1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
|
Selection Strategies: Balancing vs Stickiness
Balancing strategies:
leastConns
random
roundRobin
shuffle
Sticky strategies:
client IP
USER
HOST
What's better for FTP, SFTP connections? RoundRobin or LeastConns? And
why? The HAproxy docs for the 'balance' keyword say that 'leastconn' is
best for long-lived connections:
leastconn The server with the lowest number of connections receives the
connection. Round-robin is performed within groups of servers
of the same load to ensure that all servers will be used. Use
of this algorithm is recommended where very long sessions are
expected, such as LDAP, SQL, TSE, etc... but is not very well
suited for protocols using short sessions such as HTTP. This
algorithm is dynamic, which means that server weights may be
adjusted on the fly for slow starts for instance.
(from http://haproxy.1wt.eu/download/1.4/doc/configuration.txt)
Is this true? Corroboration or statistics showing the relationship?
Dangers of LeastConns
Consider that you are bringing up a new server, or restarting a server.
Being new, it will have no connections -- and the load balancer sees
this. Thus in a short period of time, ALL new connections may end up
slamming that server. If the server can handle this type of workload,
then you are good. If, on the other hand, the server performs warm-up
tasks (e.g. filling caches, performing discovery of necessary downstream
resources, etc), then this is quite undesirable.
This is why HAproxy has a "warm up" period.
Links/References
http://www.timgalyean.com/2011/03/load-balancing-with-haproxy-and-apache/
http://blog.loadbalancer.org/load-balancing-windows-terminal-server-haproxy-and-rdp-cookies/
http://blogs.citrix.com/2010/09/02/load-balancing-least-connections/
http://www.brocade.com/support/Product_Manuals/ServerIron_SLBGuide/health.4.19.html#49842
https://kb.wisc.edu/ns/page.php?id=13201
http://serverfault.com/questions/457506/ha-proxy-roundrobin-vs-leastconn
http://blog.exceliance.fr/2011/08/03/layer-7-load-balancing-proxy-mode/
https://devcentral.f5.com/articles/intro-to-load-balancing-for-developers-ndash-the-algorithms#.UsPAPY3TNe4
Nice description of methods, and quite a variety of them.
https://devcentral.f5.com/articles/intro-to-load-balancing-for-developers-ndash-how-they-work#.UsT6cI3TNe4
Reuse the JPG/pic there?
https://devcentral.f5.com/articles/intro-load-balancing-for-developers-ndash-the-architect-rsquos-view#.UsT6dI3TNe4
https://devcentral.f5.com/articles/advanced-load-balancing-for-developers-ndash-adcs-what-rsquos-the-difference#.UsT6XY3TNe4
The 'Programmability' bit is a good point. mod_proxy might need some
ftpdctl actions implemented for altering things on-the-fly.
https://devcentral.f5.com/articles/advanced-load-balancers-for-developers-adcs-the-code#.UsT6Y43TNe4
http://www.onjava.com/pub/a/onjava/2001/09/26/load.html
Nice discussion of DNS round robin
http://www.javaworld.com/article/2077921/architecture-scalability/server-load-balancing-architectures--part-1--transport-level-load-balancing.html
http://www.javaworld.com/article/2077922/architecture-scalability/server-load-balancing-architectures--part-2--application-level-load-balanci.html
http://www.infoq.com/articles/scalability-principles
http://www.infoq.com/articles/ebay-scalability-best-practices
Note Best Practice #4, and how that pertains to Black Pearl (think
intra-service messaging/notifications)
Aside: Software Architect Role
http://www.codingthearchitecture.com/2008/02/11/the_key_difference_between_developer_and_architect_roles.html
http://www.infoq.com/presentations/Role-of-the-Architect
Need to watch this!
|
