Hot answers tagged

41

Disclaimer: I wrote OpenTSDB. I would say that the biggest advantage of Graphite seems to be superior graphing capabilities. It offers more graph types and features. Deployment complexity is also probably a bit lower with Graphite, as it's not a distributed system and thus has fewer moving parts. OpenTSDB, on the other hand, is capable of storing a ...


28

Edit: Updating on Oct. 1 2013 - Some of my original answer has since become obsolete. I'm not sure if you're still active on this site or that you'll see this, but I wanted you to know that I read this question today and it fascinated me, and so I spent all day (when I should have been working) researching Hyper-V and Windows internals and even digging in ...


17

User interface Graphite has some superb graphing tools available. The default web interface is ugly (although functional), but you then have a wealth of great graphing and dashboard options. A few examples: Grafana dashboard, with dynamic (zoomable) graphs Vimeo's GraphExplorer dashboard Cubism.js horizon charts Look here or here to find ...


9

The routes metric is to set preference among routes with equal specificity. That is true of routing in general (i.e. Cisco, Windows, etc). So the model works like: Find the most specific route (aka the longest prefix match*) If there are multiple routes with the same specificity, pick the one with the lowest administrative distance (This distinguishes ...


8

The main metrics would be: Price. Performance. This further breaks down to small-file performance (where latency, i.e. time to first byte dominates) and large-file performance (where transfer rate dominates). Obviously, actually measuring this is darn hard, but here are some decent benchmarks to get you started. Geographical coverage for the regions you're ...


6

It sounds like you may have two problems On your monitoring server, recording the metrics for lots of servers requires more random i/o than your storage can provide. Even if all your metrics are being written to disk, the server may be too overloaded to actually generate graphs from them. On your clients being monitored, the plugins which collect the ...


5

If you wanted to just have a look on the bandwidth actually used, give nload a shot. I always prefer testing the daemon which is serving clients (wget/curl when testing a webserver, lftp for ftp-servers, etc). Artificial tests like iperf are better to check the general throughput of your routers, switches, NICs and IP stacks. HTH, PEra


5

I am not sure what you mean with task specific performance, so you will get a slightly generic answer. There are a lot of different metrics for comparing CPU's, and it depends a lot on what you intend to do with them. The most common meta statistics are: Performance/price Performance/watt And the ever popular weighted index such as Performance/(price ...


5

The metrics that matter are those which: Indicate a problem with the correct and proper operation of the services you provide; or Indicate the root cause of a problem What metrics matter to you depends on what you judge, in your professional opinion, to be the metrics that best fulfil those two criteria. If you don't have the expertise to be able to ...


4

I have extensive experience in this space; I do a lot of work for a couple fortune-5 companies who operate their data centers like an ISP would to the various company departments needing hosting & support services. They typically have two metrics called an SLA (Service Level Agreement) and an OLA (Operational Level Agreement). SLAs are met through the ...


4

I disagree that a virtual server would need to be moved to physical because of performance. Hypervisors are now so close to the metal that there is virtually (pun intended) no performance hit. Especially now that many board makers are including hypervisors on the chipset. If you took two servers with identical hardware, one running a single guest and one ...


4

For systems that are short-lived or where the infrastructure changes often, I use two different tools to handle monitoring. I added a comment asking which metrics were most important to you, and it seems like you're looking for basic "what happened when?" monitoring stats with some alerting... As systems and hardware are abstracted more via cloud services ...


4

Linux provides a number of tools for flexible routing selection. Single routing table In the simplest case, there is just one kernel routing table and no routes with the SRC attribute. This table contains a number of routes, which were placed there manually (ip route add), by the DHCP daemon, or by routing daemons. In this case, the kernel chooses: the ...


4

One way to do it is using tshark. Do the following : Capture traffic with relevant filters on both sides using tshark/tcpdump/whatever produces a pcap formatted file. Once finished, run tshark option -z over capture files. Then : to get packet and byte rates per second : tshark -q -r myfile.cap -z io,stat,1 This will produce something like this : ...


3

I appreciate that it doesn't always suit to have mod_status available and on all of the time, but it and apachetop are the best ways to diagnose these problems. However there are many ways to skin a cat. This trick is useful in a number of circumstances and isn't just Apache specific. It does depend on a number of factors however, and you need to know what ...


3

Connections have metrics that influence their preference, the lower -- the better. Windows will probably assign correct metrics automatically. See KB299540. When both connections are active verify the assignment with route print You can assign metrics manually if needed. See also KB894564.


3

The one case where I had to carry out a V2P was for an MS SQL box that had been running on dual 3.2Ghz dual core CPU's (total CPU 14.4Ghz) that we migrated to an ESX 2.5 cluster where the underlying hardware was newer with more slower (2.4Ghz IIRC) cores. Adding in the ~10% overhead even with 4 vCPU's this VM could only ever get an effective 8-8.5Ghz ...


3

Although being great tools, Munin and other RRDTool frontends (such as Cacti or Ganglia) have known i/o issues and are dificcult to scale when you monitor hundreads of nodes. There are some techniques to deal with this i/o bottleneck though. One of these thecniques is to spread writes across a large number of disks to reduce i/o in each disk. On the other ...


3

I would start with mod_log_config. Define one or more LogFormat/CustomLog setups in httpd.conf with just the stats you're interested in and any metadata about the requests you care to filter by and then you can quickly generate comparative stats from those log files. For example: LogFormat "%t %v %B %D %h %r" statlog CustomLog "|/usr/bin/cronolog logs/stat....


3

The best I have used so far is Observium hands down. It supports almost an endless amount of metrics. We looked at several other options including the ones mentioned here but ultimately decided upon Observium.


3

Why are you including the data rate? 1 / ( 0.004 + 0.0095 ) ~ 74 IOPS.


3

Your basic formulas are reasonable, but this: Data Rate = 156MB/s = (0.156MB/ms / 0.004MB) = 39 can't be right. You need to determine the time needed to read a block, which would be 4 KB / 156 MB/s = 2.5 * 10^-5s ( ~ 0.025 ms) As this is value is negligible, you might omit it for your rule-of-thumb approximation and the calculation would look like 1 ...


2

Have a look at Polymon. http://polymon.codeplex.org From what your describing, it's exactly what you're after. And free.


2

here are some zabbix templates to get started - http://www.zabbix.com/wiki/doku.php?id=contrib:templates also it is not very difficult to write your own script and include results into zabbix - give it a try! m


2

iperf for network bandwidth testing, bonnie++ and/or iozone for disk testing. All of them can be found in the debian repositories.


2

iperf. Or if you want to go old-school, ttcp. http://www.carumba.com/src/ttcp.c


2

I'm not an expert on this subject, but generally speaking: Very hungry I/O applications (especially those who write little and fast) are the ones who gets their own physical server. It's not very hard to find them either, you just run performance monitor and look for high i/o wait times. Also, high-end databases usually get its own dedicated server, ...


2

WGET will show transfer rates. Stick a large temporary file up on the server (say, dd if=/dev/zero of=tempfile bs=1M count=200) and bring it down w/ HTTP. Watch out if you're doing any kind of compression in your HTTP server-- I believe new WGET builds can do gzip encoding. Edit: One could argue that you should probably create a script w/ wget to poll a ...


2

iperf Here ya go: http://www.go2linux.org/how-to-measure-your-network-bandwidth-and-performance-with-iperf


2

I agree with user44700, it is better to separate the availability testing for your servers versus the CDN and track the two independent independently. Your true availability will be Server Avail * CDN Avail, since if either goes down - you are considering it that your page/site is down. This will also cost you less with any of the monitoring vendors. I ...



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