Using the Service Control Engine and Deep Packet Inspection in the Data Center
Available Languages
Table Of Contents
Using the Service Control Engine and Deep Packet Inspection in the Data Center
Service Control Solution Overview
Service Control Engine Insertion Strategies
Inline Multi-Gigabit Service Control Point
MGSCP Layer 2/Layer 3 Dispatch Mode
SCE Management and Policy Creation
Using the Service Control Engine and Deep Packet Inspection in the Data Center
This document is a basic primer for use and deployment of the Service Control Engine products in the data center. This document also provides a basis for further discussion of deep packet inspection (DPI) techniques and products.
Contents
Introduction
Deep packet inspection (DPI) provides the ability to look into the packet past the basic header information. DPI intelligently determines the contents of a particular packet, and then either records that information for statistical purposes or performs an action on the packet.
Applications enabled by DPI include the following:
•
Traffic Management, or the ability to control end-user applications such as peer-to-peer applications
•
•
Benefits include increased visibility into the network traffic, which enables network operators to understand usage patterns and to correlate network performance information along with providing usage base billing or even acceptable usage monitoring.
DPI can also reduce the overall costs on the network by reducing operation expenses (OpEx) and capital expenses (CapEx) by providing a more thorough understanding of what is happening with the network, and by providing the ability to direct traffic or to prioritize traffic more intelligently.
Cisco currently has two hardware-based solutions for achieving this DPI functionality: the Cisco Service Control Engine (SCE) product line, and the newly-introduced PISA hardware for the Cisco 6500/7600 Supervisor 32. This document provides basic configuration and performance information with regard to the SCE product family as well as providing comparisons between the SCE and PISA products.
Service Control Solution Overview
The Service Control solution requires the following components for implementation.
•
The Service Control solution requires that a management network be created so that the SCEs, Collection Manager, Service Control Application Suite, and the Service Control Application Suite Reporter can communicate with each other. This network can be separate, or it can be part of a larger, more-involved out-of-band management network that many customers deploy to isolate management of network elements from user data.
•
The SCE is purpose-built hardware platform that performs deep packet inspection, identifies the users, and generates the report data records. The SCE 2000 provides four Gigabit Ethernet (SX or LX) user and network ports, and two FastEthernet management ports. This device requires that a FastEthernet connection be made to the management network, and the unit must see mirrored/duplicated/couples traffic bi-directionally. This document focuses on the SCE 2000 product. (See Figure 1 and Table 1.)
Figure 1 SCE 2020
•
The SCE generates usage data and statistics from the SCE platform and forwards them as raw data records (RDRs), using a simple TCP-based protocol (RDR Protocol). The Service Control solution provides the Collection Manager software as an implementation of a collection system, listening in on RDRs from one or more SCE platforms, and processing them on the local machine. The data is then stored for analysis and reporting functions, as well as simple collection and presentation of data to other operation support and systems (OSS) such as billing. The Collection Manager runs on Sun Solaris or Red Hat Linux platforms. The Collection Manager needs to be connected to the same management network to which the SCE FastEthernet port(s) are connected.
Table 2 lists the Collection Manager software and hardware requirements.
•
The SCAS BB console is the SCE GUI used to create, modify, and apply the service configuration. The SCAS BB Console lets you define services, packages, protocols, bandwidth control, and other items in the configuration. The SCAS BB Console creates a policy configuration file (.pqb), which can then be saved and/or applied to the SCE device(s). The SCAS BB Console runs on any Windows PC (Windows 95 or greater), and it requires a connection to the management network.
•
The SCAS Reporter allows you to query the Collection Manager RDR database and to present the results in a chart or table. This tool provides a valuable resource for understanding the usage patterns and resources used by applications and users that use network resources. This tool can also help your staff to understand the operational impact of various rules and what their impact might be if they are implemented onto the network. This tool runs on any Windows PC (Windows 95 or greater), and it requires a connection to the management network.
•
This optional but recommended component is required to function as middleware software that is used to bridge between the OSS and the SCE platform(s). Subscriber information is stored in the Subscriber Manager database and can then be distributed between multiple devices according to actual subscriber placement. The Subscriber Manager runs on Sun Solaris or Red Hat Linux platforms. This element requires a Fast Ethernet connection to the management network. Software requirements are the same as for the Collection Manager.
The Subscriber Manager provides subscriber awareness, mapping network IDs to subscriber IDs. It obtains subscriber information using dedicated integration modules, which integrate with AAA devices such as RADIUS or DHCP servers.
