r = 2800 / (1+3) = 700
Copyright © 2019, Oracle and/or its affiliates. All rights reserved.
|
Planning your Deployment DRAFT |
| Previous | Next | Contents |
Before deploying GlassFish Server, first determine the performance and availability goals, and then make decisions about the hardware, network, and storage requirements accordingly.
The following topics are addressed here:
At its simplest, high performance means maximizing throughput and reducing response time. Beyond these basic goals, you can establish specific goals by determining the following:
What types of applications and services are deployed, and how do clients access them?
Which applications and services need to be highly available?
Do the applications have session state or are they stateless?
What request capacity or throughput must the system support?
How many concurrent users must the system support?
What is an acceptable average response time for user requests?
What is the average think time between requests?
You can calculate some of these metrics using a remote browser emulator (RBE) tool, or web site performance and benchmarking software that simulates expected application activity. Typically, RBE and benchmarking products generate concurrent HTTP requests and then report the response time for a given number of requests per minute. You can then use these figures to calculate server activity.
The results of the calculations described in this chapter are not absolute. Treat them as reference points to work against, as you fine-tune the performance of GlassFish Server and your applications.
The following topics are addressed here:
In broad terms, throughput measures the amount of work performed by GlassFish Server. For GlassFish Server, throughput can be defined as the number of requests processed per minute per server instance.
As described in the next section, GlassFish Server throughput is a function of many factors, including the nature and size of user requests, number of users, and performance of GlassFish Server instances and back-end databases. You can estimate throughput on a single machine by benchmarking with simulated workloads.
High availability applications incur additional overhead because they periodically save session data. The amount of overhead depends on the amount of data, how frequently it changes, and how often it is saved. The first two factors depend on the application in question; the latter is also affected by server settings.
Consider the following factors to estimate the load on GlassFish Server instances.
The following topics are addressed here:
Users interact with an application through a client, such as a web browser or Java program. Based on the user’s actions, the client periodically sends requests to the GlassFish Server. A user is considered active as long as the user’s session has neither expired nor been terminated. When estimating the number of concurrent users, include all active users.
Initially, as the number of users increases, throughput increases correspondingly. However, as the number of concurrent requests increases, server performance begins to saturate, and throughput begins to decline.
Identify the point at which adding concurrent users reduces the number of requests that can be processed per minute. This point indicates when optimal performance is reached and beyond which throughput start to degrade. Generally, strive to operate the system at optimal throughput as much as possible. You might need to add processing power to handle additional load and increase throughput.
A user does not submit requests continuously. A user submits a request, the server receives and processes the request, and then returns a result, at which point the user spends some time before submitting a new request. The time between one request and the next is called think time.
Think times are dependent on the type of users. For example, machine-to-machine interaction such as for a web service typically has a lower think time than that of a human user. You may have to consider a mix of machine and human interactions to estimate think time.
Determining the average think time is important. You can use this duration to calculate the number of requests that need to be completed per minute, as well as the number of concurrent users the system can support.
Response time refers to the amount of time GlassFish Server takes to return the results of a request to the user. The response time is affected by factors such as network bandwidth, number of users, number and type of requests submitted, and average think time.
In this section, response time refers to the mean, or average, response time. Each type of request has its own minimal response time. However, when evaluating system performance, base the analysis on the average response time of all requests.
The faster the response time, the more requests per minute are being processed. However, as the number of users on the system increases, the response time starts to increase as well, even though the number of requests per minute declines.
A system performance graph indicates that after a certain point, requests per minute are inversely proportional to response time. The sharper the decline in requests per minute, the steeper the increase in response time.
The point of the peak load is the point at which requests per minute start to decline. Prior to this point, response time calculations are not necessarily accurate because they do not use peak numbers in the formula. After this point, (because of the inversely proportional relationship between requests per minute and response time), the administrator can more accurately calculate response time using maximum number of users and requests per minute.
Use the following formula to determine Tresponse, the response time (in seconds) at peak load:
Tresponse = n/r - Tthink
where
n is the number of concurrent users
r is the number requests per second the server receives
Tthink is the average think time (in seconds)
To obtain an accurate response time result, always include think time in
the equation.
