ThousandEyes
for Enterprise and IT Operations
Performance
Challenges in Enterprise Environments
The performance of a corporate network is highly dependent on
different factors. Applications like VDI and VoIP are specially sensitive to
the real-time state of the network.
Infrastructure/Network
Performance:
Latency: Defined
as the round-trip time (RTT) between the client and server, network latency
depends mostly on the physical distance between the endpoints, as well as the
congestion state of the network. Higher latencies cause higher application
response times, as well as lower TCP (Transport Control Protocol) throughputs.
TCP throughput is actually inversely proportional to the round trip time
(WindowSize/RTT). Also, because of the slow start mechanism, TCP can be quite
inefficient in taking advantage of available bandwidth. The impact of the RTT
is exacerbated in the case of chatty protocols, that require N rounds to
complete a transaction. A transaction that would take Nx5ms inside a LAN can
take Nx100ms over the public Internet, making the application virtually
unusable.
Packet Loss: Packet
loss can trigger TCP retransmissions and bring the throughput of the connection
down. The throughput of a TCP connection is roughly proportional to sqrt(p)
where p is the packet loss probability. So a connection with 2% loss has about
70% of throughput of a 1% loss connection. In fact it can be shown that in
steady state, the window size of TCP with 1% loss is less than 10 packets. If
each packet is 1,500 bytes long and the RTT is 100ms, that’s a throughput of
150 Kbytes/s.
Capacity and Available Bandwidth: Capacity
between two endpoints is the maximum data rate that can be achieved in the
absence of any cross-traffic. Cross traffic will use a portion of the capacity,
and the remaining is the available bandwidth, which determines how fast TCP
connections can go. Available bandwidth in enterprise networks is typically
limited by cost, not so much by technology. Currently a WAN MPLS access can
cost anywhere from $300/Mbps/month to $1,000/Mbps/month (note: Access to
broadband Internet is 30x less expensive in comparison). Most enterprises still
route Internet traffic through their data centers (e.g. MPLS access) using
Internet traffic backhauling because of security and control. This consumes
expensive bandwidth in the corporate backbone.
Routing Availability: Routing
inside the enterprise is determined by interior routing protocols such as OSPF
or IGRP. But for Internet traffic, different networks need to exchange routing
information using the Border Gateway Protocol (BGP). BGP allows independent
neighboring networks to talk to each other to decide what traffic they will
exchange. This process is usually driven by economics rather than network
efficiency, therefore routes are not always optimal from a latency/bandwidth
point of view. BGP changes (e.g. misconfigurations) can render an entire
network unreachable or induce severe performance degradation on applications
(note: bgp convergence times are often in the order of 3-digit seconds.)
Wireless Access (WLAN): Bad
quality 802.11 wireless access in the branch office is a common cause for
performance degradation of applications. This can be caused by degradation of
the radio signal power (lower RSSI) that can be caused by interference or
physical distance between the client and the access point. This is often a hard
element to troubleshoot end-to-end since typically there is no access to this
information without instrumenting the client or the access point.
As enterprises adopt Software-as-a-Service (SaaS),
application performance becomes more dependent on what happens outside of the
corporate network.
ThousandEyes
for Enterprise and IT Operations
Application
Performance:
SaaS/Cloud Adoption: As
enterprises adopt Infrastructure-as-a-Service (IaaS) and boost their SaaS
consumption, application performance does not depend only on what happens
inside the corporate network anymore. Network traffic has to cross multiple
third party networks in the public Internet before it gets delivered to the
end-user. Most of these hops are outside of the control of both the consumer
and the provider (the endpoints) of the application. Traditional enterprise WAN
architectures that were designed in the days when most traffic was staying
inside the enterprise falls short when increasing volumes of traffic go through
the public Internet, mainly because of Internet traffic backhauling and limited
bandwidth in the corporate network.
VoIP/Video: Most
VoIP protocols (including RTP) use UDP (instead of TCP), and voice applications
need to use loss concealment algorithms to cope with packet loss. Voice
applications are also sensitive to jitter (variations in network delay). Each
end of the conversation needs to have buffers to accommodate variations of
delay in the network. Beyond some point, packets that suffer a very high delay
are discarded by the application. These drops can be perceptible in the audio
stream. VoIP traffic can be prioritized using different PHB strategies (Per Hop
Behavior) with different DSCP codes(e.g. EF or AF).
