
What is DWDM?
What is jumbo frame support?
What is MBGP?
What is MSDP?
What is multicast?
What is PIM-SM?
Why is quality of service not offered?
What will it cost?
The answer to this question will differ for each organisation. Connections to KAREN are available to members, associate members and partners as defined in the Network Access Policy (NAP).
Annual costs
There is an annual membership fee that is payable for access the network. The membership fee is a fixed cost, and is calculated using a variety of metrics related to the size and classification of the member. For further information on the cost of membership please contact REANNZ.
Member infrastructure costs
An appropriate network device is also necessary. It needs to support a number of connection standards, and cost will vary depending on the selection of vendor, features and general capability.
Unless organisations are located within the KAREN POP they will require an access circuit which describes the path between the organisation's edge device and the KAREN POP. The access circuit could be optical, electrical or wireless. The cost of this service will vary based on distance, complexity and a number of other geographic factors.
One-off connection costs
There are some core costs associated with connectivity that are based on the costs incurred in making the connection. These are one off costs which are passed on to the connecting organisations.
Organisations will need to cover the cost of a GBIC per physical connection. The type of GBIC required is dependant on the distance from the POP to the members terminating equipment.
LX GBIC (10km or less) $990
ZX GBIC (between 10km and 80km) $4175
Costs are based on existing stock.
Members will also need to cover the costs of making the connection logically and physically. Where multiple sites are being connected simultaneously the REANNZ labour charge may be discounted.
REANNZ Labour Charge $2000
Telstra Clear Labour Charge $1200
How fast will it go?
The KAREN core network is capable of transmission speeds of 10Gbps with very low latency and almost no jitter (variance in delay). Most operating systems and infrastructure will need to be tuned to take full advantage of the available resources.
While the network itself is a high performance network, it is not simply a case of plug and play. Using the worst case RTT measured during testing, and assuming no packet loss or bottlenecks, 6Mbps seems to be the maximum data transfer rate for a Windows 2000 or Windows XP host. This transfer rate is limited by the default TCP/IP settings chosen by Microsoft, there are a variety of approaches to increasing the throughput of systems that are explored in tuning for speed.
What can I use KAREN for?
At the most basic level KAREN is a network offering domestic layer 2 connectivity based on 802.1Q VLAN tagging, and domestic and international layer 3 connectivity based on IPv4 and IPv6 protocols (including multicast).
Any application that does not contravene the acceptable use policy and can be supported by the standards above should be able to be used over KAREN. However, it should be noted that the service level agreements and hours of support for the KAREN network may not be sufficient for business critical applications in some cases.
Some sites that may be of interest once you have access to KAREN are:
IPv4
AARNet is the Australian equivalent of KAREN. The AARNet mirror provides a variety of content that can be downloaded at comparatively high speeds. A 500MB ISO image should take approximately 5-7 minutes to download compared to 3-4 hours over the commodity internet.
IPv6
This content is coming soon.
Multicast
The Research Channel provides a great deal of multimedia content and includes a multicast stream for networks such as KAREN. To view the multimedia stream you will need to have VideoLAN Player (VLC) and enter the details provided on the Research Channel website.
Who can provide access?
REANNZ is strictly neutral about the choice members make for their last mile(s) provider. Any entity may provide last mile services providing any connection to KAREN is technically and commercially compliant.
The list below is of companies known to us who have supplied last mile services to KAREN members: there may other providers able to deliver theses services. Please note that some of these providers work in particular regions only.
TelstraClear
http://www.telstraclear.co.nz/
Contact: Jo Busche
CityLink
http://www.citylink.co.nz/
Contact: Neil De Wit
Smartlinx 3
http://www.smartlinx3.co.nz/
Contact: David Haynes
Network Tasman
http://www.networktasman.co.nz/
Contact: Des Salmon
InspireNet
http://www.inspire.net.nz/
Contact: James Watt
FX Networks
http://www.fx.net.nz/
Contact: Jamie Baddeley
Christchurch City Holdings
http://www.cchl.co.nz/
Contact: Bill Luff
Canterbury Fibre Networks
http://www.cfnl.co.nz/
Contact: Darryl Swan
Connector Systems
http://www.connectorsystems.co.nz/
Vector Communications Limited
http://www.vectorcommunications.co.nz/
Contact: Greg Davis
Telecom NZ
http://www.telecom.co.nz/
Velocity Networks Performance Broadband
http://www.velocitynetworks.co.nz/
Contact: Shane Hobson
Is VOIP supported?
There is nothing to prohibit the use of voice over IP (VOIP) between sites over KAREN. A high speed, high bandwidth network with low latency and jitter supporting connectivity at layer 2 and layer 3 will inherently support VOIP.