Example 2-1 Calculation of Response Time
If the following conditions exist:
Maximum number of concurrent users, n, that the system can support at peak load is 5,000.
Maximum number of requests, r, the system can process at peak load is 1,000 per second.
Average think time, Tthink, is three seconds per request.
Thus, the calculation of response time is:
Tresponse = n/r - Tthink = (5000/ 1000) - 3 sec. = 5 - 3 sec.
Therefore, the response time is two seconds.
After the system’s response time has been calculated, particularly at peak load, compare it to the acceptable response time for the application. Response time, along with throughput, is one of the main factors critical to GlassFish Server performance.
If you know the number of concurrent users at any given time, the response time of their requests, and the average user think time, then you can calculate the number of requests per minute. Typically, start by estimating the number of concurrent users that are on the system.
For example, after running web site performance software, the administrator concludes that the average number of concurrent users submitting requests on an online banking web site is 3,000. This number depends on the number of users who have signed up to be members of the online bank, their banking transaction behavior, the time of the day or week they choose to submit requests, and so on.
Therefore, knowing this information enables you to use the requests per minute formula described in this section to calculate how many requests per minute your system can handle for this user base. Since requests per minute and response time become inversely proportional at peak load, decide if fewer requests per minute is acceptable as a trade-off for better response time, or alternatively, if a slower response time is acceptable as a trade-off for more requests per minute.
Experiment with the requests per minute and response time thresholds that are acceptable as a starting point for fine-tuning system performance. Thereafter, decide which areas of the system require adjustment.
Solving for r in the equation in the previous section gives:
r = n/(Tresponse + Tthink)
Example 2-2 Calculation of Requests Per Second
For the values:
n = 2,800 concurrent users
Tresponse = 1 (one second per request average response time)
Tthink = 3, (three seconds average think time)
The calculation for the number of requests per second is:
r = 2800 / (1+3) = 700Therefore, the number of requests per second is 700 and the number of requests per minute is 42000.
When planning how to integrate the GlassFish Server into the network, estimate the bandwidth requirements and plan the network in such a way that it can meet users' performance requirements.
The following topics are addressed here:
You can separate external traffic, such as client requests, from the internal traffic, such as session state failover, database transactions, and messaging. Traffic separation enables you to plan a network better and augment certain parts of the network, as required.
To separate the traffic, run each server instance on a multi-homed machine. A multi-homed machine has two IP addresses belonging to different networks, an external IP and an internal IP. The objective is to expose only the external IP to user requests. The internal IP is used only by the cluster instances for internal communication. For details, see "Using the Multi-Homing Feature With GMS" in GlassFish Server Open Source Edition High Availability Administration Guide.
To plan for traffic on both networks, see Estimating Bandwidth Requirements. For external networks, follow the guidelines in Calculating Bandwidth Required and Estimating Peak Load. To size the interfaces for internal networks, see Choosing Network Cards.
To decide on the desired size and bandwidth of the network, first determine the network traffic and identify its peak. Check if there is a particular hour, day of the week, or day of the month when overall volume peaks, and then determine the duration of that peak.
During peak load times, the number of packets in the network is at its highest level. In general, if you design for peak load, scale your system with the goal of handling 100 percent of peak volume. Bear in mind, however, that any network behaves unpredictably and that despite your scaling efforts, it might not always be able handle 100 percent of peak volume.
For example, assume that at peak load, five percent of users occasionally do not have immediate network access when accessing applications deployed on GlassFish Server. Of that five percent, estimate how many users retry access after the first attempt. Again, not all of those users might get through, and of that unsuccessful portion, another percentage will retry. As a result, the peak appears longer because peak use is spread out over time as users continue to attempt access.
Based on the calculations made in Establishing Performance Goals, determine the additional bandwidth required for deploying GlassFish Server at your site.
Depending on the method of access (T-1 lines, ADSL, cable modem, and so on), calculate the amount of increased bandwidth required to handle your estimated load. For example, suppose your site uses T-1 or higher-speed T-3 lines. Given their bandwidth, estimate how many lines are needed on the network, based on the average number of requests generated per second at your site and the maximum peak load. Calculate these figures using a web site analysis and monitoring tool.
