Tag Info

Hot answers tagged

6

It's a case of top reporting memory usage per process as if it where the only process running. In reality all the apache processes have shared memory (common linked libraries), and this shared memory is being reported for each process so the total usage across all apache processes appears to be more than it really is. There are numerous posts that discuss ...


4

You forgot to calculate the SHR, or Shared portion. Try (7*40) - (3*40) = 160, which quite closely matches your used memory. Shared memory is the memory which is shared among the processes; the Apache libraries and stuff like that.


4

Preventing a DDoS is mostly about not being a target. Don't host game servers, gambling/porn sites, and other things that tend to get people annoyed. Mitigating a DDoS attack comes in two forms: being able to ignore traffic and shed excess load, which is useful when you're under an attack that tries to take you down by overloading your machines (and also ...


4

There has been a lot of talk about SSL creating excess CPU/memory load on websites. Really? Where? I have never once heard of moving to SSL causing CPU-releated problems. Unless your webserver is already CPU-bound (highly unlikely), you won't even notice the additional CPU load that SSL causes. You should be able to test HTTPS performance in the ...


3

You don't mention what kind of perimeter security you have in place. With Cisco firewalls you can limit the number of embryonic (half sessions) that your firewall will allow before it cuts them off, while still allowing full sessions to go through. By default it's unlimited, which offers no protection.


3

FreeBSD 7/8/9 and Linux 2.6 are basically equal in their out-of-box networking capacity. I've seen benchmarks where one outdoes the other, and they seem to go back and forth. I'd recommend using the platform you're most familiar with, as you'll likely be able to tune it better.


3

Although my knowledge of Linux is pretty much nonexistent, I can tell you that we had to tune several default Linux kernel networking settings to handle the traffic levels we regularly get on these sites. We use ubuntu server, but all the advice we found seemed to be distribution agnostic. My point: I don't think any distro is set up correctly for massive ...


3

Assertions. http://jakarta.apache.org/jmeter/usermanual/test_plan.html#assertions


2

I would check out this post: http://jmetertips.blogspot.com/2009/04/ajax-testing-with-jmeter.html


2

One good source for information is at this site. One measure which they only mention in passing (and which is worth researching further) is enabling SYN cookies. This prevents an entire class of DoS attacks by preventing an attacker from opening a large number of 'half-open' connections in an attempt to reach the maximum number of file descriptors permitted ...


2

Hardware-assisted load-balancers such as Foundry ServerIron's and Cisco ACEs are great for dealing with huge numbers of the main types of DOS/DDOS attacks but aren't so flexible as software solutions which can 'learn' newer techniques quicker.


2

Are you absolutely sure Apache is the real bottleneck for you? You didn't tell anything about your "fairly simple PHP application", but for me it sounds like it's using some database and that starts to slow things down quite soon during your benchmark. The traditional Apache values you should consider tuning are MaxClients, ServerLimit, MinSpareServers, ...


2

You should download and install Java 1.6 SDK from Oracle site. GNU implementation of Java (GNU GCJ) can't handle startup parameters as you can see.


2

The ramp-up time is not really important except for special cases, so they recommend in the same paragraph to keep it at the default value of 1 sec, which means all threads you enter in the number of threads field will be started within 1 second. Also, for the local OS it doesn't matter at all if you start all threads at once or with a ramp-up time, the ...


2

Well, is your server capable of running 2048 concurrent Apache processes (as you set in ServerLimit/MaxClients)? I bet it is not and actually you don't need so many concurrent Apache workers. I highly recommend you to setup an asynchronous reverse proxy like nginx or lighttpd to offload handling clients w/slow connections to that async proxy. And believe me ...


2

I suspect the answer lies in the things that browsers do, that jmeter and ab don't. Namely cookies, and javascript (and images, to some extent). Similarly, I'd argue that your benchmarks aren't really representative of real browsing. When I visit a site, I don't go to the homepage and reload the page over and over. I go and click lots of different ...


