Thursday, January 12, 2017

M.2 SSDs in PCIe slots, with "SilverStone Dual M2 to PCI-E X4 and SATA 6G Adapter Card (ECM20)"

Disclaimer, when there are Amazon links, I've used my affiliate ID; if you buy something using the link, I should get some minimal %.

Overview of my homelab:


My approach to homelabs was born from my love of gaming hardware. If I'm spending money on hardware, it better run everything I throw at it, including PC games, real good. This is typically not hard - a good CPU, max RAM and a good video card, with a power supply that can last for years and multiple expansions will get you far. This may be why I don't like Apple hardware, which typically gives you less hardware for the same dollar amount, with the excuse of "but the software is so good".

My last build was in 2011, and almost 7 years later, it's still running pretty well. Ever since I got it, I went max RAM. I have added a PCIe enterprise SSD (FusionIO doesn't show up in Windows, you will see it in the actual pics) and changed SSDs as they have come down in price.



32GBs are overkill for gaming but for VMware homelabs they were adequate in 2011 and they are the bare minimum today - RAM is the main limiting factor when running VMs. 32GB is what you can get from an Intel NUC, one of the current homelab darlings, with a normal desktop's noise levels. As you can see I have it connected to a TV, using a microsoft wireless mouse/keyboard.

M.2 SSDs

The M.2 form factor works for both SATA and NVMe SSDs. The important thing is the key, which can B (sata) or M (PCIe, used with NVMe). It's wikipedia article is a good read. M.2 SSDs are becoming more popular since they are smaller than a 2.5 drive and NVMe is much faster than SATA; but using up a PCIe slot for one storage device can be expensive and consumer/server boards normally don't have that many PCIe slots. Some boards are putting M.2 slots directly on the motherboard, both in lay-flat and vertical positions. I've seen server boards that treat them like DIMMs.

Most M.2 you would consider come in the 2280 size, but enterprise SSDs will most likely come in the 22110 size.

Selecting a M.2 PCIe adapter card

I'm a big fan of Paul Braren's TinkerTry.com site. I've called him "the God of homelabs" and I mean that - he has spent considerable time blogging in the prosumer segment of homelabs, investigating and publishing particularly useful tips for VMware homelab enthusiasts. I'm a big proponent of his SuperMicro bundles - I especially like that he always thinks of them in multiboot scenarios so they can perform more than just homelab functions. SuperMicro offers much better specs than any gaming rig you can build, since they are server boards, allowing for 128GB of RAM and on-board 10G NICs. VMware recently became his employer, which is great news for everybody, and bodes for some good VSAN content, which is known for a difficult HCL for homelabs. Don't miss out on his SuperMicro vs NUC comparison chart.

Paul did an excellent article showing PCIe adapter cards for NVMe M.2 SSDs. Ideally, if you are going to put an adapter card, you would use a x16 slot and put 4 NVME SSDs on it; using up a slot for just one drive is ok until you need the slot, so it's better to plan ahead. However, the reviewed card that can do multiple NVMe SSDs is really expensive right now, something I can't explain since in theory it's only routing the PCIe lanes and providing power. It does mark all checkboxes though - supports 22110 drives and 4 M.2 NVMe SSDs at the same time; it's the only card we've all found that can do this.

There's a cheap 4 drive card but this only works for mSATA SSDs, which are not NVMe - this is the card featured on the AzureStack mini server, the Addonics AD4MSPX2-A. If you already have mSATA SSDs this is the card to get.

SilverStone Dual M.2 (NVMe & SATA)

While checking around for a multiple NVME card I found this card



1) has two 2280 m.2 slots, one in SATA and another in NVMe, able to run both M.2 simultaneously, 
2) it's cheap, $18 last time I checked
3) it's just an adapter, no drivers or interfering silicon, using a x4 PCIe slot

This gave me the sweet spot that I was looking for, multiple M.2 slots while being cheap and simple. While I won't be able to run 22110 M.2 SSDs, I most likely won't be able to afford them anyways ;)

While waiting on a good M.2 NVMe drive to drop down in price I decided to get a cheap SATA B key M.2 SSD. I found a $100 512GB refurbished Samsung 850 EVO, which is a very good drive for cheap, and ordered it, obviously a little wary.

Good value SSD options I can recommend today:

NVME: the Samsung are probably the best performing but I can't justify the cost difference, my choice would be Intel 600p


Once everything arrived I took some pictures to share:




 2280 is pretty small!







Notice the max size is 2280 for the M.2 SSD


I only had either x1 or x16- used the last free PCIE X16 slot



Notice the card holds and powers the SATA SSD only, you still need to connect a SATA cable to it

At least the SMART data for the drive shows it healthy, and it has been fine so far

With this, once I find a good value on a 1TB NVMe SSD I can add it to the same card and get another year or so of use from this motherboard. This is a card I'm pretty sure I can move to my next build, hopefully on a x4 slot. The next build will very likely use NVMe only, and will most probably have slots for 3DXP/Optane; such a build will take at least another year to reach my budget projections.

