А будут ли добавляться другие алгоритмы распределения нагрузки?
В смысле кроме round-rbin по весам ?
Да.
А какие есть ?
Кроме привязки бэкендов к кукам и ip-адерсам ?
Игорь Сысоев
Например по количеству одновременно выполняемых запросов бэкэндами,
какой-то выполняет больше.. какой-то меньше, а MaxClients все равно у
апача есть..
Позволю себе привести список алгоритмой от циски:
To specify the load-balancing algorithm for a content rule, use the balance
command available in content configuration mode. The options are:
? balance aca - ArrowPoint Content Awareness load-balancing algorithm (see
the ?Using ArrowPoint Content Awareness Based on Server Load and
Weight? section in Chapter 1, Configuring Services). ACA balances the
traffic over the services based on load or on server weight and load.
? balance destip - Destination IP address division algorithm. The CSS
directs
all client requests with the same destination IP address to the same
service.
This option is typically used in a caching environment.
? balance domain - Domain name division algorithm. The CSS divides the
alphabet evenly across the number of caches. It parses the host tag for
the first
four letters following the first dot and then uses these characters of the
domain name to determine to which server it should forward the request. This
option is typically used in a caching environment.
? balance domainhash - Internal CSS hash algorithm based on the domain
string. The CSS parses the host tag and does an exclusive XOR hash across
the entire host name. It then uses the XOR hash value to determine to which
server to forward the request. This method guarantees that all requests with
the same host tag will be sent to the same server in order to increase the
probability of a cache hit. This option is typically used in a caching
environment.
Note If you are using the domainhash load-balancing method with proxy
cache services, you may see duplicate sites across caches because the
CSS balances on the first GET request in a persistent connection
unless the subsequent GET request does not match a rule with the
same proxy service specified. If you are concerned with duplicate hits
across caches, reset persistence to remap and disable persistence on
the rule. Issue the (config) persistence reset remap command
globally and the (config-owner-content) no persistent command on
the content rule.
? balance leastconn - Least connection algorithm. This balance method
chooses a running service that has the fewest number of connections.
We do not recommend that you use UDP content rules with the leastconn
load-balancing algorithm. The service connection counters do not increment
and remain at 0 because UDP is a connectionless protocol. Because the
counters remain at 0, the CSS will give inconsistent results.
? balance roundrobin - Roundrobin algorithm (default). The CSS resolves the
request by evenly distributing the load to resolve domain names among local
and remote content domain sites.
? balance srcip - Source IP address division algorithm. The CSS directs all
client requests coming from the same source IP address to the same service.
This option is generally used in a caching configuration.
? balance url - URL division algorithm. The CSS divides the alphabet evenly
across the number of caches. It then parses the URL for the first four
characters located after the portion of the URL matched by the rule. For
example, if the URL in a content rule is configured for "/news/*", the CSS
will balance on the first four characters following "/news/". This option is
typically used in a caching environment.
? balance weightedrr - Weighted roundrobin algorithm. The CSS uses
roundrobin but weighs some services more heavily than others depending on
the server?s configured weight. All servers have a default weight of 1.
To set
a server weight, use the add service weight command in owner-content
mode.
? balance urlhash - Internal CSS hash algorithm based on the URL string. The
CSS parses the URL and performs an XOR hash across the URL. It then uses
the XOR hash value to determine to which server to forward the request. This
method guarantees that all requests for the same URL will be sent to the
same
server in order to increase the probability of a cache hit. This option is
typically used in a caching environment.
Note A Layer 5 content rule supports the HTTP CONNECT, GET, HEAD, POST,
PUSH, and PUT methods. In addition, the CSS recognizes and forwards the
following HTTP methods directly to the destination server in a transparent
caching environment but does not load balance them: RFC 2068 - OPTIONS,
TRACE and RFC 2518 - PROPFIND, PROPPATCH, MKCOL, MOVE, LOCK,
UNLOCK, COPY, DELETE. In a transparent caching environment (for example,
no VIP address on a Layer 5 content rule), the CSS bypasses these HTTP
methods,
and they are forwarded to the destination server.
--
тел./факс: +7 (812) 323-23-23 (Санкт-Петербург)
тел./факс: +7 (095) 737-09-77 (Москва)
С уважением, Нарманов Рустам.
