We are completing a complete data migration to HPE 3PAR StoreServ Storage

[gtranslate]

Last month we informed you about the successful migration of NoLimit web hosts from old servers to new ones. It was a very important step for us to move forward. As you know, we like a unified solution that we can scale quickly, efficiently and also cheaply.

Migrating databases to NoLimit on 3PAR

Customer files migrated from the old servers directly to our new HPE 3 PAR StoreServ Storage 8450, which we started testing last year.

It is a professional AllFlash (100% SSD) storage that has full redundancy of all components and can do all the advanced features you expect from modern storage designed for datacenters. It was also crucial for us that it can also do synchronous online replication to our second datacenter. Our goal is to keep all data online and up-to-date, in two different datacentres. Thus, if one datacenter goes down, the data will be immediately available (without failure) from the other datacenter. But more about that next time 🙂

And how was the migration? Our hosting databases are replicated in master-slave mode. When something breaks, we just swap the master for the slave and we can solve serious problems very quickly. In addition, there are standard backups.

We took advantage of this when migrating databases to the new 3PAR storage. Master was on the servers and Slave moved to 3PAR. Then all we had to do was switch. No one recognized anything.

The Master-Slave concept has been retained in 3PAR. Currently we have all databases on different repositories (master is on a different one than slave).

We have previously tested the effect on performance. After moving the files to 3PAR, we saw a few percent speedup. Nothing significant, however, as we had very fast NVMe SSDs everywhere. Rather, it was noticeable that 3PAR can better handle the loads and extreme cases. Thus, the occasional spikes were reduced, which had an effect on the average.

We deployed HPE 3PAR in May 2021 on 49 NoLimit servers. The average processing time of a PHP script has been reduced by several percent.

As far as the databases are concerned, both according to the measurements of all sites on WEDOS OnLine and through the log data, the deviation was more or less unmeasurable. More like a green zero. Webs are a few milliseconds faster, but really at the limit of measurability. However, we didn’t even count on the acceleration there. We’ve been monitoring it mainly to make sure everything is working properly.

Migrating VPS ON and WEDOS Cloud

Next up was VPS ON. It was more complicated there. Unlike the NoLimit web hosting, it wasn’t as easy, but we managed to do it for a large part without any downtime. However, outages were necessary for some customers.

So we planned the shutdowns. If a customer did not like the downtime, he could agree on an individual time with the technician. Since there are still a few last VPS ON left to migrate (units of units). We migrated several thousand active and live VPS ON and WEDOS Cloud. We have only had problems with about 1%.

For VPS ON, the HPE 3PAR StoreServ Storage 8450 is more evident. There are more demanding customers out there who can put a lot more strain on the disks. 3PAR will help them in this respect. However, we don’t expect there to be any dramatic increase in disk operations by tens of percent, but it will definitely be a bit faster. Especially in the tips.

At the same time, we will finally be able to offer our customers “giant” storage devices that can have tens of TB of space and will be completely on SSD.

WMS Migration

The WMS migration will be completed by the end of the week. Many of the same things are there as with NoLimit, so we don’t foresee any problems. First we wanted to check everything on NoLimit, after all WMS is an expensive service for demanding customers.

Why we migrate data to HPE 3PAR StoreServ Storage

Currently, our goal is to simplify everything. Rely as little as possible on software, where you often have to find your own solutions, and instead rely on hardware solutions. It takes away a lot of hassle, simplifies development and makes services run faster. Of course, it costs extra.

We want to make as many things as possible “dumb”, things that can work for years without major modifications. Modern things are nice and can do a lot, but they often need to be bent and when something doesn’t work as it should, you have to communicate with the developers. So we want to bend as little as possible, we want simple things that are easy to automate and clear.

HPE 3PAR StoreServ Storage is an example of this. It replaces a software solution that has broken down on occasion, especially when unexpected events have occurred.

We currently have 6 units of HPE 3PAR StoreServ Storage 8450 in production, with more in testing and development. We have 12 more on order.

One of the first HPE 3PAR StoreServ Storage 8450 in our WEDOS DC 1 datacenter
And here you can see how they look in our rack in WEDOS DC 1.

There’s a lot more to why we chose HPE 3PAR StoreServ Storage.

  • The whole solution is very easy to scale.
  • It is fully redundant (2 nodes and each with a dual controller, plus online synchronous replication in a second datacenter) .
  • High performance (million IOPS and tens of Gbps throughput)
  • Fast backup and restore functions are available. There you will find deduplication, compression, encryption, snapshots and many other gadgets…
  • The solution is ready to run in multiple copies in different geographical locations, which we will need.

And this is the most important….

We can drown it in oil 🙂

The design of the HPE 3PAR StoreServ Storage 8450 allows it to be used in our oil baths. Of course, we won’t be able to use spinning disks, but we don’t use them for most of our services anyway (the exceptions are dedicated servers and old VPS, which can’t be bought anymore). We use server SSDs everywhere, which we buy by the pallet because of the huge consumption. No they’re not breaking, we’re just growing and we need them for new servers … Uh, actually, now we need new 3PARs.)

In the next few days we will probably conduct another experiment with drowning the entire storage in an oil bath. We expect to save mainly on electricity bills and, of course, to be able to use waste heat.

So you have something to look forward to 🙂