Standards bodies are defining the data
collection, which is a good thing and could keep costs down. Discovering
switches, the ports and E-line/E-Lan services configured in the switch
can be made available in an inventory list view. Logical elements like
services, UNIs, endpoints and profiles can also be captured by a
discovery filter. These objects can be seen both under a network
database list view and a Carrier Ethernet physical map. However, scale,
high availability and the integration story is cloudy and can ultimately
drive the costs up. Developing to a complex integration standard costs
money. The end goal really is to enable informed, proactive management
and swift problem resolution that effectively runs their operations.
To overcome the management challenge, we (and others) have
pre-built object models to support standards-based equipment and extend
the object model, which can be mapped to support various equipment.
There can be better control over networks with flow-through
automation, real-time QoS performance and bandwidth monitoring that
accelerates time-to-market and ensures customer Service Level Agreements
(SLAs) via standards.
Performance monitoring and health checking can be real-time or
historical on service and can go down to a port, EVC utilization or
transmission errors, and perform QoS thresholds and KPIs. For fault, you
can use RFC2544FdAlarm and RFC2544JitterAlarm that can are parsed and
correlated into meaningful actionable alarms. Class-of-service flows can
allow for testing of throughput, latency and jitter. The network can be
engineered for different traffic priorities.
Configuration, activation and monitoring of RFC2544 tests as well
as threshold definitions and notification reception can be supported.
Provisioning the Ethernet services and OAM profiles can be accomplished
via a user interface. Logical elements like services, endpoints, UNIs,
NNIs, and ports can be added. Various profiles like bandwidth profile,
performance profile, an RFC2544 profile and CFM profile can be added and
the same can be associated to endpoints of a service.
The scaling challenge is always present and if architected
correctly, management systems can scale to very large sizes. One way to
accommodate scale is to use multi-threading data collection in a
distributed hardware environment or virtual machines. This distributed
data collection can roll up to a centralized backend to handle the
correlation business logic, performance KPIs and reporting across the
network.
High availability can be accomplished by hardening the OS and
providing standby hardware and using database replication techniques (a
topic for a future blog).
Lastly, system integration between management systems and OSS and
BSS systems need not be expensive and standards bodies can tend to go
overboard. Technologies can be accomplished using the cloud model by
publishing an SOAP or REST API and using accepted industry protocols,
which will keep costs down. The technology exists today — use it.
Eric Wegner works for Zoho Corp, http://www.webnms.com
No comments:
Post a Comment