Example 2-3 Calculation of Bandwidth Required
A single T-1 line can handle 1.544 Mbps. Therefore, a network of four T-1 lines can handle approximately 6 Mbps of data. Assuming that the average HTML page sent back to a client is 30 kilobytes (KB), this network of four T-1 lines can handle the following traffic per second:
6,176,000 bits/10 bits = 772,000 bytes per second
772,000 bytes per second/30 KB = approximately 25 concurrent response pages per second.
With traffic of 25 pages per second, this system can handle 90,000 pages per hour (25 x 60 seconds x 60 minutes), and therefore 2,160,000 pages per day maximum, assuming an even load throughout the day. If the maximum peak load is greater than this, increase the bandwidth accordingly.
Having an even load throughout the day is probably not realistic. You need to determine when the peak load occurs, how long it lasts, and what percentage of the total load is the peak load.
Example 2-4 Calculation of Peak Load
If the peak load lasts for two hours and takes up 30 percent of the total load of 2,160,000 pages, this implies that 648,000 pages must be carried over the T-1 lines during two hours of the day.
Therefore, to accommodate peak load during those two hours, increase the number of T-1 lines according to the following calculations:
648,000 pages/120 minutes = 5,400 pages per minute
5,400 pages per minute/60 seconds = 90 pages per second
If four lines can handle 25 pages per second, then approximately four times that many pages requires four times that many lines, in this case 16 lines. The 16 lines are meant for handling the realistic maximum of a 30 percent peak load. Obviously, the other 70 percent of the load can be handled throughout the rest of the day by these many lines.
The following topics are addressed here:
To plan availability of systems and applications, assess the availability needs of the user groups that access different applications. For example, external fee-paying users and business partners often have higher quality of service (QoS) expectations than internal users. Thus, it may be more acceptable to internal users for an application feature, application, or server to be unavailable than it would be for paying external customers.
There is an increasing cost and complexity to mitigating against decreasingly probable events. At one end of the continuum, a simple load-balanced cluster can tolerate localized application, middleware, and hardware failures. At the other end of the scale, geographically distinct clusters can mitigate against major catastrophes affecting the entire data center.
To realize a good return on investment, it often makes sense to identify availability requirements of features within an application. For example, it may not be acceptable for an insurance quotation system to be unavailable (potentially turning away new business), but brief unavailability of the account management function (where existing customers can view their current coverage) is unlikely to turn away existing customers.
At the most basic level, a cluster is a group of GlassFish Server instances—often hosted on multiple physical servers—that appear to clients as a single instance. This provides horizontal scalability as well as higher availability than a single instance on a single machine. This basic level of clustering works in conjunction with the HTTP load balancer plug-in, which accepts HTTP and HTTPS requests and forwards them to one of the instances in the cluster. The ORB and integrated JMS brokers also perform load balancing to GlassFish Server clusters. If an instance fails, becomes unavailable (due to network faults), or becomes unresponsive, requests are redirected only to existing, available machines. The load balancer can also recognize when a failed instance has recovered and redistribute load accordingly.
One way to achieve high availability is to add hardware and software redundancy to the system. When one unit fails, the redundant unit takes over. This is also referred to as fault tolerance. In general, to maximize high availability, determine and remove every possible point of failure in the system.
The level of redundancy is determined by the failure classes (types of failure) that the system needs to tolerate. Some examples of failure classes are:
System process
Machine
Power supply
Disk
Network failures
Building fires or other preventable disasters
Unpredictable natural catastrophes
Duplicated system processes tolerate single system process failures, as well as single machine failures. Attaching the duplicated mirrored (paired) machines to different power supplies tolerates single power failures. By keeping the mirrored machines in separate buildings, a single-building fire can be tolerated. By keeping them in separate geographical locations, natural catastrophes like earthquakes can be tolerated.
Failover capacity planning implies deciding how many additional servers and processes you need to add to the GlassFish Server deployment so that in the event of a server or process failure, the system can seamlessly recover data and continue processing. If your system gets overloaded, a process or server failure might result, causing response time degradation or even total loss of service. Preparing for such an occurrence is critical to successful deployment.
To maintain capacity, especially at peak loads, add spare machines running GlassFish Server instances to the existing deployment.