Desktop Virtualization (VDI): VDI
protocols such as ICA or RDP run on top of TCP which means that excessive
packet loss and retransmissions can render VDI unusable. VDI is very sensitive
to network latency since the application needs to react quickly to user
interactions. Another limiting factor is available bandwidth, especially
server-side video streaming, but even in non-streaming scenarios (e.g.
printing), a >100kbps requirement per client is reasonable.
ThousandEyes
at Work
ThousandEyes provides end-to-end visibility of application
delivery in the cloud through a combination of lightweight agents and passive
analysis. ThousandEyes X-Layer technology provides deep visibility into each
layer of application delivery and a connecting thread between these layers,
making it possible to jump from layer to layer to find the root cause of
problems. Each layer maps to a different subsystem of application delivery with
its own set of unique metrics.
The following elements are unique to ThousandEyes:
X-Layer: Provides
deep visibility into each layer of application delivery and a connecting thread
between these layers, making it possible to navigate from layer to layer to
find the root cause of problems.
Deep Path Analysis: Microscopic
view of the end-to-end path between the client and server, including
localization of loss per interface, link delays, route changes and measurement
of capacity and available bandwidth.
Interactive Sharing: Enables
enterprises and application providers to share live data with one another,
eliminating ad-hoc tests, reducing manual processes and resolving problems
faster.
Using
ThousandEyes inside the Enterprise
ThousandEyes can be deployed inside the Enterprise through
Private Agents. Private agents perform periodic synthetic measurements to
specified targets and are valuable to troubleshoot performance problems due to
the depth of information they provide.
Private
Agents
Private Agents can be installed in branch offices and data
centers either as a software package or virtual appliance. They can be
configured to perform different types of tests that provide visibility over
different layers. The agents perform periodic repeatable measurements to
targets configured by the user (e.g. web servers, network devices, DNS
servers). The agents export their measurements to ThousandEyes collectors in
real-time. This data gets organized into layers of application delivery:
Web / HTTP Server Layer: HTTP
transaction data, including service availability, as well as information on
which step of the transaction is failing; in the example below some agents are
failing to establish a TCP connection to the server.
Network /
End-to-End Metrics Layer: Includes network level data to the server in the
previous step, including packet loss, latency, capacity and available
bandwidth.
Network / Path Visualization Layer: Provides
a hop-by-hop visualization of how traffic is routed between the agents and the
destination; it indicates where packets are being lost and which links are
having long delays
ThousandEyes Private Agents are valuable to troubleshoot
performance problems due to the depth of information they provide.
Conclusion
With ThousandEyes, Enterprises gain the end-to-end level of
visibility required to troubleshoot problems inside, but also outside of their
environment. ThousandEyes Private Agents are distributed as a lightweight
software package that can be installed and configured in minutes. Using the
patented X-Layer technology, ThousandEyes is able to relate different layers of
application delivery through a common thread that allows users to find the
root-cause of problems immediately. ThousandEyes can reduce the MTTR (Mean Time
to Resolution) of infrastructure problems from hours and days to minutes.
=================================================
**Important note** - contact our company for
very powerful solutions for IP management (IPv4 and IPv6, security, firewall and APT solutions:www.tabularosa.net
In addition to this blog, Netiquette IQ has a website with great assets which are being added to on a regular basis. I have authored the premiere book on Netiquette, “Netiquette IQ - A Comprehensive Guide to Improve, Enhance and Add Power to Your Email". My new book, “You’re Hired! Super Charge Your Email Skills in 60 Minutes. . . And Get That Job!” will be published soon follow by a trilogy of books on Netiquette for young people. You can view my profile, reviews of the book and content excerpts at:
www.amazon.com/author/paulbabicki
If you would like to listen to experts in all aspects of Netiquette and communication, try my radio show on BlogtalkRadio Additionally, I provide content for an online newsletter via paper.li. I have also established Netiquette discussion groups with Linkedin and Yahoo. I am also a member of the International Business Etiquette and Protocol Group and Minding Manners among others. Further, I regularly consult for the Gerson Lehrman Group, a worldwide network of subject matter experts and have been a contributor to numerous blogs and publications.
Lastly, I
am the founder and president of Tabula
Rosa Systems, a company that provides “best of breed” products for network,
security and system management and services. Tabula Rosa has a new blog and Twitter site which offers great IT
product information for virtually anyone.
==============================================
No comments:
Post a Comment