However, the service management structure that is in place was not designed with 24x365 business critical applications in mind. Support hours are between 0500 and 2100 hours Monday to Friday and service level targets may not align with the requirements of individual businesses.
KAREN does not offer quality of service (QoS) configuration in the core. QoS is often associated with VOIP services but QoS is not relevant for KAREN.
What is a LFN?
LFN is an acronym for Long Fat Network which is a term used to describe a network with a large bandwidth delay product as defined by RFC 1072. Bandwidth delay product (BDP) is the networks capacity multiplied by the expected latency and is effectively the equivalent of the amount of data on the light path (or wire) at any point in time.
BDP = bandwidth x delay product
In the case of the KAREN core the result is 2.38 Megabytes.
A BDP calculator is available at http://www.speedguide.net/bdp.php
This becomes important because protocols like TCP wait for an acknowledgement message before sending more data. The majority of operating systems will not make use of the bandwidth available on the KAREN backbone without being tuned to take advantage of the large volume of data that can be sent without waiting for an acknowledgement message.
FX Networks and the WAND Group have created a calculator that will provide information about the results that are possible when tuning operating systems for LFN’s.
http://noc.fx.net.nz/calculator/
What is path MTU discovery?
Path MTU Discovery (RFC 1191) utilises the don’t fragment (DF) bit in the TCP header to discover the maximum MTU that can be used in end to end communications.
A brief overview of the solution is that the sending device assumes that the maximum transmission value for the communication is the MTU of the next hop device and sets the DF bit on the transmission preventing the packet from being fragmented. Any device that cannot support the MTU that has been used responds with an ICMP type 3 message (datagram too big) which causes the source to reduce the assumed PMTU value for the communication.
The RFC provides more detail about how Path MTU Discovery should work. The important point to note is that devices that do not support RFC 1191 create black holes. A black hole describes a situation where data is sent to a device which discards the data without responding to the sender.
All network devices involved in end-to-end communications using KAREN should support RFC 1191.
What is the difference between a POP and an AAP?
The quick answer is nothing. The term POP (Point of Presence) and AAP (Access Aggregation Point) were used in the original RFP to differentiate between access points that were built on the DWDM core from sites that were located on spurs, and therefore not on the DWDM core. However, in the network as implemented, there is no difference between sites that may have been referred to as POP’s or AAP’s.
There are several satellite sites that are not part of the core backbone of KAREN. These sites are located at Invermay, Lincoln, Havelock North, Mount Albert and the North Shore. These locations are connected to the core at 10Gbps providing the same services as any other site. The only difference is that they rely on the upstream switch to be available to reach the core network.
Why do I need an ASN?
An ASN is used to uniquely identify the source of an IP prefix (network address) within the Border Gateway Protocol (BGP) routing domain. This allows the research and education (R&E) community to be assured that the routes they receive and the organisations they exchange data with are legitimate users of the R&E or advance networks.
More detailed information can be found by reviewing RFC 1930 which describes the appropriate use of ASN’s.
Applications can be made through APNIC in the Asia/Pacific Region. APNIC also provide a useful tool for determining the organisation to which an ASN is assigned.
Can I access the internet?
KAREN provides over 9000 routes to research and education networks and advanced networks only. The network access policy specifically prohibits the transit of traffic between commercial networks.
This group of routes should be considered as being completely separate from the commodity internet routes that are available through a commercial internet service provider (ISP). It is true that some of the routes exist in both routing tables, however this does not mean the networks are logically or physically connected.
All organisations that are connected to KAREN should maintain a separate commodity internet service if they require access to that group of routes. They should also take all steps necessary to ensure that the two networks remain logically and physically separated.
What is DWDM?
DWDM is an acronym for Dense Wave Division Multiplexing, which is a technology that relates to optical networks. In order to understand DWDM it is necessary to understand the concept of multiplexing.
Multiplexing involves the transmission of multiple discrete signals, or streams of information, over the same physical media as a single signal. This single signal is then separated on receipt so that the multiple discrete signals can continue to their destination.
One of the first methods of multiplexing was used in analog transmission and involved the use of separate frequencies to send simultaneous transmissions. This technology was referred to as frequency division multiplexing (FDM). Digital transmissions often used time division multiplexing (TDM) to split the available bandwidth into time slots which are then used to multiplex multiple signals, ISDN is an example of technology that uses TDM.
Figure 1 - Time division multiplexing
New technologies have continued to use division multiplexing to make more efficient use of available bandwidth and logical and physical network paths. In optical networks wave division multiplexing (WDM) allows more than two wavelengths (or signals) to be used on the same piece of fiber optic cable.
Dense wave division multiplexing (DWDM) provides the capability to use a far greater number of wavelengths on the same piece of fiber optic cable. By extending the spectrum of light that can be used to carry data there is the potential for more than 80 separate signals to be carried on a single fiber optic cable.