For example, consider a system with two machines running one GlassFish Server instance each. Together, these machines handle a peak load of 300 requests per second. If one of these machines becomes unavailable, the system will be able to handle only 150 requests, assuming an even load distribution between the machines. Therefore, half the requests during peak load will not be served.
Design decisions include whether you are designing the system for peak or steady-state load, the number of machines in various roles and their sizes, and the size of the administration thread pool.
The following topics are addressed here:
In a typical deployment, there is a difference between steady state and peak workloads:
If the system is designed to handle peak load, it can sustain the expected maximum load of users and requests without degrading response time. This implies that the system can handle extreme cases of expected system load. If the difference between peak load and steady state load is substantial, designing for peak loads can mean spending money on resources that are often idle.
If the system is designed to handle steady state load, it does not have all the resources required to handle the expected peak load. Thus, the system has a slower response time when peak load occurs.
How often the system is expected to handle peak load will determine whether you want to design for peak load or for steady state.
If peak load occurs often—say, several times per day—it may be worthwhile to expand capacity to handle it. If the system operates at steady state 90 percent of the time, and at peak only 10 percent of the time, then it may be preferable to deploy a system designed around steady state load. This implies that the system’s response time will be slower only 10 percent of the time. Decide if the frequency or duration of time that the system operates at peak justifies the need to add resources to the system.
Based on the load on the GlassFish Server instances and failover requirements, you can determine the number of applications server instances (hosts) needed. Evaluate your environment on the basis of the factors explained in Estimating Load on GlassFish Server Instances to each GlassFish Server instance, although each instance can use more than one Central Processing Unit (CPU).
The default admin-thread-pool size of 50 should be adequate for most
cluster deployments. If you have unusually large clusters, you may need
to increase this thread pool size. In this case, set the
max-thread-pool-size attribute to the number of instances in your
largest cluster, but not larger than the number of incoming
synchronization requests that the DAS can handle.
The Java Message Service (JMS) API is a messaging standard that allows Java EE applications and components to create, send, receive, and read messages. It enables distributed communication that is loosely coupled, reliable, and asynchronous. Message Queue, which implements JMS, is integrated with GlassFish Server, enabling you to create components that send and receive JMS messages, including message-driven beans (MDBs).
Message Queue is integrated with GlassFish Server using a resource adapter also known as a connector module. A resource adapter is a Java EE component defined according to the Java EE Connector Architecture (JCA) Specification. This specification defines a standardized way in which application servers such as GlassFish Server can integrate with enterprise information systems such as JMS providers. GlassFish Server includes a resource adapter that integrates with its own JMS provider, Message Queue. To use a different JMS provider, you must obtain and deploy a suitable resource adapter that is designed to integrate with it.
Creating a JMS resource in GlassFish Server using the Administration
Console creates a preconfigured connector resource that uses the Message
Queue resource adapter. To create JMS Resources that use any other
resource adapter (including GenericJMSRA), you must create them under
the Connectors node in the Administration Console.
In addition to using resource adapter APIs, GlassFish Server uses additional Message Queue APIs to provide better integration with Message Queue. This tight integration enables features such as connector failover, load balancing of outbound connections, and load balancing of inbound messages to MDBs. These features enable you to make messaging traffic fault-tolerant and highly available.
The following topics are addressed here:
Message Queue supports using multiple interconnected broker instances known as a broker cluster. With broker clusters, client connections are distributed across all the brokers in the cluster. Clustering provides horizontal scalability and improves availability.
A single message broker scales to about eight CPUs and provides sufficient throughput for typical applications. If a broker process fails, it is automatically restarted. However, as the number of clients connected to a broker increases, and as the number of messages being delivered increases, a broker will eventually exceed limitations such as number of file descriptors and memory.
Having multiple brokers in a cluster rather than a single broker enables you to:
Provide messaging services despite hardware failures on a single machine.
Minimize downtime while performing system maintenance.
Accommodate workgroups having different user repositories.
Deal with firewall restrictions.
Message Queue allows you to create conventional or enhanced broker clusters. Conventional broker clusters offer service availability. Enhanced broker clusters offer both service and data availability. For more information, see "Configuring and Managing Broker Clusters" in Open Message Queue Administration Guide.
