Does anyone have any perspective on how well Asterisk performs and scales inside a Xen hypervisor environment? Obviously, the answer depends largely on what sort of hardware it's running on, whether it's in PAE mode, whether it's a newer CPU that has some paravirtualisation instruction sets available to assist it, how much memory is allocated to each VM, and other architectural considerations. Any perspective would be helpful, however. -- Alex Balashov Evariste Systems Web : http://www.evaristesys.com/ Tel : (+1) (678) 954-0670 Direct : (+1) (678) 954-0671 Mobile : (+1) (706) 338-8599
>Does anyone have any perspective on how well Asterisk performs and >scales inside a Xen hypervisor environment?I tried on many different pieces of hardware with various recent Xen versions and it always had some level of unpredictability and was not as reliable as running on bare hardware. I wouldn't do it for production but it was fine for testing (sort of :>). This was of course w/ ztdummy in a pure sip env. jlc
My experience is very limited, but you asked for any perspective, so... I put an Asterisk with freePBX on a linode server (linode.com), just to play with it a few months ago. I can say that it worked to the point of being able to dial out with my Polycom phone on a FiOS connection, through the * box, and a SIP termination service like Vitelity, and to receive calls in the other direction. No problems with that, and kinda cool to be able to throw a virtual PBX out there with so little expense. I did not stress test it, nor did I examine resource usage to gain perspective on scalability. More of a proof of concept. One issue that comes up with regard to this is about timing sources for MOH, etc. Related to this and of general use to know, I believe one can associate PCI cards with particular VMs, but it's been a few months since I configured a Xen box of my own, so the details have already fled from my feeble brain... But I hope that's helpful. Alex Balashov wrote:>Does anyone have any perspective on how well Asterisk performs and >scales inside a Xen hypervisor environment?>Obviously, the answer depends largely on what sort of hardware it's >running on, whether it's in PAE mode, whether it's a newer CPU that has >some paravirtualisation instruction sets available to assist it, how >much memory is allocated to each VM, and other architectural >considerations.>Any perspective would be helpful, however.-------------- next part -------------- A non-text attachment was scrubbed... Name: smime.p7s Type: application/x-pkcs7-signature Size: 3234 bytes Desc: S/MIME Cryptographic Signature Url : http://lists.digium.com/pipermail/asterisk-users/attachments/20081001/4ccf9dad/attachment.bin
Alex Balashov wrote:> Does anyone have any perspective on how well Asterisk performs and > scales inside a Xen hypervisor environmentI did asterisk in xen recently (ubuntu hardy, xen 3.2). 10 sip users, 1 fxs, 1 e1. used xen pci passthrough for digium cards. no problems at this day. same machine hosts vms for fileserver, sms-gateway, imap, smtp, squid and router Alexander
On Tue, 2008-09-30 at 19:05 -0400, Alex Balashov wrote:> Does anyone have any perspective on how well Asterisk performs and > scales inside a Xen hypervisor environment? > > Obviously, the answer depends largely on what sort of hardware it's > running on, whether it's in PAE mode, whether it's a newer CPU that has > some paravirtualisation instruction sets available to assist it, how > much memory is allocated to each VM, and other architectural > considerations. > > Any perspective would be helpful, however. >We have been doing some test. To avoid single point of failure (as musch as possible) we had mysql, ldap and asterisk14 and asterisk16 in separate virtual machines. No problem what so ever. DOM-u's are easy to scale up, considering mem or cpu. Would suggest to have either channel-banks/ata's or PRI-boards in a separate machine(s). never got the pci forwarding from a hypervisor to a dom-U properly working Hans