Feel free to ask me questions on this build on twitter.

Monday, November 21, 2016

#vDM30in30 11-12-2016 IT certifications - my 2016 results so far

I realize this post is not something to proudly announce to the world, but I also feel that I get enough questions from people just starting out that I thought I would share my experience - if nothing else, so others can learn from it :)

So far since I started with the goal of attaining IT certifications, I've only been able to successfully pass one relatively tough cert per year - and all have been VMware certifications (VCA doesn't count, that was a free voucher for an entry-level exam). 




When I started this year, I really wanted to "level up" my resume in regards to IT certifications. When I planned out 2016 back in January, I wanted to achieve at least two more VMware certifications, a Cisco certification, and a Linux Foundation certification. You can see the original plan from this post (which will be edited for 2017 in January again).

Results so far

The results so far fall in three categories:

1) Failed to even sit

There was one exam that, although I tried to study, I failed to cover the material in time and thus didn't even schedule it before the version was upgraded:

August 20, 2016 (last date exam was active) - Cisco CCNA 200-120 - only covered 30% of the material by exam due date - did not sit.

2) Sat (with mixed results)

I've sat three exams this year - all were taken at that time because of the discount offered, not because I had actually prepared properly... And you can tell they were taken at different times from my original plans:

June 26, 2016 - VCAP6-DCV deploy exam - half price during beta of $200 - passed!
August 29, 2016 - VCP6-DCV in VMworld 2016 - half price during VMworld of $113 - failed! 
September 22, 2016 - VCP6-NX for NSX 6.2  - super discount during beta of $50 - still hasn't been scored but I'm not sure I will pass it.

3) Still have to sit

In Black Friday 2015 I purchased a Linux Foundation bundle that included video training and the exam voucher. Last date I can take the exam for it to be valid is December 1st (in roughly a week). I just remembered that I hadn't scheduled it yet, so I've just done so



To note - my current employer does not provide incentives or pre-pay for certifications exams taken, so all risk is mine if I don't pass an exam. I'm almost sure that they will pay me back for a passed VMware exam since that is my main job - but that is the reason why I've taken exams in these dates, rather than actually preparing and paying for them until I felt ready, like I did my first few ones.

Actually, that's not entirely true. The VCAP5-DCD I was halfway through preparation when a voucher fell into my lap, to take the exam that week. I was very lucky to pass that one. And the LFCS exam, I've had a full year to prepare for, so there's no excuse there.

Analysis

(ノ≧∇≦)ノ ミ ┸━┸

I think anyone can agree that my results have been terrible so far. There was bad planning, bad execution, and definitely no real commitment. With the excuse of "reach for the starts, if you fail you will land in the clouds" I've basically spent more money than I could have, while attaining only one new certification for now.

Clearly the time to take cheap VMware certification exams is around fall. I spent valuable time studying for CCNA in the summer that I should have focused on preparing for the VCP6-DCV exam that I failed (from overconfidence and little preparation). I should have tackled the CCNA in the first few months of the year, which was going to be the same cost no matter when I did it, and then focused on the VMware certs.

What was different with the exam I did pass? One reason why I passed the VCAP6 Deploy is that it maps much closer to my day-to-day job. This was an exam that I could lab, there were active HOLs that I could use to practice, and that tests tasks which I'm very familiar with. The rest of the exams, including several topics of the VCP which are not in use in my job, don't map as well to my job description and thus don't get the benefit of me needing and using the knowledge 8 hours a day.

For the VCP6-NV beta, I won't feel bad if I miss it. I took it on very short notice because the price was cheap (even for a beta), I have genuine interest in NSX, and I had just bought Elver's excellent study guide, which had been released during VMworld. I didn't have enough time to prepare for it, but I can live with a fail that pushed me to study and cost only $50, as this is an exam that I do want to pass for the long term.

Next steps

Clearly I have to invest this remaining week on my LFCS studies so that I don't add another failure to these results. I believe that the future is in Linux/BSD and that if I don't address this ASAP my relevance as an IT professional will go down.

I've already committed publicly to doing VCDX5-DCV by March, and I'm coming to the realization that I'm probably setting myself up for failure. If I'm going to submit a VCDX5 design document that I'm not ashamed of, I need to find a way to concentrate and dedicate hours to it - much more than I've been doing for everything else. That submission is $1200 or so by the way - not something to do halfheartedly.

If I don't pass the VCP6-NV, I will still want to do it; and I still want the CCNA and VCP6-DCV in 2017. The fact of the matter is that I won't feel satisfied until I pass these exams, since I don't feel I'm very far away and they are still relevant to me.