In a conventional cluster, having multiple brokers does not ensure that transactions in progress at the time of a broker failure will continue on the alternate broker. Although Message Queue reestablishes a failed connection with a different broker in a cluster, transactions owned by the failed broker are not available until it restarts. Except for failed in-progress transactions, user applications can continue on the failed-over connection. Service failover is thus ensured.
In an enhanced cluster, transactions and persistent messages owned by the failed broker are taken over by another running broker in the cluster and non-prepared transactions are rolled back. Data failover is ensured for prepared transactions and persisted messages.
In a configuration for a conventional broker cluster, each destination is replicated on all of the brokers in a cluster. Each broker knows about message consumers that are registered for destinations on all other brokers. Each broker can therefore route messages from its own directly-connected message producers to remote message consumers, and deliver messages from remote producers to its own directly-connected consumers.
In a cluster configuration, the broker to which each message producer is directly connected performs the routing for messages sent to it by that producer. Hence, a persistent message is both stored and routed by the message’s home broker.
Whenever an administrator creates or destroys a destination on a broker, this information is automatically propagated to all other brokers in a cluster. Similarly, whenever a message consumer is registered with its home broker, or whenever a consumer is disconnected from its home broker—either explicitly or because of a client or network failure, or because its home broker goes down—the relevant information about the consumer is propagated throughout the cluster. In a similar fashion, information about durable subscriptions is also propagated to all brokers in a cluster.
A shared database of cluster change records can be configured as an alternative to using a master broker. For more information, see "Configuring and Managing Broker Clusters" in Open Message Queue Administration Guide and "Using Message Queue Broker Clusters With GlassFish Server" in GlassFish Server Open Source Edition High Availability Administration Guide.
By default, Message Queue brokers (JMS hosts) run in the same JVM as the GlassFish Server process. However, Message Queue brokers (JMS hosts) can be configured to run in a separate JVM from the GlassFish Server process. This allows multiple GlassFish Server instances or clusters to share the same set of Message Queue brokers.
The GlassFish Server’s Java Message Service represents the connector
module (resource adapter) for Message Queue. You can manage the Java
Message Service through the Administration Console or the asadmin
command-line utility.
In GlassFish Server, a JMS host refers to a Message Queue broker. The GlassFish Server’s Java Message Service configuration contains a JMS Host List (also called AddressList) that contains all the JMS hosts that will be used.
There are three types of integration between GlassFish Server and Message Queue brokers: embedded, local, and remote. You can set this type attribute on the Administration Console’s Java Message Service page.
Embedded Java Message Service
If the Type attribute is EMBEDDED, GlassFish Server and the JMS broker are colocated in the same virtual machine. The JMS Service is started in-process and managed by GlassFish Server. In EMBEDDED mode, JMS operations on stand-alone server instances bypass the networking stack, which leads to performance optimization. The EMBEDDED type is most suitable for stand-alone GlassFish Server instances. EMBEDDED mode is not supported for enhanced broker clusters.
With the EMBEDDED type, use the Start Arguments attribute to specify Message Queue broker startup parameters.
With the EMBEDDED type, make sure the Java heap size is large enough to allow GlassFish Server and Message Queue to run in the same virtual machine.
Local Java Message Service
If the Type attribute is LOCAL, GlassFish Server starts and stops the Message Queue broker. When GlassFish Server starts up, it starts the Message Queue broker specified as the Default JMS host. Likewise, when the GlassFish Server instance shuts down, it shuts down the Message Queue broker. The LOCAL type is most suitable for use with enhanced broker clusters, and for other cases where the administrator prefers the use of separate JVMs.
With the LOCAL type, use the Start Arguments attribute to specify Message Queue broker startup parameters.
Remote Java Message Service
If the Type attribute is REMOTE, GlassFish Server uses an externally configured broker or broker cluster. In this case, you must start and stop Message Queue brokers separately from GlassFish Server, and use Message Queue tools to configure and tune the broker or broker cluster. The REMOTE type is most suitable for brokers running on different machines from the server instances (to share the load among more machines or for higher availability), or for using a different number of brokers and server instances.
With the REMOTE type, you must specify Message Queue broker startup parameters using Message Queue tools. The Start Arguments attribute is ignored.
