Corvil » Questions and Answers
Questions & Answers
View our archive of the most common questions and answers.
Data Sheet
CorvilNet Features and Benefits
Q. What are Service Channels?
A. A Service Channel represents a monitored set of traffic flowing across the network between points where CorvilNet appliances are deployed. A single Service Channel can be used to match all of the traffic present or multiple (potentially overlapping) Channels can be defined to represent different traffic subsets. Service Channels provide two main capabilities in CorvilNet. Firstly, they measure the point-to-point latency, jitter and loss experienced by the traffic within each Channel. Secondly, the measured performance and traffic rate are continuously compared against user-specified targets, generating alerts and triggering Smart Packet Capture whenever violations are detected.
Q. How does Service Channel performance measurement operate?
A. Service Channel measurement is passive in the sense that it does not inject instrumented packets into the network. Instead it watches the flow of application traffic at both ends and returns time stamps and packet signatures to one end to allow the latency, jitter and loss to be calculated. The latency and loss of each and every packet between two points can be measured – not just a small set of 'probe' packets. Results are measured and reported independently for each direction within each Service Channel.
Q. What is the difference between Service Channel measurement and ICMP Ping?
A. Performance measurement using Service Channels differs from ICMP Ping in several important respects:
- Service Channels measure the performance of the actual application packets. ICMP Ping only measures the performance of ping packets.
- One-way latency and loss are resolved in each direction independently.
- Service Channels report the latencies experienced by all network traffic instead of just the small fraction represented by ICMP ping packets. In other words, the Channel will detect if even a single packet is lost or excessively delayed.
- In a multi-class network, Service Channels can be defined to report the different latency and loss characteristics of each class of traffic independently, whereas ICMP traffic will only probe one class.
Q. Do Service Channels require GPS synchronisation of the CorvilNet appliances?
A. GPS is supported but is not required. GPS hardware can be useful when the propagation delay (that is, the minimum delay) between two sites is asymmetric. Without GPS, CorvilNet assumes equal propagation delay in each direction. Measurement of the end to end loss and variation in latency (jitter) is virtually unaffected by the presence of GPS synchronisation.
Q. What is the network impact of Service Channels?
A. CorvilNet batches time stamp reports together, and the overhead is approximately 12 bytes per measured packet. For a typical mixture of packets sizes (IMIX) this translates into a network overhead of about 4% when measuring all packets. Service Channels can be configured to measure less than 100% of packets, generating controllable and arbitrarily small network overheads.
Q. Do Service Channels require packets or packet headers to remain unmodified at different network points?
A. CorvilNet uses packet signatures to uniquely identify packets as they travel along a Service Channel. The signatures can be computed from any part of the packet that remains unmodified during its journey. The user can specify for each Service Channel which fields or sets of fields in the packet header or payload should be used for this purpose. For example, a financial transaction may require multiple processing stages in a data centre, each of which modifies packet headers. However the transaction ID is not modified and can be used by CorvilNet to track the progress of the packets within a Service Channel.
Q. Are Service Channels compatible with NAT?
A. NAT modifies source and/or destination addresses within packet headers. However CorvilNet does not have to rely on these fields to determine packet signatures for Service Channels. Therefore Service Channels are fully compatible with NAT.
Q. What is Congestion-Aware Monitoring?
A. CorvilNet provides Congestion-Aware Monitoring by calculating the amount of queuing delay and loss that measured traffic will induce at network bottleneck points. This feature takes account of the way bandwidth and resources are shared out between different traffic classes in a multi-class network. Congestion-Aware Monitoring can be used in conjunction with Service Channel measurement to determine how much of the measured point-to-point latency is due to congestion. It can also be used to monitor remote bottleneck interfaces where no CorvilNet Engine has been deployed.
Q. What is the Microburst Detection feature?
A. Microburst detection uses a unique algorithm to calculate traffic activity over very short time scales, with an adjustable duration of 1 millisecond to 1 second. This algorithm measures every packet to report microbursts on Service Channels and on monitored network interfaces or traffic classes.