Subscriber information may be obtained in one of two ways:
–
–
Figure 2 shows a high level connectivity diagram of each of the components of the Service Control solution.
Figure 2 Service Control Solution Components
Service Control Engine Insertion Strategies
There are various ways of positioning the SCE products for use in the network. As part of a data center deployment, only a subset of these were used. This is based on the requirements for maintaining Layer 3 functionality between the aggregation and the core layers of the data center network as well as the requirements of moving to 10 Gig throughput rates. The two methods used allow for a customer to first become operationally experienced with the SCE as well as to begin to baseline traffic patterns on their network to begin the policy creation phase. Policy creation should not be attempted without an understanding of the applications on the network and the traffic flow that these applications place on the network.
Note
Port Mirror
In this deployment scenario, the SCE is deployed in a monitor-only method, and the actual port mirroring can be done at either the core or aggregation layers. This allows the data center operations staff to create the SCE management network and to test functionality, as well as begin to gather baseline traffic analysis on their network.
Inline Multi-Gigabit Service Control Point
In a inline multi-gigabit service control point (MGSCP) configuration, the SCE is inserted between the core data center routers. (See Figure 3.)
Figure 3 Inline MGSCP Deployment
The Service Control solution can also be deployed between the core and aggregation nodes. However, with the requirement of all flows traversing the same SCE, it is more practical to deploy the SCE devices at a layer in the data center and not between. Optional external optical bypass modules and redundant cascade operation provides high availability. The inline deployment permits dynamic control of the traffic including throttling, blocking, and redirection. The benefits of this solution are that any single SCE failure does not result in an outage, and the solution can be architected to provide for an N+1 redundancy model. In addition, from a sizing perspective, this solution allows the data center operators to add additional SCE units as bandwidth needs dictate.
Basics of Dispatch Operation
When configured in dispatch mode, the Catalyst 6500/7600 requires high speed links facing the host device and the network. The Catalyst 6500 dispatches traffic flows to a unique port on the EtherChannel served by an SCE, which performs the configured analysis of the traffic and returns the packets to the original data path. The reason for the unique port is that traffic in the flow must be dispatched to the same SCE to maintain state information.
The Catalyst 6500 gives the network operations staff the flexibility to use Layer 2 or Layer 3 dispatching, depending on their data center deployment. When using Layer 2, packets are dispatched into the EtherChannel using a combination of Layer 2 switching and flow hashing. When Layer 3 routing is used, you need to use policy-based routing (PBR) in combination with flow hashing in the EtherChannel. It is also possible to use both Layer 2 and Layer 3 methods of dispatch to the same SCE cluster group.
For optimum flow dispersion, the number of links in the EtherChannel bundle should be 2^N.
MGSCP Options
Figure 4 shows the Cisco Catalyst 6500/7600 in dispatch mode.
Figure 4 Catalyst 6500/7600 in Dispatch Mode
MGSCP Layer 2 Dispatch Mode
When both the host and the network ports are configured as Layer 2 VLAN ports, the MGSCP solution can be deployed in Layer 2 operational mode. (See Figure 5.)
Figure 5 Layer 2 Dispatch Mode
EtherChannel bundles need to be created to process both the host-side VLAN traffic and the network-side VLAN traffic. The SCE performs its packet inspection and configured policy enforcement, and returns the packet to the original data path. Any single flow must be dispatched to the same SCE to maintain state information. By default, the Layer 2 switching and EtherChannel hash algorithm balances the flows based on the IP header information, and the resulting path selection is the same for all packets in a particular flow. Because most customers deploy redundant cores and aggregation layers, it is worth mentioning how asymmetric flows are handled by the SCE. Because the EtherChannel hash is predictable given a consistent set of inputs, cabling the SCE devices in a uniform fashion to redundant 6500/7600 chassis results in the flows being directed to the same SCE unit, regardless of which chassis the packet traverses.
VLAN translation is then used to avoid misdirecting Layer 2 switched packets back on an EtherChannel. The EtherChannel interface is configured as a trunk port, the trunk port carries the VLAN ID in the .1Q header payload and the .1Q VLAN ID is automatically rewritten at the network EtherChannel connection or by the SCE device from the User VLAN ID to the network EtherChannel VLAN ID. Without the use of VLAN translation, the packet is automatically dropped by the 6500/7600 line card port ASICs.