1

As a child to Thread Group add Loop Controller and put everything other inside that Loop Controller. At the end of Loop Controller add Constant Timer. It should work that way.


1

As Tom said, you should simulate real browser behavior. For JMeter it's quite easy: Add HTTP Cookie Manager Add HTTP Cache Manager Add HTTP Header Manager, and configure User-Agent of browser which you usually use. In HTTP Request or HTTP Request Defaults in Option task check settings: Retrieve All Embedded Resources from HTML Files to simulate a browser ...


1

check your page with a tool like "YSLOW", that might give a hint where the timelags begin. and i'd always check for mysql slow_queries and use a tuning-script to evaluate, if your bottleneck is your db. if not -> (you did already) check the bare performance of your webserver with a simple static file; then with ab; then investigate the whole page with a ...


1

First thing is that, you should try different URLs for testing, not just one, which I believe you are trying. Also, try increasing the values for MaxServers and MaxClients so that you don't hit a bottleneck. You can check the same whether you are hitting a bottleneck for the number of processes by using this command # watch "ps ax | grep -http | wc -l" ...


1

I don't know whether Cloud solution would be acceptable for you but there is a company which is offering Jmeter as a Service - ie basically SaaS solution. They claim that they have rich scripting capabilities so it might be one of the options for you. http://aws.amazon.com/customerapps/3299 http://blazemeter.com/ I never used their services myself though ...


1

How high is that number? Can you tell us more about the application that goes on top? How serious deployment are you planing? If you're on Windows XP there's something called half-open connection limit, which has caused problems in the past. It's similar to your case, everything works fine until it reaches a certain number of parallel connections. Here's ...


1

That message shows that JMeter attempted to connect to your app but was unable to. This usually means that Tomcat has stopped responding to requests because you have hit some limit. There are limits on the number of threads that Tomcat starts to handle requests as well as limitations imposed by the Java VM that can stop requests from being handled (it is ...


1

I had the same problem, which turned out to be the limit of half-open socket connections in XP. I used this tool to solve the problem: http://www.megaleecher.net/Tcpip.sys_Patch_To_Increase_Windows_XP_Connection_Limit


1

My guess would be network congestion or network overhead. Your jmeter machines not only connect to the client, but they are also sending back results to the jmeter master machine. More jmeter machines does not always mean more load on your application. Try to find httpd timing logs, and try to find network througput for all machines and see what you come up ...


1

Can you post jMeter profile - depending on your cookie configuration, you will find differing results from EE's FPC Also, how have you configured your cache backend, to use File/Backend/DB/Memcache/Redis? http://www.sonassi.com/knowledge-base/magento-kb/what-is-memcache-actually-caching-in-magento/) Both of the above points will effect the frequency and ...


1

We were also having a transient "boxes fail health checks for no good reason" problem and from working with Amazon support it turns out there is an interaction between the ELBs and the Apache KeepaliveTimeout. If the health check interval is larger than the timeout then the healch checker can try to reuse a bad connection and it fails the test and tosses ...


1

The best way to stress test ELB is to get the ips used behind the cname they provide. Used those to hit the load balancer. Make sure there is at lease one image in every az you selected for the ELB. Amazon dynamically scales the ips behind the ELB, Your load balancer is probably hitting just a single ip. I'm not sure about the sporadic behavior you're ...


1

One way to do is it to create lots of aliases for your eth0, like for each in $(seq 1 254); do ifconfig eth0:$each 192.168.1.$each; done And then just do a little bit of NAT voodoo: iptables -t nat -A POSTROUTING -o eth0 -j SNAT --to 192.168.1.1-192.168.1.254 But that may change the source IP too often for you ... I'm not 100% sure what you're going to ...


1

There isn't a good number for all those parameters, other than as high as possible for your particular testing subject machine. Just load up as many concurrent connections you can with Jmeter (toying with the balance of instances and threads), and fire to your machine. Do so until a certain component (ie MySQL) fails, then try to optimize that component and ...



Only top voted, non community-wiki answers of a minimum length are eligible