Q. What is the Network Service Index?
A. Network Service Index is a highly summarised figure-of-merit representing the service level being achieved for each Service Channel or interface in the system. The index vale measures how close the Service Channel is to its defined service level targets, and indicates whether the targets have been breached. This feature enables network managers to quickly review the status of hundreds of Service Channels and to rank them according to problem severity.
Q. What is Corvil Bandwidth?
A. CorvilNet calculates directly the amount of the bandwidth that should be allocated to measured traffic in order to maintain a user-specified latency and/or packet-loss target. This value is called Corvil Bandwidth. It can be computed either per network interface or per traffic class in a multi-class network.
Q. What is Smart Packet Capture?
A. Smart Packet Capture is a triggered packet-capture facility which records network traffic during anomalous events together with performance measurements and analytics that help to diagnose and resolve problems. CorvilNet allows any of its comprehensive performance and bit-rate measurements to act as threshold triggers for Smart Packet Capture. When any threshold is violated an event trace is recorded to disk which includes the 10 seconds of traffic preceding the event, all traffic during the event, and 10 seconds of traffic after the event is cleared. Per-packet latency, jitter, packet-loss and microburst bit-rates are saved alongside the captured packets, as well as the recognised application IDs. The user can later analyse the captured data using a drill-down GUI which presents results in graphical form down to the level of per-packet granularity.
Q. Can packet captures be exported for analysis by other tools?
A. Yes. CorvilNet will export packet captures in standard PCAP format.
Q. How much packet data is captured by Smart Packet Capture?
A. Smart Packet Capture in its default configuration captures minimal packet header and no packet payload. The user can configure the SNAP length either system-wide or per Service Channel (or monitored interface) to enable up to full payload and header capture. This can be useful when combined with the ability to export the data to external applications.
Q. What is TCP Micro-Analytics?
A. TCP Micro-Analytics combines CorvilNet's Rapid Traffic Sequencing technology and Smart Packet Capture with TCP quality metrics to rapidly identify and trouble-shoot TCP related network quality issues. Available TCP measurements include:
- Application Response Time – The delay experienced beginning with a client's request and ending upon receipt of a response. This corresponds to an end user's view of the time it takes for a response after initiating an action
- Network Roundtrip Time – This measurement isolates the roundtrip network delay between the client and server
- Out-of-sequence/Retransmitted Packets – This measurement provides a count of all out-of-sequence packets in a TCP session due to re-ordering or re-transmission
- Goodput/Throughput – Goodput measures the data successfully delivered within a TCP session minus all packet overheads (including TCP headers) and re-transmitted packets. Throughput provides the measure of all data transmitted in the TCP session including TCP headers and re-transmitted packets
- Concurrent Connections – Measures the number of concurrent, successful TCP connections.
Q. What is Live View?
A. The Live View capability in CorvilNet provides immediate access to the measured data and analytics in the form of animated graphical charts which are updated once per second. Results presented include microburst and average bit-rates, distribution of measured latency and jitter, packet-loss rate, Corvil Bandwidth, and TCP micro-analytics. Live View allows you to view traffic and network behaviour in real-time and to spot the impact of changes immediately.
Q. How does CorvilNet recognize applications?
A. CorvilNet includes an application recognition engine based on layer-7 deep packet inspection. CorvilNet provides tables of top applications for each monitored Service Channel, interface, or traffic class, as well as top talkers, listeners, and conversations. CorvilNet will recognise more than 350 common applications out-of-the-box, and allows users to easily add support for custom applications.
Q. Does CorvilNet store historical performance measurements?
A. Yes. CorvilNet stores all measured data with 5-minute granularity. Onboard storage on each appliance has capacity for at least 60 days of storage, after which data will optionally be transferred automatically to an external database. The stored data can be viewed with 5-minute granularity over any timeframe during the past year. CorvilNet also stores Smart Packet Captures with per-packet granularity for up to 1 year.