Message 308060 - Python tracker [proxy]

Message308060

Author	yselivanov
Recipients	asvetlov, lukasz.langa, methane, pitrou, yselivanov
Date	2017-12-11.16:50:18
SpamBayes Score	-1.0
Marked as misclassified	Yes
Message-id	<1513011018.08.0.213398074469.issue32251@psf.upfronthosting.co.za>
In-reply-to

Content
>> Looks nice. Can it speed up aiohttp too? > Yes. > aiohttp uses own streams but public API and internal implementation are pretty close to asyncio streams. > Moreover C accelerated HTTP parser should work with proposed BufferedProtocol seamlessly. I did some benchmarks, and it looks like BufferedProtocol can make httptools up to 5% faster for relatively small requests < 10Kb. Don't expect big speedups there. For asyncpg, benchmarks that fetch a lot of data (50-100Kb) get faster up to 15%. So we'll definitely use the BufferedProtocol in asyncpg. For applications that need to handle megabytes of data per request (like Antoine's benchmark) the speedup will be up to 2x.

Content

>> Looks nice.  Can it speed up aiohttp too?

> Yes.
> aiohttp uses own streams but public API and internal implementation are pretty close to asyncio streams.
> Moreover C accelerated HTTP parser should work with proposed BufferedProtocol seamlessly.

I did some benchmarks, and it looks like BufferedProtocol can make httptools up to 5% faster for relatively small requests < 10Kb.  Don't expect big speedups there.

For asyncpg, benchmarks that fetch a lot of data (50-100Kb) get faster up to 15%.  So we'll definitely use the BufferedProtocol in asyncpg.

For applications that need to handle megabytes of data per request (like Antoine's benchmark) the speedup will be up to 2x.

History
Date	User	Action	Args
2017-12-11 16:50:18	yselivanov	set	recipients: + yselivanov, pitrou, asvetlov, methane, lukasz.langa
2017-12-11 16:50:18	yselivanov	set	messageid: <1513011018.08.0.213398074469.issue32251@psf.upfronthosting.co.za>
2017-12-11 16:50:18	yselivanov	link	issue32251 messages
2017-12-11 16:50:18	yselivanov	create