Coyote Point Brings Load Balancing to Virtual Servers
Coyote Point Brings Load Balancing to Virtual Servers
By: Frank Ohlhorst dnu
Savvy network managers are using open-source software to provision new, inexpensive virtual servers, fueling the exponential growth of server clusters, server farms and Web applications. However, just throwing additional servers into the mix is an inefficient way to add capacity. Efficiency requires that loads be properly spread across resources, and that's where load balancing solutions prove their mettle. Load balancing reduces latency, improves throughput and eliminates bottlenecks.
Coyote Point Systems takes the appliance approach to load balancing, offering four different appliances. These devices differ based on design traffic load and subfeatures, but they all share the same management console and basic feature set. I tested one of these units, Coyote Point's top-tier E650GX (V8.6) appliance, for ease of use, feature set and suitability to task.
With this product, which has an MSRP of $14,395 and includes a year of support, Coyote Point has reinvented the idea of load balancing by shifting traffic shaping from basic Layer 4 algorithms to Layer 7 application-aware calculations. When combined with VMware vSphere support, that makes the device a complete traffic-acceleration solution.
The physical installation of the E650GX, which includes 22 Gigabit Ethernet interfaces, was a snap: Other than installing the unit in a rack, administrators will probably spend most of their time routing the appropriate Ethernet cables from the servers that are going to participate in load balancing.
That is simple on smaller networks: You can simply plug the connection from your firewall into the external port on the E650GX and then plug each segment of the LAN into the internal ports on the device. All ports on the E650GX are Gigabit Ethernet and support full-duplex operation. That means it's very unlikely that the device will introduce any bottlenecks into LAN or WAN connections, and none were detected during performance testing.
Setting Up Basic Parameters
I found it pretty easy to get load balancing working on the E650GX simply by setting up some basic parameters, such as defining server clusters and some load balancing rules. For example, if you had nine servers running a Web application, you would plug each of those servers into an internal port on the E650GX appliance and then place those servers into logical clusters-perhaps creating three logical clusters with three servers occupying each cluster.
Since cluster definitions are done logically, it is a simple matter to make any changes, which provides a great deal of deployment flexibility. Administrators can also define VLANs (virtual LANs) based on ip address segments, which can speed up local requests and reduce latency by isolating local traffic to the same logical segment.
The E650GX offers numerous load balancing options. These include "Match Rules and Custom Load-Balancing Policies," which define policies based on Layer 4 requests, Layer 7 requests or using Boolean logic. The simplest load balancing policies are defined using Layer 4 parameters, such as least connections, fastest response, adaptive and round robin, as well as a server-based, agent-based algorithm. The agent-based algorithm proves to be the most accurate, as long as the monitoring agent is installed on each server in the cluster.
Policies based on Layer 7 are a little more sophisticated and incorporate traffic analysis to determine how to load balance the member clusters. That works by identifying administrator-defined protocols to trigger a load balancing policy to route traffic to a particular cluster.
Policies that incorporate Boolean logic can handle particular requests based on administrator-defined events. Those policies can be used to reroute traffic if a server fails to respond (failover routing) or to route based on a schedule.
Cluster definition is only part of the story here. As mentioned above, administrators also have the ability to define rules to handle traffic flow and load balancing decisions. Coyote Point calls those definitions "Smart Events."
The rules work with a number of parameters, such as server load, traffic type, server weighting and connection persistence. While the underlying technology used for traffic-routing decisions is very complex, the E650GX offers rule-creation wizards and a common-sense procedural layout to simplify rule definition.
Of course, most of those load balancing and traffic-routing capabilities are expected from a device such as the E650GX. However, the real magic of the E650GX-and the biggest value it offers-comes from its support for VMware's vSphere products. Simply put, that means the E650GX brings application load balancing and traffic shaping to the world of virtual servers.
Support for vSphere clusters is powered by Coyote Point's ability to use VMware's APIs, which allows the E650GX to judge loads on virtual servers and then route requests based on virtual server loads as well as on administrator-defined load balancing policies.
Coyote Point's support for IPMI (Intelligent Platform Management Interface)-capable servers brings additional capabilities to the E650GX, allowing even more flexibility when it comes to balancing and shaping traffic for virtual and physical servers. The IPMI specification allows third-party products to power servers on and off and remotely execute commands. Administrators can use IPMI-based policies to automatically power up additional servers when traffic loads hit a certain level, and shut down those servers once traffic load drops.
I found the interface easy to use in many ways, making it simple to set up many of the unit's capabilities. That simplicity extends to the device's dashboards and reporting menus. The dashboards offer real-time information on how well the device is handling traffic, while reports offer historical information on monitored parameters.
The EG650GX offers additional features, such as SSL (Secure Sockets Layer) acceleration, HTTP compression and global/geographic load balancing. SSL acceleration helps to reduce the latency found in HTTPS (HTTP Secure) requests by offloading the packet encryption onto the device. HTTP compression helps to reduce latency by compressing and optimizing HTTP requests.
Global/geographic load balancing can be used to balance traffic across geographical clusters, placing requests on servers that are closest to the user, as far as latency and bandwidth are concerned. The EG650GX also offers support for connection persistence, which may be needed for transaction based e-commerce solutions and failover support if a second device is added to the network.