CODEWITHSUNDEEP
pythonamazon-simpledbboto
Chris Hayes
Chris Hayes

Reputation: 4035

improve read spead simple db boto

I have a script that reads simpledb domains and writes them to s3. the performance is mediocre at best. Is there a way to increase the read speed?

import boto
import datetime
from xml.dom.minidom import Document
from boto.s3.key import Key

awsa = "myawsaccesskey"
awss = "myawssecretkey"

conn = boto.connect_sdb(awsa, awss)
domains = conn.get_all_domains()
s3conn = boto.connect_s3(awsa, awss)
archbucket = s3conn.get_bucket("simpledbbu")
for d in domains:
    print d.name
    doc = None
    doc = Document()
    root = doc.createElement("items")
    doc.appendChild(root)
    countermax = 0
    counter = 0
    for item in d:
        print "loading {0} of {1}".format(counter,countermax)
        counter += 1
        node = doc.createElement("item")
        node.setAttribute("itemName", item.name)
        for k,v in item.items():
            if not isinstance(v, basestring):
                i = 0
                for val in v:
                    node.setAttribute("{0}::{1}".format(k,i),val)
                    i += 1
            else:
                node.setAttribute(k,v)
        root.appendChild(node)
    k = Key(archbucket)
    k.key = "{0}/{1}".format(datetime.date.today().strftime("%Y%m%d"),d.name)
    x = doc.toprettyxml(indent="  ")
    k.set_contents_from_string(x)

here's the profile:

ncalls  tottime percall cumtime percall filename:lineno(function)
2035    1.312   0.001   1.312   0.001   {built-in method read}
2   0.445   0.223   0.445   0.223   {built-in method do_handshake}
17  0.355   0.021   0.355   0.021   {built-in method write}
2   0.321   0.161   0.321   0.161   {_ssl.sslwrap}
2   0.292   0.146   0.292   0.146   {_socket.getaddrinfo}
2   0.177   0.089   0.177   0.089   {method 'connect' of '_socket.socket' objects}
14  0.012   0.001   0.077   0.005   {built-in method Parse}
2   0.01    0.005   0.047   0.023   __init__.py:24(<module>)
3369    0.01    0   0.012   0   item.py:71(endElement)
1   0.009   0.009   3.185   3.185   backupSimpleDb_0.0.py:1(<module>)
3508    0.009   0   0.03    0   expatreader.py:300(start_element)
4145    0.008   0   0.011   0   StringIO.py:208(write)
3508    0.007   0   0.019   0   handler.py:31(startElement)
3508    0.007   0   0.02    0   handler.py:37(endElement)
1   0.006   0.006   0.006   0.006   {nt.urandom}
3114    0.006   0   0.009   0   item.py:58(startElement)
1208    0.005   0   0.005   0   minidom.py:343(__init__)
1208    0.005   0   0.025   0   minidom.py:686(setAttribute)
258/3   0.005   0   0.024   0.008   minidom.py:794(writexml)
1   0.004   0.004   0.007   0.007   exception.py:26(<module>)
1   0.004   0.004   0.007   0.007   expatreader.py:4(<module>)
1   0.004   0.004   0.007   0.007   urllib.py:23(<module>)
1   0.004   0.004   0.006   0.006   utils.py:5(<module>)

Upvotes: 0

Views: 362

Answers (2)

Ryan
Ryan

Reputation: 392

I agree with the accepted answer but if you're just looking to backup your simpledb domains on s3, I've had decent performance with boto's to_xml() function. Plus, you don't need to roll your own simpledb parser.

Upvotes: 0

Aaron Digulla
Aaron Digulla

Reputation: 328754

Run 5 or 10 threads in parallel and have each copy one domain.

This is actually pretty simple if you use a queue for domains to copy and have the threads wait for elements in the queue.

http://www.ibm.com/developerworks/aix/library/au-threadingpython/

Upvotes: 3

Related Questions