Configuration Example
In the configuration of the MGSCP solution, remember to enable load balancing on both the ingress and egress ports for the traffic going through the EtherChannel. Before 12.2(18) SXF, you can use the per-source/destination IP load balancing from the global command line. The load balancing hash should be performed on the Supervisor CPU using the following command:
<port-channel load-balance src-ip><port-channel load-balance dst-ip>With releases after 12.2(18) SXF, if you are using line cards with DCEF capabilities, you can now do per-line card load balancing using the following command:
<port-channel load-balance dst-ip module <1-9>>The following examples show the actual switch configuration for MGSCP for a Layer 2 MGSCP configuration.
Global Configuration
!interface Port-channel10switchportswitchport trunk encapsulation dot1qswitchport trunk allowed vlan 201switchport mode trunkno ip addresslacp max-bundle 2!interface Port-channel20switchportswitchport trunk encapsulation dot1qswitchport trunk allowed vlan 100switchport mode trunkno ip addresslacp max-bundle 2!!port-channel per-module load-balanceport-channel load-balance dst-ipport-channel load-balance src-ip module 1!vlan 100name Host-side!vlan 201name Network-sideHost Side Configuration
•
!interface GigabitEthernet1/1switchportswitchport access vlan 100switchport mode accessno ip addressload-interval 30!•
!interface GigabitEthernet2/13description <<< Connected to Backup host side >>>switchportswitchport trunk encapsulation dot1qswitchport trunk allowed vlan 100switchport mode trunkswitchport vlan mapping enableswitchport vlan mapping 201 100no ip addressno cdp enablechannel-protocol lacpchannel-group 20 mode activelacp port-priority 65535!interface GigabitEthernet2/14description <<< Connected to SCE-2 host side >>>switchportswitchport trunk encapsulation dot1qswitchport trunk allowed vlan 100switchport mode trunkswitchport vlan mapping enableswitchport vlan mapping 201 100no ip addresschannel-protocol lacpchannel-group 20 mode active!interface GigabitEthernet2/15description <<< Connected to SCE-1 host side >>>switchportswitchport trunk encapsulation dot1qswitchport trunk allowed vlan 100switchport mode trunkswitchport vlan mapping enableswitchport vlan mapping 201 100no ip addresschannel-protocol lacpchannel-group 20 mode active!Network Side Configuration
•
!interface GigabitEthernet5/1switchportswitchport access vlan 201switchport mode accessno ip addressload-interval 30!•
!interface GigabitEthernet1/13description <<< Connected to Backup Net >>>switchportswitchport trunk encapsulation dot1qswitchport trunk allowed vlan 201switchport mode trunkswitchport vlan mapping enableswitchport vlan mapping 100 201no ip addressno cdp enablechannel-protocol lacpchannel-group 10 mode activelacp port-priority 65535!interface GigabitEthernet1/14description <<< Connected to SCE-2 Net >>>switchportswitchport trunk encapsulation dot1qswitchport trunk allowed vlan 201switchport mode trunkswitchport vlan mapping enableswitchport vlan mapping 100 201no ip addresschannel-protocol lacpchannel-group 10 mode active!interface GigabitEthernet1/15description <<< Connected to SCE-1 Net >>>switchportswitchport trunk encapsulation dot1qswitchport trunk allowed vlan 201switchport mode trunkswitchport vlan mapping enableswitchport vlan mapping 100 201no ip addresschannel-protocol lacpchannel-group 10 mode active!MGSCP Layer 2/Layer 3 Dispatch Mode
When deploying the MGSCP solution in a mixed Layer 2/Layer 3 environment; for example, at the aggregation layer where the host-facing ports are Layer 2 and the core-facing ports are Layer 3, the only change needed is for a switched virtual interface (SVI) to allow packets to be routed in and out of the network EtherChannel. (See Figure 6.)
Figure 6 Layer 2/Layer 3 Deployment
MGSCP Layer 3 Dispatch Mode
In the event that MGSCP needs to be deployed in an all-Layer 3 environment; for example, at the core layer of the data center or at the aggregation layer when running a completely routed data center design, you need to use an ACL on the 6500/7600 to define an action for a range of IP addresses. The EtherChannel is now configured as Layer 3 with an assigned IP address, but the hashing algorithm still occurs based on the flow IP address. (See Figure 7.)
