PDA

View Full Version : Shared Hosting Service Improvements


FH-Dave
03-22-2006, 01:53 AM
I am creating this special thread so people can freely discuss about our service.

As part of the preparetion that we have done in the last 2 weeks, in the last two days, Hary had been working in both Equinix and Telehouse data centers to push ahead for the service improvements that we had been planning. So these are the service improvements that you will be seeing coming to you shortly. Please do not hesitate to post your questions/concerns/comments.

A. Mail service improvements

As I have noted elsewhere, a major mprovement on mail service that we have made is to ensure none of our mail servers will be listed in major RBLs for more than 5 minutes. I will even argue that this improvement is more improtant than other mail server improvements. What good will it be to have a secondary MX, RAID1 protected mail server, when incoming emails can't be received and outgoing emails can't be delivered because the mail sever IP is listed in major RBLs? what we do is to make sure that the IP use to accept incoming mails will never be listed in any RBLs and the IP used to deliver emails will never be listed in any major RBLs for a period of longer than 5 minutes. Status: This has been in implementation for the last 3 months and so far this has worked very well.

Mail servers migration to RAID1 capable mail servers. While installing replacement server for mail2, we just realized that the mail1 server has been ready for the last 62 days. Again, the last several mail issues have forced me to postpone scheduling another mail service interruption. However, as this has been brought up multiple times, I have no problem scheduling this migration within the next 7 days. I hope no customers will have any problem with it. We will give no less than 72 hours advance notifications to all customers for this migration. Status: All of our mail servers equipped with SCSI drives in RAID1 since April 5th, 2006

Backup mail server (acting as secondary MX) will be offered within the next few days. As we have several mail servers, one mail server will act as relay for other mail servers. Please note again that the backup mail server will only accept incoming emails when the primary mail servers are down. The backup mail servers will not allow anybody to have any mail service when the primary mail servers are down. For this reason, I hate to use the term "backup mail server" as it gives wrong impression of what it does. In any case, as this is sensitive for some customers to have (even though under the wrong impression), I simply agree to just do it at this moment. Status: As part of our default service, secondary offsite MX record has been implemented since 05/01/2006

Backing Up mail servers. In the event of a complete mail failure, we can still recreate all mailboxes as the information is stored in HSphere database. Backing up all mailboxes has been suggested. However, this decision is still pending. For one, mail server is not meant to be used as online file storage. With the IMAP capability, more and more users are now not only store their mails on the mail server, but also store their files as well. We do not want to encourage customers to treat the mail servers as remote file servers. Furthermore, backing up the mail servers (very I/O intensive) will further increase load on mail servers. At this present moment, we are processing roughly about 300000-500000 mail deliveries per day on each mail server, which is also very I/O intensive. Again, we are still evaluating whether we want to backup the mail servers at this moment. Status: Daily Backup on all inboxes/mail servers has beend one since April 14th, 2006


B. Name Service Improvements

We have prepared a dedicated server (RAID1) that we will allocate only for NS1. Migration of NS1 to its own dedicated server will be done within the next 3 weeks. Status: Migration of NS1 to its own dedicated server was performed on 04/10/2006
We have also prepared another dedicated server (RAID1) that we will allocate for NS3 / forums. NS3 will be put on our Telehouse data center. Migration of NS3 to its offsite dedicated server will be done in the next 2 weeks, after both mail servers migration. Status: Migration of NS3 to its own dedicated server was performed on 04/10/2006
We have upgraded all of our name servers to BIND 9.3.1, which (according to PSOFT) should eliminate the mysterious hostname disappearing problem that we have encountered numerous times in the past.


C. Network Improvements

Even though we did not make any formal notification about this, we have turned up our BGP session with NAC. NAC is a regional ISP which has strong peering/transit in the east coast/UK. This will simply make our network more redundant.
Joe (forum nick: FH-Joe) has been constantly working on tweaking and improving our network beyond standar BGP implementation.