What is jumbo frame support?
A jumbo frame is a term used to describe a frame that is larger than the traditional Ethernet frame size (1518 bytes). This size was selected when high error rates and comparatively low transmission speeds characterised network communications.
Larger bandwidth, more reliable transmission and larger data volumes led to the consideration of larger frame sizes. A larger frame size will allow higher throughput and reduce the demand on resources (router CPU, memory etc.) involved in the transmission and receipt of data.
The 9000 byte IP MTU that is used within the KAREN network is aligned to the MTU that was selected for use across the Abilene core by Internet2. KAREN peers with Abilene and receives a significant number of research and education routes from this network.
For the 9000 byte IP MTU to be effective it must be enabled on every device that is involved in communications between peer systems. It is important to ensure that network interface cards, switches, routers and firewalls all support a 9000 byte MTU.
There are many good explanations about jumbo MTU throughput such as:
- Phil Dykstra Gigabit Ethernet Jumbo Frames – and why you should care
- Alteon Networks Jumbo Frames. Yes!
- Joe St Sauver Practical Issues Associated with 9K MTU’s
What is MBGP?
Multiprotocol BGP is defined in RFC 2858 as an extension to the BGP unicast routing protocol that allows different types of addresses (i.e. multicast addresses) to be distributed in parallel with BGP unicast routes.
The effect is that routers that support multiprotocol BGP will support two separate routing topologies for unicast and multicast routes. The multicast routing topology provides information about the multicast routes, but further action is required in order to make use of those routes. Protocol independent multicast (PIM) is used to build the distribution tree, based on the multicast routes received using multiprotocol BGP, which enables multicast content to be forwarded.
Multiprotocol BGP should be supported by edge devices that are connected to KAREN, where organisations wish to use multicast services.
What is MSDP?
Multicast Source Discovery Protocol (MSDP) provides for the interconnection of multiple PIM-SM domains, allowing for rendezvous point (RP) redundancy and inter-domain multicast.
The use of MSDP to provide inter-domain multicast is a critical point for organisations wishing to separate their multicast cloud from KAREN.
What is multicast?
Multicast is a means of delivering IP datagrams to interested receivers using the strategy of sending data over each network link only once, creating copies only where the destinations split. Most data communications are based on either unicast (sending to a single destination) or broadcast (sending to all destinations). Multicast allows for the sending of data to a group of destinations that register their interest in the data.
Figure 1: Unicast vs multicast traffic
In order for multicast to work effectively it is necessary to address both multicast routing, and multicast forwarding. Multicast routing provides the routing information required to find multicast sources while multicast forwarding deals with the way the requested information reaches the end user.
What is PIM-SM?
Protocol Independent Multicast (PIM) is a multicast routing protocol that is used to build distribution tree’s for multicast data flow. It does not include a topology discovery mechanism, relying on routing information from other sources such as MBGP. There are four variants of PIM:
- PIM Sparse Mode (PIM-SM), defined in RFC 4601 builds unidirectional shared tree’s rooted at rendezvous points (RP’s). Scales well and designed for use across wide areas
- PIM Dense Mode (PIM-DM), defined in RFC 3973 floods multicast traffic and then prunes where instructed to do so. Does not scale very well
- Bidirectional PIM, builds bi-directional trees but does not focus on the shortest path and is therefore not the most efficient
- PIM Source Specific Multicast (PIM-SSM), defined in RFC 3569, builds trees that are rooted at just one source. PIM-SSM is limited to certain applications
PIM-SM has the widest deployment and is the protocol that is in use within KAREN. PIM-SM is built on the assumption that few hosts will subscribe to any multicast session and rather than flooding the network with traffic builds specific trees that are used to deliver multicast traffic.
PIM-SM is a stateful protocol and requires PIM-SM enabled routers to keep track of other routers and forwarding trees. Instead of requiring each router to keep track of stateful information a rendezvous point (RP) is elected and used to centralise the functions. The RP becomes the root of the distribution tree for that particular multicast tree.
PIM-SM must be supported on edge devices connected to KAREN where organisations wish to use multicast services.
Why is quality of service not offered?
Quality of Service (QoS) is a control mechanism that allows for different priorities to be assigned to pre-determined flows of data.
QoS is very important where network capacity is limited. It provides a guarantee that data flows that are defined as critical will receive priority increasing the likelihood that the application associated with that data flow will continue to function when the network is stressed.
KAREN has been proven to retain the characteristics of low latency and jitter for all traffic at close to the 10Gbps theoretical maximum capacity of the network. Until the point at which the network utilisation approaches 10Gbps QoS is not a relevant technology for KAREN.