Figure 7 Complete Layer 3 Deployment
N+1 Redundancy
The MGSCP solution takes advantage of the network infrastructure to provide for an N+1 level of failover support. EtherChannel links can be monitored using Link Aggregation Control Protocol (LACP). LACP allows for some links in the bundle to be active while others are waiting in standby mode. When an active link fails, one of the standby links is moved into the active list. Because the SCE devices are "bump in the wire", if a link on one side of the SCE fails, the opposite link must be brought down as well. The SCE devices can be configured with a "link reflection" for sub-second recognition. LACP also has a fast timeout configuration to detect a failed link in a bundle. This detection can occur within 1-2 seconds. (See Figure 8.)
Figure 8 Link Failure
SCE Management and Policy Creation
The SCE solution relies heavily on the management stations for configuration and day-to-day operation. The Engage Console becomes the primary operational window that operators use to control network policy and to generate reports, as well as to configure SCE units operations. (See Figure 9.)
Figure 9 SCAS BB Console Window
For example, the Network Navigator tab allows the operator to see the SCE units in operation and to group them in a logical collection. (See Figure 10.)
This is also where policies can be deployed to the SCE units; either to a single SCE device, or to a collection of SCE units. It may be useful for an operator to configure a policy and deploy it to one SCE device. Then, upon checking for issues such as performance impact and desired behavior, they can deploy the policy to the rest of the SCEs in the network.
Figure 10 Network Navigator Tab
Deploying a New SCE
This section outlines the steps necessary to install a new SCE into the network and bring it into production.
Note
Step 1
Figure 11 Basic Configuration of SCE
Step 2
Ensure that you have provisioned the network completely, as previously outlined.
Figure 12 Cable the SCE to the Network
Step 3
You need to provide the IP address and the name of the device.
Figure 13 Adding the SCE to the Network Navigator
Step 4
Figure 14 Downloading Latest Firmware to the New SCE
Step 5
When this step is complete, the unit is now in sync with any other SCE engines deployed in the cluster.
Figure 15 Deploy Site Policy Configuration (PQB)
Changing a Policy
As a routine course of operation, the policies in place may need to be adjusted for things such as new applications, higher or lower bandwidth demands, and so on. The steps to edit a policy are as follows.
Step 1
You can adjust the bandwidth in a uni-directional fashion so care should be taken to ensure you understand which traffic flow you want to impact.
Figure 16 Policy Editor Window
Step 2
Figure 17 Applying Policy to the SCE
Updating Signatures
This section deals with the routine updating of the signature files. Cisco publishes protocol packs that customers can download. Cisco recommends that customers use the latest release of the protocol pack because these updates provide greater application awareness.
After downloading the protocol pack, the network operator uses the SCAS BB Console to select the SCE(s) to which to deploy the protocol pack. The updates can then be installed. (See Figure 18.)
It is useful to keep all the SCE(s) in one location or cluster on the same protocol pack.
Figure 18 Updating the Signature Pack
Reports
The ability to collect, analyze, and report are a key feature of the SCE product. Following is a list of many of the reports that are part of the Service Control solution. Users can customize existing reports or create new report types. Through the use of the protocol packs that Cisco distributes routinely, the SCE solution stays current with new signature files and port-based applications. Users can again define their own signatures, but the product has built-in support for 100+ protocols.
The following are examples of the reports generated by the Service Control solution.
•
–
–
–
–
–
–
–
–
–
•
–
–
–
–
–
•
–
–
•
–
–
–
•
–
–
–
–
•
–
–
–
–
–
–
Example reports are shown in the following figures.
Figure 19 Global Bandwidth per Service
Figure 20 Global Bandwidth per Service—Downstream Direction
Figure 21 Daily Usage Volume
Figure 22 Top Subscribers in Selected Service
Figure 23 Subscriber Usage Bandwidth
Figure 24 VoIP MOS
Appendix A—SCE and PISA
Recently, Supervisor 32 with PISA hardware was introduced. This section provides a comparison of the SCE and PISA products.
At this time, testing of the Supervisor 32 with PISA has not been performed, but may be undertaken and documented in the future.
Supervisor Engine 32 PISA is primarily targeted at the enterprise in the WAN edge:
•
•
•
•
SCE had traditionally been positioned for the service provider market by offering the following:
•
•
•
The ability to cluster the SCE and provide a multi-gigabit capability, however, make this product suited for inclusion in the data center as a method for inspecting traffic between the campus network and the data center, or in the high speed Metro Ethernet connections that can exist between multi-site data center environments.
The following tables provide a comparison of the two technologies.
Feedback