ACE 4710 without Switch

[brandongohyf]

Hi All:

We recently configured a setup to loadbalance 2 application server using ACE4710.  Initially the configuration was to link two app servers directly to ACE4710 without connecting to a Switch, but later, it was advised that ACE4710 is not able to work without connecting to a switch.  I would appreciate if any expert here can clarify and confirm the followings:-

1. ACE4710 is not able to link directly to APP/WEB server, but it must go through a network Switch.

2. If item-1 above is true.  We used to have a older Cisco Loadbalancer which is able to link directly to WEB/APP servers.  What is the reason or advantage of removing this feature?

Thank you very much in advance.

Brandon Goh YF

[rodrguti_2]

Hello Brandon,

What you have been told is true.

The  reason is the following, the ACE has just 4 interfaces, we can just  configure 1 vlan per interface we cannot enable the same vlan in 2  different interfaces, either on a trunk or as switchport access.

If  your 2 web/app servers must reside on the same subnet, this wont work  on the ACE. In other words, you will need to have 1 vlan for the VIP  subnet, 1 vlan for 1 server and another vlan for the second servers, in  total you will be wasting 2 subnets.

In other load  balancers like the CSS, we have the option to use interfaces in the same  vlan, like a regular switch, this is because the CSS support spanning  tree, the ACE does not support spanning tree either. So the ACE has to  have a mechanism to avoid loops.

You can try it, if you  enable lets say vlan 10 on interface g1/1, and then you try to  configure interface g1/2 on vlan 10 as well, the ACE is going to give  you an error message saying that vlan 10 is already configured on  another interface.

Hope this answer your question.

Rodrigo.

[brandongohyf]

Hi Rodrigo,

Thank you very much for attending to my question.

I understand that it is due to some technical limitation that we are unable to implement ACE without a Switch in between.

What I would like to know is that would there be any strong reason for Cisco to drop the feature that is previously supported on CSS and now not on ACE?  such as best practice or performance?.  The problem we are facing now is; the configuration that we proposed does not work due to the limitation on ACE, which now it requires an additional Switch that is seen to be another point of failure for the entire setup. 

Many thanks!

[rodrguti_2]

Hello Brandon,

Scalability is one of the strongest reasons, since the ACE has been designed to operate in large and big datacenter deployments, it is like having a lamborgini, it has just 2 seats but the power is awesome, and the CSS is more like a Family car now, in other words, with the ACE Cisco removed other features and added other in order to meet the new needs, the CSS has been a great device and still it is, but it was a combination of switch-router functionalities besides the load balancing options, it supports spanning tree, ospf, rip, etc.... but not PVST for example, spanning tree is limited, the use of the routing protocol is almost none, so all these lead to create a device more focus on load balancing, and include more strong features that help to meet the new requirements for load balancing and let the other specialized devices to complete a proper design.

So spanning tree was sacrificed, routing protocols as well. Remember that the CSS was created long time ago, when the datacenters were not that large as of now, the serverfarms were not that big, we had 3-4 servers, and with a CSS capable of handling 16 ports, we are ok.

ACE avoids a lot of other issues that the CSS can be either involve or create with the features that are not related with load balancing, like been the spanning tree root for example.

The ACE performance is lot better than the CSS, it also support a lot of new features that the CSS is missing, you have now virtualization, so you can have more than 1 load balancer available and you do not need to buy a new equipment which is very expensive, so comparing that to add a new switch that does not have to be a super switch, you have gain a lot, and you have a very good device.

Hope this answer your question.

Rodrigo.

[brandongohyf]

Hi Rodrigo,

Thank you again for your reply. You mentioned that ACE come with "virtualization" feature, so it allows more than one load balancers.  Can you elaborate this point?  how could it be seen as an advantage or cost saving?  Sorry but I have only very basic knowledge about load balancer / networking.

Brandon Goh YF

[rodrguti_2]

Hi Brandon,

Virtualization help to  reduce power consumption and air conditioning needs and trim the  building space and land requirements that have always been associated  with server farm growth. Virtualization also provides high availability  for critical applications, and streamlines application deployment and  migrations. Virtualization can simplify IT operations.

You can have a production Virtual LB, and another one just for testing, in one single box, and you do not need to buy a new piece of hardware just to do this, you can work on the test environment during business hours, and you are not going to risk the production setup since it is working on a "different" LB.

By default you can have 5 Virtual LB's with the ACE. So you can use a single context just to load balance some specific applications, or in the case that you host your ACE to provide LB for different companies, you can provide a dedicated LB for each company.

Hope this give you a better idea about why Virtualization is an advantage, and you will find the same with Virtual servers.

Rodrigo.

[brandongohyf]

Hi Rodrigo,

Thank you so much for taking the time to reply my posting so quickly.  Really appreciate.

I am now clear about the Virtualization feature, but it is not quite applicable to our environment.  Instead of Virtualization, we are actually looking for Clustering feature.  Our current setup has mutiple db servers, multiple web/app servers, multiple network switch (for redundancy), redundant network card, redundant power supply, etc... With this, the singtle ACE loadbalancer seems to be the point of failure.  May I seek your advice where ACE4710 is able to support clustering?  Can we setup 2 ACEs to receiving requests from one common IP address?  This means we need to implement two ACEs on ACTIVE-ACTIVE mode, and they are doing self-loadbalancing.

Thanks again for your help.

[rodrguti_2]

Hi Brandon,

The ACE does not support Active - Active for the same Virtual address, however it does support connection replication, so the standby ACE will have all the connections replicated in case of a failover, of course there are some limitations that you can read in here:

http://www.cisco.com/en/US/docs/app_ntwk_services/data_center_app_services/ace_appliances/vA4_2_0/configuration/administration/guide/redundcy.html#wp999587

If you have 2 different VIP's you can use the 2 ACE's, but based on your description that does not seem to be the case.

Rodrigo.