Hot answers tagged

3

Use the 10 Gb/s link. Otherwise you restrict the VMs to 1 Gb/s when communicating with multiple clients or with other VMs (communication to other VMs should be through the virtual switch and not the physical NIC). If your network card supports it, use VMQ and / or SR-IOV. Ideally you should create an SR-IOV virtual NIC inside every VM. Most cards should be ...


2

Running PHP is expensive, in terms of CPU usage and RAM. Any time you can avoid invoking the PHP interpreter, you should. Caching is the primary way you do this, so long as the content isn't custom generated for every user. My little AWS t2.micro can serve around 2500 pages per second from its Nginx cache, but about 20 pages per second if PHP (HHVM) has to ...


2

You can find physical core use on the host with “\Hyper-V Hypervisor Logical Processor(_Total)\% Total Run Time” (LPTR) If you have more virtual processors than logical processors (cores) they will context switch and you get low virtual use; do not oversubscribe processors. The hypervisor has a small amount of overhead so you need a core or two more than ...


1

As user the-wabbit suggests, there is request-merging going on. You can see that in the column avgrq-sz, the average request size - which shows a significant increase. Now 'await' is the time spent in the queue plus the time spent servicing those requests. If a small request, let's call it 'x', is merged with a couple of other requests (y and z, issued ...


1

a. Have you check wait state of the running query ? Select * from sys.sysprocesses will do b. is there any resource governor in place ? c. have you use recompile hint to test it ? d. have you check the query plan ? e. have you tried to force the plan to run in paralel mode using dbcc cpuweight ? f. have you update the statistics of the tables?


1

I received an answer from microsoft support which sort of makes sense. Anyway for future searches: No matter what the platform, the editions (not Web/Biz) provides capacity to ensure the DTU and response constraints as specified by https://azure.microsoft.com/en-us/documentation/articles/sql-database-benchmark-overview/#metrics (look at the ...


1

"t" == Terabytes. Move the decimal three places to the right, and you have 12GB.


1

I don't recommend this type of passthrough setup anymore. It's a lose-lose on most counts, especially with reliability and performance. It can certainly be done, but the safest solution (especially with a hypervisor like ESXi) is to just use a supported RAID controller and local storage. If you want ZFS storage, build a standalone ZFS storage system.


1

You could enable HTTP/2, which will reduce the number of connections for compatible browsers, reducing page load time. This won't reduce latency for that connection though. Demo here. See also this question - latency is key in this situation. I ran a speed test against my https server in Sydney, Australia. Testing from another Sydney server the SSL ...


1

HTTPS does slow your website down despite all claims to the contrary. This is because the client and server need to negotiate the SSL/TLS ciphers before it can start. However after that, the slowdown is negligible for most sites, and there are massive benefits to SSL. Additionally the default is http, so someone entering that for a https-only site will need ...


1

It's possible to set on the kernel command line with e.g. elevator=deadline. I don't know for sure, but I consider it very unlikely that the default changed between 573.3.1 and 573.8.1 of Red Hat's 2.6.32 kernel. You can inspect your kernel command line by reading /proc/cmdline.


1

It turns out the difference in response length is a methodological one. I didn't notice the Apache server was returning all 301 codes during this test, because the URLs were being re-written. I had to change the server URL and path to exactly match what the re-write rules enforce. After that, the content lengths matched up perfectly, and the Apache server ...



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