D. Support Improvements

We are adding new personnel to our system admin staff. Please welcome our new junior admin, Jeff Jones (forum nick: FH-Jeff). I have known Jeff since before I started Fluid Hosting and I believe he will be a valuable addition to our admin team.
Our highest priority at this moment is to expand our support department. To this, we will be adding three additional L1 support personnel. Initially they will be under intensive training and I hope within the next 2-3 months, they will be ready to work by themselves.
In addition to 24/7 helpdesk support, we now offer you 24/7 Live Support. Phone support is still kept to 12/5 at this moment, at least until new support personnels have completed their training.

FH-John
04-06-2006, 06:59 AM
Wanted to update everyone that's been watching this thread. Both mail1, and mail2 were successfuly moved to newer servers which employ RAID1 for disk fault tolerance.

FH-John
04-12-2006, 02:08 AM
For those that don't keep close tabs on our announcements. ns1 and ns3 were migrated recently, and ns3 was moved off-site to Telehouse.

FH-Dave
04-12-2006, 02:20 AM
Backing up folders/mailboxes on mail servers are currently being tested. At this moment, we are testing this on mail2. We will see how much additional stress this will cause to the mail server as well as how much space we will need on the backup server.

backup MX records are still being worked on. It's currently available since it's already part of HSphere feature. However, what we are trying to do is to have our backup MX records located off-network. This is not supported by HSphere. We are trying to see if there is a work around this.

FH-Dave
04-21-2006, 12:23 PM
After some testing, I am happy to announce that all mail servers are currently being backed up every day. We will keep the backup for 7 days.

theOG
04-21-2006, 08:25 PM
Thanks, dave... john... others

dagta
04-22-2006, 04:01 AM
After some testing, I am happy to announce that all mail servers are currently being backed up every day. We will keep the backup for 7 days.


Awesome! Thank you!! :)

FH-John
05-02-2006, 04:52 AM
Another happy announcement. We have finalized the setup of our off-site backup MX, and enabled it on all accounts.

The delay in making this possible was due to a limitation in H-Sphere. H-Sphere while supporting backup MXs, is unable to be configured so that one or more systems will be the only backup MXs.

During the past few weeks, we've been working on a solution, and eventually came to developing our own add-on to H-Sphere which makes this possible.

In other words, it was a lot of work to make happen, but it's here now. Enjoy. :)

theOG
05-02-2006, 05:04 AM
Grand... Thanks a lot!

I don't totally understand know how this helps... but any backup should be good!

Thanks.

FH-John
05-02-2006, 05:09 AM
I'd be happy to explain how this helps, since I imagine others may not know as well.

A backup MX handles the reception of inbound e-mail in the event that a domain's primary mail server is non-responsive, or down. By placing our backup MX off-site, should we have a network outage, or mail server issue, your domain's e-mail will be spooled up until the mail server for your domain is running again.

FH-Dave
05-02-2006, 09:15 AM
I am not only happy, but proud to announce this as I believe we are the first HSphere host out there having the ability to offer our customers offsite backup MX record as part of our default service. It is a shame that HSphere is quite limited in this regards. We simply could not designate one (or more) server as the backup mail server. Rather, what HSphere has is the ability for one mail server to be a backup mail server for the other. However, we insited to have the backup mail server located off-network. The backup mail server is located in our rack in Telehouse, NYC (Manhattan). Thus, in the rare even that our whole Equinix network is down, all emails will still be accepted by the backup mail server in NYC.

John has been the primary person working on this project. Thanks John!

With this, I believe our mail service improvements are completed. Should you have any other suggestions on any improvements we can make, feel free to let us know.

dagta
05-02-2006, 08:13 PM
Thank you very much! I've been very impressed with the improvements announced in this thread and am very happy to be a FluidHosting customer! :)

Trip
05-03-2006, 11:21 PM
Fantastic news, guys! Webhosts are usually not recommend for email hosting is because so many host their email right on the same servers as their webhost accounts, which means when sites go down, bye bye mail. Better webhosts host mail on separate servers, but you guys have seemed to procede a couple steps further still. I'm very comfortable having my forwarding/mail accounts hosted with you guys. Well done!

Trip

FH-John
05-04-2006, 01:46 AM
We aim to provide the best service possible.