In the Administration Console, you can set JMS properties using the Java Message Service node for a particular configuration. You can set properties such as Reconnect Interval and Reconnect Attempts. For more information, see "Administering the Java Message Service (JMS)" in GlassFish Server Open Source Edition Administration Guide.
The JMS Hosts node under the Java Message Service node contains a list of JMS hosts. You can add and remove hosts from the list. For each host, you can set the host name, port number, and the administration user name and password. By default, the JMS Hosts list contains one Message Queue broker, called "default_JMS_host," that represents the local Message Queue broker integrated with GlassFish Server.
In REMOTE mode, configure the JMS Hosts list to contain all the Message Queue brokers in the cluster. For example, to set up a cluster containing three Message Queue brokers, add a JMS host within the Java Message Service for each one. Message Queue clients use the configuration information in the Java Message Service to communicate with Message Queue broker.
In addition to the Administration Console, you can use the asadmin
command-line utility to manage the Java Message Service and JMS hosts.
Use the following asadmin commands:
Configuring Java Message Service attributes: asadmin set
Managing JMS hosts:
asadmin create-jms-host
asadmin delete-jms-host
asadmin list-jms-hosts
Managing JMS resources:
asadmin create-jms-resource
asadmin delete-jms-resource
asadmin list-jms-resources
For more information on these commands, see the GlassFish
Server Open Source Edition Reference Manual or the corresponding man
pages.
You can specify the default JMS Host in the Administration Console Java Message Service page. If the Java Message Service type is LOCAL, GlassFish Server starts the default JMS host when the GlassFish Server instance starts. If the Java Message Service type is EMBEDDED, the default JMS host is started lazily when needed.
In REMOTE mode, to use a Message Queue broker cluster, delete the default JMS host, then add all the Message Queue brokers in the cluster as JMS hosts. In this case, the default JMS host becomes the first JMS host in the JMS host list.
You can also explicitly set the default JMS host to one of the JMS hosts. When the GlassFish Server uses a Message Queue cluster, the default JMS host executes Message Queue-specific commands. For example, when a physical destination is created for a Message Queue broker cluster, the default JMS host executes the command to create the physical destinations, but all brokers in the cluster use the physical destination.
To accommodate your messaging needs, modify the Java Message Service and JMS host list to suit your deployment, performance, and availability needs. The following sections describe some typical scenarios.
For best availability, deploy Message Queue brokers and GlassFish Servers on different machines, if messaging needs are not just with GlassFish Server. Another option is to run a GlassFish Server instance and a Message Queue broker instance on each machine until there is sufficient messaging capacity.
Installing the GlassFish Server automatically creates a domain administration server (DAS). By default, the Java Message Service type for the DAS is EMBEDDED. So, starting the DAS also starts its default Message Queue broker.
Creating a new domain also creates a new broker. By default, when you add a stand-alone server instance or a cluster to the domain, its Java Message Service is configured as EMBEDDED and its default JMS host is the broker started by the DAS.
In EMBEDDED or LOCAL mode, when a GlassFish Server is configured, a Message Queue broker cluster is auto-configured with each GlassFish Server instance associated with a Message Queue broker instance.
In REMOTE mode, to configure a GlassFish Server cluster to use a Message Queue broker cluster, add all the Message Queue brokers as JMS hosts in the GlassFish Server’s Java Message Service. Any JMS connection factories created and MDBs deployed then uses the JMS configuration specified.
In some cases, an application may need to use a different Message Queue
broker cluster than the one used by the GlassFish Server cluster. To do
so, use the AddressList property of a JMS connection factory or the
activation-config element in an MDB deployment descriptor to specify
the Message Queue broker cluster.
For more information about configuring connection factories, see "Administering JMS Connection Factories and Destinations" in GlassFish Server Open Source Edition Administration Guide. For more information about MDBs, see "Using Message-Driven Beans" in GlassFish Server Open Source Edition Application Development Guide.
When an application client or standalone application accesses a JMS administered object for the first time, the client JVM retrieves the Java Message Service configuration from the server. Further changes to the JMS service will not be available to the client JVM until it is restarted.
| Previous | Next | Contents |
Copyright © 2019, Oracle and/or its affiliates. All rights reserved.
DRAFT